You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by tq...@apache.org on 2022/04/16 00:34:18 UTC
[tvm-site] branch asf-site updated: deploying docs (apache/tvm@fafabc96c1ba1a5f987c2402fcc2ce4d1bad5cc8)

This is an automated email from the ASF dual-hosted git repository.

tqchen pushed a commit to branch asf-site
in repository https://gitbox.apache.org/repos/asf/tvm-site.git


The following commit(s) were added to refs/heads/asf-site by this push:
     new 13e368c21 deploying docs (apache/tvm@fafabc96c1ba1a5f987c2402fcc2ce4d1bad5cc8)
13e368c21 is described below

commit 13e368c216a711610049a7f6fccafe3acc33db8d
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Sat Apr 16 00:34:13 2022 +0000

    deploying docs (apache/tvm@fafabc96c1ba1a5f987c2402fcc2ce4d1bad5cc8)
---
 .../how_to/compile_models/from_mxnet.rst.txt       |    2 +-
 .../how_to/compile_models/from_paddle.rst.txt      |    2 +-
 .../how_to/compile_models/from_pytorch.rst.txt     |    2 +-
 .../compile_models/sg_execution_times.rst.txt      |   20 +-
 .../deploy_models/deploy_model_on_android.rst.txt  |    2 +-
 .../deploy_object_detection_pytorch.rst.txt        |    4 +-
 .../deploy_models/deploy_prequantized.rst.txt      |    6 +-
 .../deploy_prequantized_tflite.rst.txt             |    4 +-
 .../how_to/deploy_models/deploy_quantized.rst.txt  |    2 +-
 .../deploy_models/deploy_ssd_gluoncv.rst.txt       |    4 +-
 .../deploy_models/sg_execution_times.rst.txt       |   18 +-
 .../extend_tvm/bring_your_own_datatypes.rst.txt    |    2 +-
 .../how_to/extend_tvm/sg_execution_times.rst.txt   |   10 +-
 .../how_to/extend_tvm/use_pass_instrument.rst.txt  |   16 +-
 .../optimize_operators/opt_conv_cuda.rst.txt       |    2 +-
 .../optimize_operators/opt_conv_tensorcore.rst.txt |    2 +-
 .../how_to/optimize_operators/opt_gemm.rst.txt     |   16 +-
 .../optimize_operators/sg_execution_times.rst.txt  |    8 +-
 .../sg_execution_times.rst.txt                     |   16 +-
 .../tune_conv2d_layer_cuda.rst.txt                 |  790 ++++++++++---
 .../tune_network_cuda.rst.txt                      |    2 +-
 .../tune_network_x86.rst.txt                       |    4 +-
 .../tune_sparse_x86.rst.txt                        | 1175 +++++++-------------
 .../tune_with_autotvm/sg_execution_times.rst.txt   |   12 +-
 .../tune_with_autotvm/tune_conv2d_cuda.rst.txt     |   34 +-
 .../work_with_microtvm/micro_autotune.rst.txt      |   16 +-
 .../work_with_microtvm/sg_execution_times.rst.txt  |   10 +-
 .../work_with_relay/sg_execution_times.rst.txt     |    8 +-
 .../work_with_schedules/sg_execution_times.rst.txt |   18 +-
 .../how_to/work_with_schedules/tensorize.rst.txt   |    2 +-
 .../tutorials/autotvm/sg_execution_times.rst.txt   |    6 +-
 .../frontend/deploy_classification.rst.txt         |    2 +-
 .../tutorials/frontend/deploy_detection.rst.txt    |    2 +-
 .../tutorials/frontend/sg_execution_times.rst.txt  |    6 +-
 .../tutorials/optimize/sg_execution_times.rst.txt  |    6 +-
 .../topic/vta/tutorials/sg_execution_times.rst.txt |    6 +-
 .../tutorial/auto_scheduler_matmul_x86.rst.txt     |    2 +-
 docs/_sources/tutorial/autotvm_relay_x86.rst.txt   |   56 +-
 .../tutorial/cross_compilation_and_rpc.rst.txt     |    2 +-
 docs/_sources/tutorial/intro_topi.rst.txt          |    2 +-
 docs/_sources/tutorial/sg_execution_times.rst.txt  |   26 +-
 .../tutorial/tensor_expr_get_started.rst.txt       |   49 +-
 docs/commit_hash                                   |    2 +-
 docs/how_to/compile_models/from_mxnet.html         |    2 +-
 docs/how_to/compile_models/from_paddle.html        |    2 +-
 docs/how_to/compile_models/from_pytorch.html       |    8 +-
 docs/how_to/compile_models/sg_execution_times.html |   20 +-
 .../deploy_models/deploy_model_on_android.html     |    2 +-
 .../deploy_object_detection_pytorch.html           |   19 +-
 docs/how_to/deploy_models/deploy_prequantized.html |    6 +-
 .../deploy_models/deploy_prequantized_tflite.html  |    4 +-
 docs/how_to/deploy_models/deploy_quantized.html    |    2 +-
 docs/how_to/deploy_models/deploy_ssd_gluoncv.html  |   41 +-
 docs/how_to/deploy_models/sg_execution_times.html  |   18 +-
 .../extend_tvm/bring_your_own_datatypes.html       |    2 +-
 docs/how_to/extend_tvm/sg_execution_times.html     |   10 +-
 docs/how_to/extend_tvm/use_pass_instrument.html    |   16 +-
 docs/how_to/optimize_operators/opt_conv_cuda.html  |    2 +-
 .../optimize_operators/opt_conv_tensorcore.html    |    2 +-
 docs/how_to/optimize_operators/opt_gemm.html       |   16 +-
 .../optimize_operators/sg_execution_times.html     |    8 +-
 .../sg_execution_times.html                        |   14 +-
 .../tune_conv2d_layer_cuda.html                    |  790 ++++++++++---
 .../tune_with_autoscheduler/tune_network_cuda.html |    2 +-
 .../tune_with_autoscheduler/tune_network_x86.html  |    4 +-
 .../tune_with_autoscheduler/tune_sparse_x86.html   | 1175 +++++++-------------
 .../tune_with_autotvm/sg_execution_times.html      |   12 +-
 .../how_to/tune_with_autotvm/tune_conv2d_cuda.html |   34 +-
 docs/how_to/work_with_microtvm/micro_autotune.html |   16 +-
 .../work_with_microtvm/sg_execution_times.html     |   10 +-
 .../how_to/work_with_relay/sg_execution_times.html |    8 +-
 .../work_with_schedules/sg_execution_times.html    |   18 +-
 docs/how_to/work_with_schedules/tensorize.html     |    2 +-
 .../api/doxygen/classtvm_1_1LinkedParamNode.html   |    2 +-
 .../classtvm_1_1LinkedParamNode__coll__graph.svg   |  160 +--
 .../doxygen/classtvm_1_1relay_1_1ConstantNode.html |    2 +-
 ...sstvm_1_1relay_1_1ConstantNode__coll__graph.svg |  360 +++---
 .../classtvm_1_1runtime_1_1NDArray-members.html    |   16 +-
 .../doxygen/classtvm_1_1runtime_1_1NDArray.html    |   91 +-
 ...classtvm_1_1runtime_1_1NDArray__coll__graph.svg |  142 +--
 ...sstvm_1_1runtime_1_1NDArray__inherit__graph.svg |  104 +-
 docs/reference/api/doxygen/functions_f.html        |    7 +-
 docs/reference/api/doxygen/functions_func_f.html   |    5 +-
 docs/reference/api/doxygen/functions_func_n.html   |    3 +
 docs/reference/api/doxygen/functions_func_s.html   |    2 +-
 docs/reference/api/doxygen/functions_func_t.html   |    6 +-
 docs/reference/api/doxygen/functions_n.html        |    3 +
 docs/reference/api/doxygen/functions_s.html        |   10 +-
 docs/reference/api/doxygen/functions_t.html        |    4 +-
 docs/reference/api/doxygen/functions_v.html        |    4 +-
 docs/reference/api/doxygen/ndarray_8h_source.html  |   48 +-
 .../api/doxygen/packed__func_8h_source.html        |   10 +-
 docs/reference/api/doxygen/search/all_11.js        |    2 +-
 docs/reference/api/doxygen/search/all_13.js        |    4 +-
 docs/reference/api/doxygen/search/all_14.js        |   10 +-
 docs/reference/api/doxygen/search/all_15.js        |    4 +-
 docs/reference/api/doxygen/search/all_16.js        |    2 +-
 docs/reference/api/doxygen/search/all_17.js        |    2 +-
 docs/reference/api/doxygen/search/all_18.js        |    2 +-
 docs/reference/api/doxygen/search/all_7.js         |    1 +
 docs/reference/api/doxygen/search/all_e.js         |    4 +-
 docs/reference/api/doxygen/search/all_f.js         |    1 +
 docs/reference/api/doxygen/search/functions_10.js  |    2 +-
 docs/reference/api/doxygen/search/functions_13.js  |    6 +-
 docs/reference/api/doxygen/search/functions_14.js  |    2 +-
 docs/reference/api/doxygen/search/functions_15.js  |    2 +-
 docs/reference/api/doxygen/search/functions_16.js  |    2 +-
 docs/reference/api/doxygen/search/functions_6.js   |    1 +
 docs/reference/api/doxygen/search/functions_d.js   |    4 +-
 docs/reference/api/doxygen/search/functions_e.js   |    1 +
 .../api/doxygen/structural__hash_8h_source.html    |    2 +-
 docs/reference/api/python/auto_scheduler.html      |    4 +-
 .../api/typedoc/classes/bytestreamreader.html      |   12 +-
 .../api/typedoc/classes/cachedcallstack.html       |   34 +-
 docs/reference/api/typedoc/classes/dldatatype.html |   12 +-
 docs/reference/api/typedoc/classes/dldevice.html   |   10 +-
 .../reference/api/typedoc/classes/environment.html |   12 +-
 docs/reference/api/typedoc/classes/ffilibrary.html |   20 +-
 .../api/typedoc/classes/graphexecutor.html         |   16 +-
 docs/reference/api/typedoc/classes/instance.html   |   40 +-
 docs/reference/api/typedoc/classes/memory.html     |   34 +-
 docs/reference/api/typedoc/classes/module.html     |   10 +-
 docs/reference/api/typedoc/classes/ndarray.html    |   22 +-
 .../api/typedoc/classes/packedfunccell.html        |    6 +-
 docs/reference/api/typedoc/classes/rpcserver.html  |   14 +-
 docs/reference/api/typedoc/classes/scalar.html     |    6 +-
 .../api/typedoc/classes/webgpucontext.html         |   12 +-
 docs/reference/api/typedoc/enums/argtypecode.html  |   30 +-
 .../api/typedoc/enums/aynccallbackcode.html        |    4 +-
 .../api/typedoc/enums/dldatatypecode.html          |    8 +-
 .../api/typedoc/enums/rpcserverstate.html          |   12 +-
 docs/reference/api/typedoc/enums/sizeof.html       |   18 +-
 docs/reference/api/typedoc/index.html              |  112 +-
 .../api/typedoc/interfaces/disposable.html         |    2 +-
 .../api/typedoc/interfaces/functioninfo.html       |    6 +-
 .../api/typedoc/interfaces/libraryprovider.html    |    4 +-
 docs/searchindex.js                                |    2 +-
 .../vta/tutorials/autotvm/sg_execution_times.html  |    6 +-
 .../tutorials/frontend/deploy_classification.html  |    2 +-
 .../vta/tutorials/frontend/deploy_detection.html   |    2 +-
 .../vta/tutorials/frontend/sg_execution_times.html |    6 +-
 .../vta/tutorials/optimize/sg_execution_times.html |    6 +-
 docs/topic/vta/tutorials/sg_execution_times.html   |    6 +-
 docs/tutorial/auto_scheduler_matmul_x86.html       |    2 +-
 docs/tutorial/autotvm_relay_x86.html               |  172 +--
 docs/tutorial/cross_compilation_and_rpc.html       |    2 +-
 docs/tutorial/intro_topi.html                      |    2 +-
 docs/tutorial/sg_execution_times.html              |   26 +-
 docs/tutorial/tensor_expr_get_started.html         |   45 +-
 149 files changed, 3329 insertions(+), 3084 deletions(-)

diff --git a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
index f2cc5fde3..b33aec6fc 100644
--- a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
@@ -98,7 +98,7 @@ In this section, we download a pretrained imagenet model and classify an image.
 
  .. code-block:: none
 
-    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip67db9577-e96a-4c5c-8273-04062e07c5aa from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip6d3e2761-0eae-4f2a-aa04-e3b1d5eca3da from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
     x (1, 3, 224, 224)
 
 
diff --git a/docs/_sources/how_to/compile_models/from_paddle.rst.txt b/docs/_sources/how_to/compile_models/from_paddle.rst.txt
index a0c9134e3..ed33ca0ac 100644
--- a/docs/_sources/how_to/compile_models/from_paddle.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_paddle.rst.txt
@@ -201,7 +201,7 @@ Look up prediction top 1 index in 1000 class synset.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  4.500 seconds)
+   **Total running time of the script:** ( 1 minutes  3.900 seconds)
 
 
 .. _sphx_glr_download_how_to_compile_models_from_paddle.py:
diff --git a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
index ab8075dae..1f81d9bd7 100644
--- a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
@@ -79,7 +79,7 @@ Load a pretrained PyTorch model
  .. code-block:: none
 
     Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /workspace/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
-
      0%|          | 0.00/44.7M [00:00<?, ?B/s]
      1%|1         | 472k/44.7M [00:00<00:09, 4.80MB/s]
     10%|#         | 4.68M/44.7M [00:00<00:01, 27.9MB/s]
     70%|#######   | 31.4M/44.7M [00:00<00:00, 143MB/s] 
    100%|##########| 44.7M/44.7M [00:00<00:00, 135MB/s]
+
      0%|          | 0.00/44.7M [00:00<?, ?B/s]
      7%|6         | 3.05M/44.7M [00:00<00:01, 32.0MB/s]
     17%|#7        | 7.61M/44.7M [00:00<00:00, 41.0MB/s]
     69%|######8   | 30.7M/44.7M [00:00<00:00, 133MB/s] 
    100%|##########| 44.7M/44.7M [00:00<00:00, 131MB/s]
 
 
 
diff --git a/docs/_sources/how_to/compile_models/sg_execution_times.rst.txt b/docs/_sources/how_to/compile_models/sg_execution_times.rst.txt
index 109b28e41..93a7a29b4 100644
--- a/docs/_sources/how_to/compile_models/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/compile_models/sg_execution_times.rst.txt
@@ -5,14 +5,14 @@
 
 Computation times
 =================
-**04:39.864** total execution time for **how_to_compile_models** files:
+**04:38.468** total execution time for **how_to_compile_models** files:
 
-- **01:04.500**: :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)
-- **00:59.553**: :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``)
-- **00:55.423**: :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)
-- **00:25.357**: :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)
-- **00:20.788**: :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)
-- **00:20.742**: :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)
-- **00:18.969**: :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)
-- **00:12.078**: :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)
-- **00:02.454**: :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)
+- **01:03.900**: :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)
+- **00:59.085**: :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``)
+- **00:55.277**: :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)
+- **00:25.181**: :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)
+- **00:20.886**: :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)
+- **00:20.807**: :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)
+- **00:18.207**: :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)
+- **00:12.435**: :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)
+- **00:02.691**: :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)
diff --git a/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt b/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt
index 40ab19085..dff75d791 100644
--- a/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt
@@ -393,7 +393,7 @@ Execute on TVM
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      15.6234      15.6349      15.7135      15.5296       0.0604   
+      15.9951      15.9650      16.4304      15.8805       0.1509   
                
 
 
diff --git a/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt b/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt
index 51ba210d3..8635eca43 100644
--- a/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt
@@ -108,7 +108,7 @@ Load pre-trained maskrcnn from torchvision and do tracing
  .. code-block:: none
 
     Downloading: "https://download.pytorch.org/models/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth" to /workspace/.cache/torch/hub/checkpoints/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth
-
      0%|          | 0.00/170M [00:00<?, ?B/s]
     10%|#         | 17.8M/170M [00:00<00:00, 186MB/s]
     25%|##4       | 41.7M/170M [00:00<00:00, 224MB/s]
     39%|###8      | 65.6M/170M [00:00<00:00, 236MB/s]
     52%|#####2    | 88.3M/170M [00:00<00:00, 236MB/s]
     65%|######5   | 111M/170M [00:00<00:00, 166MB/s] 
     78%|#######8  | 133M/170M [00:00<00:00, 183MB/s]
     92%|#########2| 157M/170M [00:00<00:00, 202MB/s]
    100%|##########| 170M/170M [00:00<00:00, 204MB/s]
+
      0%|          | 0.00/170M [00:00<?, ?B/s]
      3%|2         | 4.52M/170M [00:00<00:03, 47.3MB/s]
      6%|5         | 9.66M/170M [00:00<00:03, 51.2MB/s]
     20%|##        | 34.1M/170M [00:00<00:00, 145MB/s] 
     36%|###5      | 60.7M/170M [00:00<00:00, 198MB/s]
     51%|#####1    | 87.4M/170M [00:00<00:00, 227MB/s]
     67%|######6   | 114M/170M [00:00<00:00, 243MB/s] 
     82%|########2 | 139M/170M [00:00<00:00, 252MB/s]
     98%|#########7| 166M/170M [00:00<00:00, 261MB/s]
    100%|##########| 170M/170M [00:00<00:00, 217MB/s]
     /usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:3878: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
       for i in range(dim)
     /usr/local/lib/python3.7/dist-packages/torchvision/models/detection/anchor_utils.py:127: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the 'trunc' function NOT 'floor'). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode='trunc'), or for actual floor division, use torch.div(a, b, rounding_mode='floor').
@@ -253,7 +253,7 @@ Get boxes with score larger than 0.9
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 2 minutes  58.290 seconds)
+   **Total running time of the script:** ( 3 minutes  0.840 seconds)
 
 
 .. _sphx_glr_download_how_to_deploy_models_deploy_object_detection_pytorch.py:
diff --git a/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt b/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
index da37e7c3f..0e5e74bab 100644
--- a/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
@@ -187,7 +187,7 @@ training. Other models require a full post training calibration.
  .. code-block:: none
 
     Downloading: "https://download.pytorch.org/models/mobilenet_v2-b0353104.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2-b0353104.pth
-
      0%|          | 0.00/13.6M [00:00<?, ?B/s]
    100%|##########| 13.6M/13.6M [00:00<00:00, 149MB/s]
+
      0%|          | 0.00/13.6M [00:00<?, ?B/s]
    100%|##########| 13.6M/13.6M [00:00<00:00, 185MB/s]
 
 
 
@@ -344,7 +344,7 @@ Here we give an example of how to measure performance of TVM compiled models.
 
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      90.1806      90.1495      91.6179      89.9756       0.1878   
+      90.1193      90.0943      91.0812      89.8953       0.1627   
                
 
 
@@ -384,7 +384,7 @@ TODO
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  3.795 seconds)
+   **Total running time of the script:** ( 1 minutes  4.586 seconds)
 
 
 .. _sphx_glr_download_how_to_deploy_models_deploy_prequantized.py:
diff --git a/docs/_sources/how_to/deploy_models/deploy_prequantized_tflite.rst.txt b/docs/_sources/how_to/deploy_models/deploy_prequantized_tflite.rst.txt
index 2196766b3..a5f3881ad 100644
--- a/docs/_sources/how_to/deploy_models/deploy_prequantized_tflite.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_prequantized_tflite.rst.txt
@@ -351,7 +351,7 @@ Here we give an example of how to measure performance of TVM compiled models.
 
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      118.9514     118.9186     121.8738     118.0731      0.4844   
+      119.4266     119.4160     120.6976     118.3389      0.4110   
                
 
 
@@ -385,7 +385,7 @@ Here we give an example of how to measure performance of TVM compiled models.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  57.873 seconds)
+   **Total running time of the script:** ( 1 minutes  54.100 seconds)
 
 
 .. _sphx_glr_download_how_to_deploy_models_deploy_prequantized_tflite.py:
diff --git a/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt b/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
index 0f43d247e..596872e48 100644
--- a/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
@@ -221,7 +221,7 @@ We create a Relay VM to build and execute the model.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  9.391 seconds)
+   **Total running time of the script:** ( 1 minutes  9.426 seconds)
 
 
 .. _sphx_glr_download_how_to_deploy_models_deploy_quantized.py:
diff --git a/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt b/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt
index 50fbb77b7..4baebef85 100644
--- a/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt
@@ -137,7 +137,7 @@ Convert and compile model for CPU.
             data: None
       input_sym_arg_type = in_param.infer_type()[0]
     Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
-
      0%|          | 0/132723 [00:00<?, ?KB/s]
      2%|1         | 2122/132723 [00:00<00:06, 21216.88KB/s]
      5%|5         | 7018/132723 [00:00<00:03, 37533.45KB/s]
     11%|#1        | 14657/132723 [00:00<00:02, 55273.91KB/s]
     17%|#6        | 22292/132723 [00:00<00:01, 63591.79KB/s]
     23%|##2       | 29989/132723 [00:00<00:01, 68413.06KB/s]
     28%|##7       | 36831/132723 [00:00<00:01, 63368.68KB/s]
     34%|###3      | 44568/132723 [00:00<00:01, 67707.28KB/s]
     39%|###9      | 52285/132723 [00:00<00:01, 70612.24KB/s]
     45%|####5     | 60150/132723 [00:00<00:00, 73061.44KB/s]
     51%|#####1    | 67967/132723 [00:01<00:00, 74605.57KB/s]
     57%|#####7    | 75785/132723 [00:01<00:00, 75683.36KB/s]
     63%|######2   | 83613/132723 [00:01<00:00, 76462.65KB/s]
     69%|######8   | 91424/132723 [00:01<00:00, 76956.22KB/s]
     75%|#######4  | 99307/132723 [00:01<00:00, 77518.30KB/s]
     81%|########  | 107068/132723 [00:01<00:00, 76879.52KB/s]
     86%|########6 |
  114764/132723 [00:01<00:00, 65402.06KB/s]
     92%|#########2| 122373/132723 [00:01<00:00, 68244.07KB/s]
     98%|#########8| 130274/132723 [00:01<00:00, 71208.53KB/s]
    100%|##########| 132723/132723 [00:01<00:00, 69214.61KB/s]
+
      0%|          | 0/132723 [00:00<?, ?KB/s]
      3%|3         | 4011/132723 [00:00<00:03, 40104.97KB/s]
      9%|8         | 11818/132723 [00:00<00:01, 62429.31KB/s]
     15%|#4        | 19487/132723 [00:00<00:01, 68937.87KB/s]
     20%|#9        | 26381/132723 [00:00<00:01, 64125.14KB/s]
     25%|##4       | 32841/132723 [00:00<00:01, 51078.33KB/s]
     30%|##9       | 39514/132723 [00:00<00:01, 55473.10KB/s]
     36%|###5      | 47266/132723 [00:00<00:01, 61784.39KB/s]
     41%|####      | 54021/132723 [00:00<00:01, 63440.57KB/s]
     47%|####6     | 62075/132723 [00:00<00:01, 68450.30KB/s]
     52%|#####2    | 69109/132723 [00:01<00:00, 67496.95KB/s]
     57%|#####7    | 75990/132723 [00:01<00:01, 53175.22KB/s]
     62%|######1   | 81847/132723 [00:01<00:00, 51028.31KB/s]
     66%|######5   | 87318/132723 [00:01<00:00, 46698.56KB/s]
     72%|#######1  | 95496/132723 [00:01<00:00, 55166.11KB/s]
     76%|#######6  | 101424/132723 [00:01<00:00, 55132.32KB/s]
     81%|########  
 | 107224/132723 [00:01<00:00, 51928.00KB/s]
     86%|########6 | 114680/132723 [00:02<00:00, 50506.20KB/s]
     90%|######### | 119903/132723 [00:02<00:00, 46107.30KB/s]
     96%|#########6| 128043/132723 [00:02<00:00, 54593.47KB/s]
    100%|##########| 132723/132723 [00:02<00:00, 56199.73KB/s]
 
 
 
@@ -202,7 +202,7 @@ Display result
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 2 minutes  21.103 seconds)
+   **Total running time of the script:** ( 2 minutes  21.536 seconds)
 
 
 .. _sphx_glr_download_how_to_deploy_models_deploy_ssd_gluoncv.py:
diff --git a/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt b/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt
index 4182e629a..ac0e79099 100644
--- a/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt
@@ -5,13 +5,13 @@
 
 Computation times
 =================
-**10:19.589** total execution time for **how_to_deploy_models** files:
+**10:19.507** total execution time for **how_to_deploy_models** files:
 
-- **02:58.290**: :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``)
-- **02:21.103**: :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)
-- **01:57.873**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)
-- **01:09.391**: :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)
-- **01:03.795**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)
-- **00:27.182**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)
-- **00:21.776**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)
-- **00:00.179**: :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)
+- **03:00.840**: :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``)
+- **02:21.536**: :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)
+- **01:54.100**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)
+- **01:09.426**: :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)
+- **01:04.586**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)
+- **00:27.722**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)
+- **00:21.110**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)
+- **00:00.188**: :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)
diff --git a/docs/_sources/how_to/extend_tvm/bring_your_own_datatypes.rst.txt b/docs/_sources/how_to/extend_tvm/bring_your_own_datatypes.rst.txt
index 10fad31f4..a7463d024 100644
--- a/docs/_sources/how_to/extend_tvm/bring_your_own_datatypes.rst.txt
+++ b/docs/_sources/how_to/extend_tvm/bring_your_own_datatypes.rst.txt
@@ -423,7 +423,7 @@ First let us define two helper functions to get the mobilenet model and a cat im
 
  .. code-block:: none
 
-    Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip87c00332-7f47-42da-8a75-970612d26974 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+    Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip1c0edb4c-2721-4485-a82f-207087b6065e from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
 
 
 
diff --git a/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt b/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt
index 8a0caa413..123d8f362 100644
--- a/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt
@@ -5,9 +5,9 @@
 
 Computation times
 =================
-**00:37.555** total execution time for **how_to_extend_tvm** files:
+**00:37.772** total execution time for **how_to_extend_tvm** files:
 
-- **00:34.169**: :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``)
-- **00:02.187**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)
-- **00:01.007**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)
-- **00:00.192**: :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)
+- **00:34.314**: :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``)
+- **00:02.210**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)
+- **00:01.048**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)
+- **00:00.200**: :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)
diff --git a/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt b/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt
index c79251d9b..cc6b5bdc2 100644
--- a/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt
+++ b/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt
@@ -199,10 +199,10 @@ profile the execution time of each passes.
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 5881us [5881us] (44.91%; 44.91%)
-    FoldScaleAxis: 7213us [2us] (55.09%; 55.09%)
-            FoldConstant: 7211us [1499us] (55.07%; 99.97%)
-                    InferType: 5712us [5712us] (43.62%; 79.21%)
+    InferType: 5825us [5825us] (44.98%; 44.98%)
+    FoldScaleAxis: 7124us [2us] (55.02%; 55.02%)
+            FoldConstant: 7122us [1502us] (55.00%; 99.97%)
+                    InferType: 5620us [5620us] (43.40%; 78.92%)
 
 
 
@@ -239,10 +239,10 @@ Refer to following sections and :py:func:`tvm.instrument.pass_instrument` for th
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 5735us [5735us] (44.33%; 44.33%)
-    FoldScaleAxis: 7201us [2us] (55.67%; 55.67%)
-            FoldConstant: 7199us [1533us] (55.65%; 99.97%)
-                    InferType: 5666us [5666us] (43.80%; 78.70%)
+    InferType: 5729us [5729us] (44.57%; 44.57%)
+    FoldScaleAxis: 7124us [2us] (55.43%; 55.43%)
+            FoldConstant: 7122us [1488us] (55.41%; 99.98%)
+                    InferType: 5634us [5634us] (43.83%; 79.10%)
 
 
 
diff --git a/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt b/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt
index 95acc2dec..34f06eceb 100644
--- a/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt
@@ -295,7 +295,7 @@ latency of convolution.
 
  .. code-block:: none
 
-    Convolution: 33.608237 ms
+    Convolution: 44.464928 ms
 
 
 
diff --git a/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt b/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt
index 2839c7f3e..383641187 100644
--- a/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt
@@ -626,7 +626,7 @@ be able to run on our build server
 
  .. code-block:: none
 
-    conv2d with tensor core: 8.191855 ms
+    conv2d with tensor core: 6.923151 ms
 
 
 
diff --git a/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt b/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
index 88750a1f9..9fb039ad2 100644
--- a/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
@@ -118,8 +118,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
 
  .. code-block:: none
 
-    Numpy running time: 0.018082
-    Baseline: 3.226418
+    Numpy running time: 0.018355
+    Baseline: 3.388145
 
 
 
@@ -209,7 +209,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
 
  .. code-block:: none
 
-    Opt1: 0.302598
+    Opt1: 0.304776
 
 
 
@@ -307,7 +307,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
 
  .. code-block:: none
 
-    Opt2: 0.333762
+    Opt2: 0.339592
 
 
 
@@ -398,7 +398,7 @@ the access pattern for A matrix is more cache friendly.
 
  .. code-block:: none
 
-    Opt3: 0.115119
+    Opt3: 0.116929
 
 
 
@@ -516,7 +516,7 @@ flattening.
 
  .. code-block:: none
 
-    Opt4: 0.112020
+    Opt4: 0.110666
 
 
 
@@ -633,7 +633,7 @@ write to C when all the block results are ready.
 
  .. code-block:: none
 
-    Opt5: 0.111046
+    Opt5: 0.110533
 
 
 
@@ -753,7 +753,7 @@ Futhermore, we can also utilize multi-core processors to do the thread-level par
 
  .. code-block:: none
 
-    Opt6: 0.144175
+    Opt6: 0.145002
 
 
 
diff --git a/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt b/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt
index 7dfc2cb8d..8c2f0c6de 100644
--- a/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt
@@ -5,8 +5,8 @@
 
 Computation times
 =================
-**00:34.203** total execution time for **how_to_optimize_operators** files:
+**00:34.973** total execution time for **how_to_optimize_operators** files:
 
-- **00:31.709**: :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)
-- **00:01.346**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``)
-- **00:01.148**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)
+- **00:32.353**: :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)
+- **00:01.398**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``)
+- **00:01.221**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/sg_execution_times.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/sg_execution_times.rst.txt
index 04df6fe2c..311b5c1e8 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/sg_execution_times.rst.txt
@@ -5,11 +5,11 @@
 
 Computation times
 =================
-**04:56.539** total execution time for **how_to_tune_with_autoscheduler** files:
-
-- **02:17.771**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``)
-- **01:19.538**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)
-- **00:39.1000**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)
-- **00:22.188**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)
-- **00:08.630**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)
-- **00:08.413**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)
+**04:53.427** total execution time for **how_to_tune_with_autoscheduler** files:
+
+- **02:20.154**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``)
+- **01:18.957**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)
+- **00:40.191**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)
+- **00:17.223**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)
+- **00:08.638**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)
+- **00:08.264**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.rst.txt
index 2766600dd..0254c8e03 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.rst.txt
@@ -223,116 +223,347 @@ cooperative fetching, unrolling and operator fusion.
       buffer_map = {data_1: data, kernel_1: kernel, bias_1: bias, compute_1: compute} {
       attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 32;
       allocate(conv2d_nchw: Pointer(local float32), float32, [28]), storage_scope = local;
-      allocate(pad_temp.shared: Pointer(shared float32), float32, [252]), storage_scope = shared;
-      allocate(kernel.shared: Pointer(shared float32), float32, [192]), storage_scope = shared;
+      allocate(pad_temp.shared: Pointer(shared float32), float32, [504]), storage_scope = shared;
+      allocate(kernel.shared: Pointer(shared float32), float32, [384]), storage_scope = shared;
       attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28 {
-        conv2d_nchw_1: Buffer(conv2d_nchw, float32, [49], [], scope="local", align=16)[0] = 0f32
-        conv2d_nchw_1[7] = 0f32
-        conv2d_nchw_1[14] = 0f32
-        conv2d_nchw_1[21] = 0f32
-        conv2d_nchw_1[1] = 0f32
+        conv2d_nchw_1: Buffer(conv2d_nchw, float32, [16], [], scope="local", align=16)[0] = 0f32
+        conv2d_nchw_1[4] = 0f32
         conv2d_nchw_1[8] = 0f32
-        conv2d_nchw_1[15] = 0f32
-        conv2d_nchw_1[22] = 0f32
-        conv2d_nchw_1[2] = 0f32
-        conv2d_nchw_1[9] = 0f32
+        conv2d_nchw_1[12] = 0f32
         conv2d_nchw_1[16] = 0f32
-        conv2d_nchw_1[23] = 0f32
-        conv2d_nchw_1[3] = 0f32
-        conv2d_nchw_1[10] = 0f32
-        conv2d_nchw_1[17] = 0f32
+        conv2d_nchw_1[20] = 0f32
         conv2d_nchw_1[24] = 0f32
-        conv2d_nchw_1[4] = 0f32
-        conv2d_nchw_1[11] = 0f32
-        conv2d_nchw_1[18] = 0f32
-        conv2d_nchw_1[25] = 0f32
+        conv2d_nchw_1[1] = 0f32
         conv2d_nchw_1[5] = 0f32
-        conv2d_nchw_1[12] = 0f32
-        conv2d_nchw_1[19] = 0f32
-        conv2d_nchw_1[26] = 0f32
-        conv2d_nchw_1[6] = 0f32
+        conv2d_nchw_1[9] = 0f32
         conv2d_nchw_1[13] = 0f32
-        conv2d_nchw_1[20] = 0f32
+        conv2d_nchw_1[17] = 0f32
+        conv2d_nchw_1[21] = 0f32
+        conv2d_nchw_1[25] = 0f32
+        conv2d_nchw_1[2] = 0f32
+        conv2d_nchw_1[6] = 0f32
+        conv2d_nchw_1[10] = 0f32
+        conv2d_nchw_1[14] = 0f32
+        conv2d_nchw_1[18] = 0f32
+        conv2d_nchw_1[22] = 0f32
+        conv2d_nchw_1[26] = 0f32
+        conv2d_nchw_1[3] = 0f32
+        conv2d_nchw_1[7] = 0f32
+        conv2d_nchw_1[11] = 0f32
+        conv2d_nchw_1[15] = 0f32
+        conv2d_nchw_1[19] = 0f32
+        conv2d_nchw_1[23] = 0f32
         conv2d_nchw_1[27] = 0f32
-        for (rc.outer.outer: int32, 0, 128) {
-          for (rx.outer.outer: int32, 0, 3) {
-            let cse_var_2: int32 = (rc.outer.outer*196)
-            let cse_var_1: int32 = (rc.outer.outer*36)
+        for (rc.outer.outer: int32, 0, 64) {
+          for (ry.outer.outer: int32, 0, 3) {
+            let cse_var_4: int32 = (rc.outer.outer*392)
+            let cse_var_3: int32 = (ry.outer.outer*7)
+            let cse_var_2: int32 = (rc.outer.outer*72)
+            let cse_var_1: int32 = (ry.outer.outer*3)
              {
               attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1: Buffer(pad_temp.shared, float32, [252], [], scope="shared")[threadIdx.x_1] = @tir.if_then_else((((7 <= threadIdx.x_1) && (1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((cse_var_2 + threadIdx.x_1) + rx.outer.outer) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1: Buffer(pad_temp.shared, float32, [504], [], scope="shared")[threadIdx.x_1] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 9) + ry.outer.outer)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[((((cse_var_4 + (floordiv(threadIdx.x_1, 9)*7)) + cse_var_3) + floormod(threadIdx.x_1, 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 28)] = @tir.if_then_else(((((floordiv((threadIdx.x_1 + 28), 9) + ry.outer.outer) < 8) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 28), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 1), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 56)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 56), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 56), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 56), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 84)] = @tir.if_then_else(((1 <= floormod((threadIdx.x_1 + 3), 9)) && (floormod((threadIdx.x_1 + 3), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 84), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 3), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 112)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 112), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 112), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 4), 9))) && (floormod((threadIdx.x_1 + 4), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 112), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 4), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 140)] = @tir.if_then_else(((1 <= floormod((threadIdx.x_1 + 5), 9)) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 140), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 5), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 168)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 168), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 168), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 168), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 6), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 <= (floordiv(floormod((threadIdx.x_1 + 196), 63), 9) + ry.outer.outer)) && (1 <= floormod((threadIdx.x_1 + 7), 9))) && (floormod((threadIdx.x_1 + 7), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 196), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 7), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 224), 63), 9) + ry.outer.outer) < 8) && (1 <= floormod((threadIdx.x_1 + 8), 9))) && (floormod((threadIdx.x_1 + 8), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 224), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 8), 9)) - 8)], 0f32, dtype=float32)
+              attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              pad_temp.shared_1[(threadIdx.x_1 + 252)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 9) + ry.outer.outer)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[((((cse_var_4 + (floordiv(threadIdx.x_1, 9)*7)) + cse_var_3) + floormod(threadIdx.x_1, 9)) + 188)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 28)] = @tir.if_then_else(((1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[((((cse_var_2 + ((floordiv(threadIdx.x_1, 7) + 4)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1[(threadIdx.x_1 + 280)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 280), 63), 9) + ry.outer.outer) < 8) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 280), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 1), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 56)] = @tir.if_then_else(((((1 <= floormod((floordiv(threadIdx.x_1, 7) + 8), 9)) && (floormod((floordiv(threadIdx.x_1, 7) + 8), 9) < 8)) && (1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 8), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 8), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtyp [...]
+              pad_temp.shared_1[(threadIdx.x_1 + 308)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 308), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 308), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 308), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 84)] = @tir.if_then_else(((1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 12), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 3), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1[(threadIdx.x_1 + 336)] = @tir.if_then_else(((1 <= floormod((threadIdx.x_1 + 3), 9)) && (floormod((threadIdx.x_1 + 3), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 336), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 3), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 112)] = @tir.if_then_else(((((1 <= floormod((floordiv(threadIdx.x_1, 7) + 7), 9)) && (floormod((floordiv(threadIdx.x_1, 7) + 7), 9) < 8)) && (1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 16), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 7), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dt [...]
+              pad_temp.shared_1[(threadIdx.x_1 + 364)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 364), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 364), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 4), 9))) && (floormod((threadIdx.x_1 + 4), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 364), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 4), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 140)] = @tir.if_then_else(((1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 20), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 2), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else(((1 <= floormod((threadIdx.x_1 + 5), 9)) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 392), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 5), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 168)] = @tir.if_then_else(((((1 <= floormod((floordiv(threadIdx.x_1, 7) + 6), 9)) && (floormod((floordiv(threadIdx.x_1, 7) + 6), 9) < 8)) && (1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 24), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 6), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dt [...]
+              pad_temp.shared_1[(threadIdx.x_1 + 420)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 420), 63), 9) + ry.outer.outer)) && ((floordiv(floormod((threadIdx.x_1 + 420), 63), 9) + ry.outer.outer) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 420), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 6), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else(((1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 28), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 1), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1[(threadIdx.x_1 + 448)] = @tir.if_then_else((((1 <= (floordiv(floormod((threadIdx.x_1 + 448), 63), 9) + ry.outer.outer)) && (1 <= floormod((threadIdx.x_1 + 7), 9))) && (floormod((threadIdx.x_1 + 7), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 448), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 7), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else((((floormod((floordiv(threadIdx.x_1, 7) + 5), 9) < 8) && (1 <= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) && ((rx.outer.outer + floormod(threadIdx.x_1, 7)) < 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 32), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 5), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+              pad_temp.shared_1[(threadIdx.x_1 + 476)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 476), 63), 9) + ry.outer.outer) < 8) && (1 <= floormod((threadIdx.x_1 + 8), 9))) && (floormod((threadIdx.x_1 + 8), 9) < 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 476), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 8), 9)) - 8)], 0f32, dtype=float32)
               attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1: Buffer(kernel.shared, float32, [192], [], scope="shared")[threadIdx.x_2] = kernel[(((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 12)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer)]
+              kernel.shared_1: Buffer(kernel.shared, float32, [384], [], scope="shared")[threadIdx.x_2] = kernel[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 24)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1[(threadIdx.x_2 + 28)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 7), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 4), 12)*3)) + rx.outer.outer)]
+              kernel.shared_1[(threadIdx.x_2 + 28)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 7), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 28), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1[(threadIdx.x_2 + 56)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 14), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 8), 12)*3)) + rx.outer.outer)]
+              kernel.shared_1[(threadIdx.x_2 + 56)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 14), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 56), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1[(threadIdx.x_2 + 84)] = kernel[((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 3)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer) + 32256)]
+              kernel.shared_1[(threadIdx.x_2 + 84)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 21), 6)*4608)) + cse_var_2) + (floormod((floordiv(threadIdx.x_2, 3) + 4), 8)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1[(threadIdx.x_2 + 112)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 28), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 4), 12)*3)) + rx.outer.outer)]
+              kernel.shared_1[(threadIdx.x_2 + 112)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 28), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 112), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              kernel.shared_1[(threadIdx.x_2 + 140)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 35), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 8), 12)*3)) + rx.outer.outer)]
+              kernel.shared_1[(threadIdx.x_2 + 140)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 35), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 140), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
               attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
-              if @tir.likely((threadIdx.x_2 < 24), dtype=bool) {
-                kernel.shared_1[(threadIdx.x_2 + 168)] = kernel[((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 3)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer) + 64512)]
+              kernel.shared_1[(threadIdx.x_2 + 168)] = kernel[(((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 6)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 32256)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 196)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 49), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 196), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 224)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 56), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 224), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 252)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 63), 6)*4608)) + cse_var_2) + (floormod((floordiv(threadIdx.x_2, 3) + 4), 8)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 280)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 70), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 280), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 308)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 77), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 308), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              kernel.shared_1[(threadIdx.x_2 + 336)] = kernel[(((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 6)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 64512)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 28;
+              if @tir.likely((threadIdx.x_2 < 20), dtype=bool) {
+                kernel.shared_1[(threadIdx.x_2 + 364)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 91), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 364), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
               }
-              for (ry.outer.inner: int32, 0, 3) {
-                for (rc.inner: int32, 0, 4) {
-                  conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                  conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-                  conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-                  conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-                  conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-                }
+              for (rx.outer.inner: int32, 0, 3) {
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+                conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
               }
             }
           }
         }
-        for (i2.inner: int32, 0, 7) {
-          compute[((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[i2.inner] + bias[((blockIdx.x*16) + floordiv(threadIdx.x, 7))]), 0f32)
-          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 196)] = max((conv2d_nchw_1[(i2.inner + 7)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 4)]), 0f32)
-          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 392)] = max((conv2d_nchw_1[(i2.inner + 14)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 8)]), 0f32)
-          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 588)] = max((conv2d_nchw_1[(i2.inner + 21)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 12)]), 0f32)
+        for (i1.inner: int32, 0, 4) {
+          compute[((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[i1.inner] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 7)] = max((conv2d_nchw_1[(i1.inner + 4)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 14)] = max((conv2d_nchw_1[(i1.inner + 8)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 21)] = max((conv2d_nchw_1[(i1.inner + 12)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 28)] = max((conv2d_nchw_1[(i1.inner + 16)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 35)] = max((conv2d_nchw_1[(i1.inner + 20)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+          compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 42)] = max((conv2d_nchw_1[(i1.inner + 24)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
         }
       }
     }
@@ -385,7 +616,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 0.351 ms
+    Execution time of this operator: 0.315 ms
 
 
 
@@ -430,33 +661,33 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
     conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
     conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
-    conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
+    conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=4)
     conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=4)
-    conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=4)
-    conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=7)
+    conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=1)
+    conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=1)
     conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
     conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
-    conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
+    conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=7)
     conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
     conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
     conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=7)
     conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
-    conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
+    conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=8)
     conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=1)
     conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=1)
-    conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
+    conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=1)
     conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=1)
-    conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+    conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
     s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2 [...]
     compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
     compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
     compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-    compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
+    compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=4)
     compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=4)
-    compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=4)
-    compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=7)
+    compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
+    compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
     compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
-    compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
+    compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=7)
     compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
     compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=7)
     compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
@@ -485,7 +716,7 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     s[pad_temp_shared].vectorize(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i)
     pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=28)
     s[pad_temp_shared].bind(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i, te.thread_axis("threadIdx.x"))
-    s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "auto_unroll_max_step", 64)
+    s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "auto_unroll_max_step", 512)
     s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "unroll_explicit", True)
 
     CUDA source code:
@@ -505,97 +736,310 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     #endif
     extern "C" __global__ void __launch_bounds__(28) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
       float conv2d_nchw[28];
-      __shared__ float pad_temp_shared[252];
-      __shared__ float kernel_shared[192];
+      __shared__ float pad_temp_shared[504];
+      __shared__ float kernel_shared[384];
       conv2d_nchw[0] = 0.000000e+00f;
-      conv2d_nchw[7] = 0.000000e+00f;
-      conv2d_nchw[14] = 0.000000e+00f;
-      conv2d_nchw[21] = 0.000000e+00f;
-      conv2d_nchw[1] = 0.000000e+00f;
+      conv2d_nchw[4] = 0.000000e+00f;
       conv2d_nchw[8] = 0.000000e+00f;
-      conv2d_nchw[15] = 0.000000e+00f;
-      conv2d_nchw[22] = 0.000000e+00f;
-      conv2d_nchw[2] = 0.000000e+00f;
-      conv2d_nchw[9] = 0.000000e+00f;
+      conv2d_nchw[12] = 0.000000e+00f;
       conv2d_nchw[16] = 0.000000e+00f;
-      conv2d_nchw[23] = 0.000000e+00f;
-      conv2d_nchw[3] = 0.000000e+00f;
-      conv2d_nchw[10] = 0.000000e+00f;
-      conv2d_nchw[17] = 0.000000e+00f;
+      conv2d_nchw[20] = 0.000000e+00f;
       conv2d_nchw[24] = 0.000000e+00f;
-      conv2d_nchw[4] = 0.000000e+00f;
-      conv2d_nchw[11] = 0.000000e+00f;
-      conv2d_nchw[18] = 0.000000e+00f;
-      conv2d_nchw[25] = 0.000000e+00f;
+      conv2d_nchw[1] = 0.000000e+00f;
       conv2d_nchw[5] = 0.000000e+00f;
-      conv2d_nchw[12] = 0.000000e+00f;
-      conv2d_nchw[19] = 0.000000e+00f;
-      conv2d_nchw[26] = 0.000000e+00f;
-      conv2d_nchw[6] = 0.000000e+00f;
+      conv2d_nchw[9] = 0.000000e+00f;
       conv2d_nchw[13] = 0.000000e+00f;
-      conv2d_nchw[20] = 0.000000e+00f;
+      conv2d_nchw[17] = 0.000000e+00f;
+      conv2d_nchw[21] = 0.000000e+00f;
+      conv2d_nchw[25] = 0.000000e+00f;
+      conv2d_nchw[2] = 0.000000e+00f;
+      conv2d_nchw[6] = 0.000000e+00f;
+      conv2d_nchw[10] = 0.000000e+00f;
+      conv2d_nchw[14] = 0.000000e+00f;
+      conv2d_nchw[18] = 0.000000e+00f;
+      conv2d_nchw[22] = 0.000000e+00f;
+      conv2d_nchw[26] = 0.000000e+00f;
+      conv2d_nchw[3] = 0.000000e+00f;
+      conv2d_nchw[7] = 0.000000e+00f;
+      conv2d_nchw[11] = 0.000000e+00f;
+      conv2d_nchw[15] = 0.000000e+00f;
+      conv2d_nchw[19] = 0.000000e+00f;
+      conv2d_nchw[23] = 0.000000e+00f;
       conv2d_nchw[27] = 0.000000e+00f;
-      for (int rc_outer_outer = 0; rc_outer_outer < 128; ++rc_outer_outer) {
-        for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
+      for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
+        for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
           __syncthreads();
-          pad_temp_shared[((int)threadIdx.x)] = ((((7 <= ((int)threadIdx.x)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((rc_outer_outer * 196) + ((int)threadIdx.x)) + rx_outer_outer) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 28)] = (((1 <= (rx_outer_outer + (((int)threadIdx.x) % 7))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((rc_outer_outer * 196) + ((int)threadIdx.x)) + rx_outer_outer) + 20)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 56)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 56) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 84)] = (((1 <= (rx_outer_outer + (((int)threadIdx.x) % 7))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 84) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 3) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 112) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 140)] = (((1 <= (rx_outer_outer + (((int)threadIdx.x) % 7))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 140) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 2) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 168)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 168) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 196)] = (((1 <= (rx_outer_outer + (((int)threadIdx.x) % 7))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 196) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 1) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 224)] = ((((((int)threadIdx.x) < 21) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 224) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 5) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-          kernel_shared[((int)threadIdx.x)] = kernel[(((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer)];
-          kernel_shared[(((int)threadIdx.x) + 28)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 28) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 4) % 12) * 3)) + rx_outer_outer)];
-          kernel_shared[(((int)threadIdx.x) + 56)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 56) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 8) % 12) * 3)) + rx_outer_outer)];
-          kernel_shared[(((int)threadIdx.x) + 84)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer) + 32256)];
-          kernel_shared[(((int)threadIdx.x) + 112)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 112) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 4) % 12) * 3)) + rx_outer_outer)];
-          kernel_shared[(((int)threadIdx.x) + 140)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 140) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 8) % 12) * 3)) + rx_outer_outer)];
-          if (((int)threadIdx.x) < 24) {
-            kernel_shared[(((int)threadIdx.x) + 168)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer) + 64512)];
+          pad_temp_shared[((int)threadIdx.x)] = ((((1 <= ((((int)threadIdx.x) / 9) + ry_outer_outer)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 28)] = (((((((((int)threadIdx.x) + 28) / 9) + ry_outer_outer) < 8) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 28) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 56)] = (((((1 <= ((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 56) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 84)] = (((1 <= ((((int)threadIdx.x) + 3) % 9)) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 84) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 112) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 140)] = (((1 <= ((((int)threadIdx.x) + 5) % 9)) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 140) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 168)] = (((((1 <= ((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 168) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 <= ((((((int)threadIdx.x) + 7) % 63) / 9) + ry_outer_outer)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 196) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 224)] = ((((((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) < 8) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 224) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 252)] = ((((1 <= ((((int)threadIdx.x) / 9) + ry_outer_outer)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) + 188)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 280)] = ((((((((((int)threadIdx.x) + 28) % 63) / 9) + ry_outer_outer) < 8) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 280) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 308)] = (((((1 <= ((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 308) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 336)] = (((1 <= ((((int)threadIdx.x) + 3) % 9)) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 336) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 364)] = (((((1 <= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 364) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 392)] = (((1 <= ((((int)threadIdx.x) + 5) % 9)) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 392) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 420)] = (((((1 <= ((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 420) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 448)] = ((((1 <= ((((((int)threadIdx.x) + 7) % 63) / 9) + ry_outer_outer)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 448) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+          pad_temp_shared[(((int)threadIdx.x) + 476)] = ((((((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) < 8) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 476) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
+          kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 28)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 28) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 56)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 56) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 84)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 84) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) / 3) + 4) & 7) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 112)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 112) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 140)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 140) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 20) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 168)] = kernel[(((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 32256)];
+          kernel_shared[(((int)threadIdx.x) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 196) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 224) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 252)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 252) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) / 3) + 4) & 7) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 280)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 308)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 308) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 20) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 64512)];
+          if (((int)threadIdx.x) < 20) {
+            kernel_shared[(((int)threadIdx.x) + 364)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 364) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
           }
           __syncthreads();
-          for (int ry_outer_inner = 0; ry_outer_inner < 3; ++ry_outer_inner) {
-            for (int rc_inner = 0; rc_inner < 4; ++rc_inner) {
-              conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-              conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-              conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-              conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-              conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-            }
+          for (int rx_outer_inner = 0; rx_outer_inner < 3; ++rx_outer_inner) {
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+            conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
           }
         }
       }
-      for (int i2_inner = 0; i2_inner < 7; ++i2_inner) {
-        compute[((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[i2_inner] + bias[((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
-        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 196)] = max((conv2d_nchw[(i2_inner + 7)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 4)]), 0.000000e+00f);
-        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 392)] = max((conv2d_nchw[(i2_inner + 14)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 8)]), 0.000000e+00f);
-        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 588)] = max((conv2d_nchw[(i2_inner + 21)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 12)]), 0.000000e+00f);
+      for (int i1_inner = 0; i1_inner < 4; ++i1_inner) {
+        compute[((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[i1_inner] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 7)] = max((conv2d_nchw[(i1_inner + 4)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 14)] = max((conv2d_nchw[(i1_inner + 8)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 21)] = max((conv2d_nchw[(i1_inner + 12)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 28)] = max((conv2d_nchw[(i1_inner + 16)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 35)] = max((conv2d_nchw[(i1_inner + 20)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+        compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 42)] = max((conv2d_nchw[(i1_inner + 24)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
       }
     }
 
@@ -654,7 +1098,7 @@ In the example below we resume the status and do more 5 trials.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 2 minutes  17.771 seconds)
+   **Total running time of the script:** ( 2 minutes  20.154 seconds)
 
 
 .. _sphx_glr_download_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py:
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
index 6bbd851e8..4ea38d9b8 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
@@ -614,7 +614,7 @@ so we can read the log file and load the best schedules.
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-       9.6715       9.6668       9.6970       9.6506       0.0193   
+       9.8233       9.8322       9.8639       9.7739       0.0372   
                
 
 
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_x86.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_x86.rst.txt
index 3d287b295..5857df18b 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_x86.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_x86.rst.txt
@@ -633,7 +633,7 @@ so we can read the log file and load the best schedules.
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      762.0476     761.2159     765.8830     759.0438      2.8534   
+      755.6295     751.7929     764.4012     750.6943      6.2188   
                
 
 
@@ -658,7 +658,7 @@ Other Tips
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  19.538 seconds)
+   **Total running time of the script:** ( 1 minutes  18.957 seconds)
 
 
 .. _sphx_glr_download_how_to_tune_with_autoscheduler_tune_network_x86.py:
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
index 6b7dcda10..2a4d835fb 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
@@ -362,790 +362,407 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
                  placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [65536], []),
                  compute: Buffer(compute_2: Pointer(float32), float32, [65536], [])}
       buffer_map = {placeholder_5: placeholder, placeholder_6: placeholder_1, placeholder_7: placeholder_2, placeholder_8: placeholder_3, placeholder_9: placeholder_4, compute_1: compute} {
-      for (i0.outer: int32, 0, 2) "parallel" {
-        allocate(compute_3: Pointer(global float32), float32, [1024]), storage_scope = global;
-        for (i1.outer: int32, 0, 32) {
-          for (i.outer.inner: int32, 0, 4) {
-            let cse_var_1: int32 = (i.outer.inner*256)
+      for (i0.outer: int32, 0, 16) "parallel" {
+        allocate(compute_3: Pointer(global float32), float32, [256]), storage_scope = global;
+        for (i1.outer: int32, 0, 16) {
+          for (nb_j.inner: int32, 0, 2) {
+            let cse_var_2: int32 = (nb_j.inner*16)
+            let cse_var_1: int32 = ((i1.outer*2) + nb_j.inner)
              {
-              compute_4: Buffer(compute_3, float32, [1024], [])[cse_var_1] = 0f32
-              compute_4[(cse_var_1 + 1)] = 0f32
-              compute_4[(cse_var_1 + 2)] = 0f32
-              compute_4[(cse_var_1 + 3)] = 0f32
-              compute_4[(cse_var_1 + 4)] = 0f32
-              compute_4[(cse_var_1 + 5)] = 0f32
-              compute_4[(cse_var_1 + 6)] = 0f32
-              compute_4[(cse_var_1 + 7)] = 0f32
-              compute_4[(cse_var_1 + 8)] = 0f32
-              compute_4[(cse_var_1 + 9)] = 0f32
-              compute_4[(cse_var_1 + 10)] = 0f32
-              compute_4[(cse_var_1 + 11)] = 0f32
-              compute_4[(cse_var_1 + 12)] = 0f32
-              compute_4[(cse_var_1 + 13)] = 0f32
-              compute_4[(cse_var_1 + 14)] = 0f32
-              compute_4[(cse_var_1 + 15)] = 0f32
-              compute_4[(cse_var_1 + 16)] = 0f32
-              compute_4[(cse_var_1 + 17)] = 0f32
-              compute_4[(cse_var_1 + 18)] = 0f32
-              compute_4[(cse_var_1 + 19)] = 0f32
-              compute_4[(cse_var_1 + 20)] = 0f32
-              compute_4[(cse_var_1 + 21)] = 0f32
-              compute_4[(cse_var_1 + 22)] = 0f32
-              compute_4[(cse_var_1 + 23)] = 0f32
-              compute_4[(cse_var_1 + 24)] = 0f32
-              compute_4[(cse_var_1 + 25)] = 0f32
-              compute_4[(cse_var_1 + 26)] = 0f32
-              compute_4[(cse_var_1 + 27)] = 0f32
-              compute_4[(cse_var_1 + 28)] = 0f32
-              compute_4[(cse_var_1 + 29)] = 0f32
-              compute_4[(cse_var_1 + 30)] = 0f32
-              compute_4[(cse_var_1 + 31)] = 0f32
-              compute_4[(cse_var_1 + 32)] = 0f32
-              compute_4[(cse_var_1 + 33)] = 0f32
-              compute_4[(cse_var_1 + 34)] = 0f32
-              compute_4[(cse_var_1 + 35)] = 0f32
-              compute_4[(cse_var_1 + 36)] = 0f32
-              compute_4[(cse_var_1 + 37)] = 0f32
-              compute_4[(cse_var_1 + 38)] = 0f32
-              compute_4[(cse_var_1 + 39)] = 0f32
-              compute_4[(cse_var_1 + 40)] = 0f32
-              compute_4[(cse_var_1 + 41)] = 0f32
-              compute_4[(cse_var_1 + 42)] = 0f32
-              compute_4[(cse_var_1 + 43)] = 0f32
-              compute_4[(cse_var_1 + 44)] = 0f32
-              compute_4[(cse_var_1 + 45)] = 0f32
-              compute_4[(cse_var_1 + 46)] = 0f32
-              compute_4[(cse_var_1 + 47)] = 0f32
-              compute_4[(cse_var_1 + 48)] = 0f32
-              compute_4[(cse_var_1 + 49)] = 0f32
-              compute_4[(cse_var_1 + 50)] = 0f32
-              compute_4[(cse_var_1 + 51)] = 0f32
-              compute_4[(cse_var_1 + 52)] = 0f32
-              compute_4[(cse_var_1 + 53)] = 0f32
-              compute_4[(cse_var_1 + 54)] = 0f32
-              compute_4[(cse_var_1 + 55)] = 0f32
-              compute_4[(cse_var_1 + 56)] = 0f32
-              compute_4[(cse_var_1 + 57)] = 0f32
-              compute_4[(cse_var_1 + 58)] = 0f32
-              compute_4[(cse_var_1 + 59)] = 0f32
-              compute_4[(cse_var_1 + 60)] = 0f32
-              compute_4[(cse_var_1 + 61)] = 0f32
-              compute_4[(cse_var_1 + 62)] = 0f32
-              compute_4[(cse_var_1 + 63)] = 0f32
-              compute_4[(cse_var_1 + 64)] = 0f32
-              compute_4[(cse_var_1 + 65)] = 0f32
-              compute_4[(cse_var_1 + 66)] = 0f32
-              compute_4[(cse_var_1 + 67)] = 0f32
-              compute_4[(cse_var_1 + 68)] = 0f32
-              compute_4[(cse_var_1 + 69)] = 0f32
-              compute_4[(cse_var_1 + 70)] = 0f32
-              compute_4[(cse_var_1 + 71)] = 0f32
-              compute_4[(cse_var_1 + 72)] = 0f32
-              compute_4[(cse_var_1 + 73)] = 0f32
-              compute_4[(cse_var_1 + 74)] = 0f32
-              compute_4[(cse_var_1 + 75)] = 0f32
-              compute_4[(cse_var_1 + 76)] = 0f32
-              compute_4[(cse_var_1 + 77)] = 0f32
-              compute_4[(cse_var_1 + 78)] = 0f32
-              compute_4[(cse_var_1 + 79)] = 0f32
-              compute_4[(cse_var_1 + 80)] = 0f32
-              compute_4[(cse_var_1 + 81)] = 0f32
-              compute_4[(cse_var_1 + 82)] = 0f32
-              compute_4[(cse_var_1 + 83)] = 0f32
-              compute_4[(cse_var_1 + 84)] = 0f32
-              compute_4[(cse_var_1 + 85)] = 0f32
-              compute_4[(cse_var_1 + 86)] = 0f32
-              compute_4[(cse_var_1 + 87)] = 0f32
-              compute_4[(cse_var_1 + 88)] = 0f32
-              compute_4[(cse_var_1 + 89)] = 0f32
-              compute_4[(cse_var_1 + 90)] = 0f32
-              compute_4[(cse_var_1 + 91)] = 0f32
-              compute_4[(cse_var_1 + 92)] = 0f32
-              compute_4[(cse_var_1 + 93)] = 0f32
-              compute_4[(cse_var_1 + 94)] = 0f32
-              compute_4[(cse_var_1 + 95)] = 0f32
-              compute_4[(cse_var_1 + 96)] = 0f32
-              compute_4[(cse_var_1 + 97)] = 0f32
-              compute_4[(cse_var_1 + 98)] = 0f32
-              compute_4[(cse_var_1 + 99)] = 0f32
-              compute_4[(cse_var_1 + 100)] = 0f32
-              compute_4[(cse_var_1 + 101)] = 0f32
-              compute_4[(cse_var_1 + 102)] = 0f32
-              compute_4[(cse_var_1 + 103)] = 0f32
-              compute_4[(cse_var_1 + 104)] = 0f32
-              compute_4[(cse_var_1 + 105)] = 0f32
-              compute_4[(cse_var_1 + 106)] = 0f32
-              compute_4[(cse_var_1 + 107)] = 0f32
-              compute_4[(cse_var_1 + 108)] = 0f32
-              compute_4[(cse_var_1 + 109)] = 0f32
-              compute_4[(cse_var_1 + 110)] = 0f32
-              compute_4[(cse_var_1 + 111)] = 0f32
-              compute_4[(cse_var_1 + 112)] = 0f32
-              compute_4[(cse_var_1 + 113)] = 0f32
-              compute_4[(cse_var_1 + 114)] = 0f32
-              compute_4[(cse_var_1 + 115)] = 0f32
-              compute_4[(cse_var_1 + 116)] = 0f32
-              compute_4[(cse_var_1 + 117)] = 0f32
-              compute_4[(cse_var_1 + 118)] = 0f32
-              compute_4[(cse_var_1 + 119)] = 0f32
-              compute_4[(cse_var_1 + 120)] = 0f32
-              compute_4[(cse_var_1 + 121)] = 0f32
-              compute_4[(cse_var_1 + 122)] = 0f32
-              compute_4[(cse_var_1 + 123)] = 0f32
-              compute_4[(cse_var_1 + 124)] = 0f32
-              compute_4[(cse_var_1 + 125)] = 0f32
-              compute_4[(cse_var_1 + 126)] = 0f32
-              compute_4[(cse_var_1 + 127)] = 0f32
-              compute_4[(cse_var_1 + 128)] = 0f32
-              compute_4[(cse_var_1 + 129)] = 0f32
-              compute_4[(cse_var_1 + 130)] = 0f32
-              compute_4[(cse_var_1 + 131)] = 0f32
-              compute_4[(cse_var_1 + 132)] = 0f32
-              compute_4[(cse_var_1 + 133)] = 0f32
-              compute_4[(cse_var_1 + 134)] = 0f32
-              compute_4[(cse_var_1 + 135)] = 0f32
-              compute_4[(cse_var_1 + 136)] = 0f32
-              compute_4[(cse_var_1 + 137)] = 0f32
-              compute_4[(cse_var_1 + 138)] = 0f32
-              compute_4[(cse_var_1 + 139)] = 0f32
-              compute_4[(cse_var_1 + 140)] = 0f32
-              compute_4[(cse_var_1 + 141)] = 0f32
-              compute_4[(cse_var_1 + 142)] = 0f32
-              compute_4[(cse_var_1 + 143)] = 0f32
-              compute_4[(cse_var_1 + 144)] = 0f32
-              compute_4[(cse_var_1 + 145)] = 0f32
-              compute_4[(cse_var_1 + 146)] = 0f32
-              compute_4[(cse_var_1 + 147)] = 0f32
-              compute_4[(cse_var_1 + 148)] = 0f32
-              compute_4[(cse_var_1 + 149)] = 0f32
-              compute_4[(cse_var_1 + 150)] = 0f32
-              compute_4[(cse_var_1 + 151)] = 0f32
-              compute_4[(cse_var_1 + 152)] = 0f32
-              compute_4[(cse_var_1 + 153)] = 0f32
-              compute_4[(cse_var_1 + 154)] = 0f32
-              compute_4[(cse_var_1 + 155)] = 0f32
-              compute_4[(cse_var_1 + 156)] = 0f32
-              compute_4[(cse_var_1 + 157)] = 0f32
-              compute_4[(cse_var_1 + 158)] = 0f32
-              compute_4[(cse_var_1 + 159)] = 0f32
-              compute_4[(cse_var_1 + 160)] = 0f32
-              compute_4[(cse_var_1 + 161)] = 0f32
-              compute_4[(cse_var_1 + 162)] = 0f32
-              compute_4[(cse_var_1 + 163)] = 0f32
-              compute_4[(cse_var_1 + 164)] = 0f32
-              compute_4[(cse_var_1 + 165)] = 0f32
-              compute_4[(cse_var_1 + 166)] = 0f32
-              compute_4[(cse_var_1 + 167)] = 0f32
-              compute_4[(cse_var_1 + 168)] = 0f32
-              compute_4[(cse_var_1 + 169)] = 0f32
-              compute_4[(cse_var_1 + 170)] = 0f32
-              compute_4[(cse_var_1 + 171)] = 0f32
-              compute_4[(cse_var_1 + 172)] = 0f32
-              compute_4[(cse_var_1 + 173)] = 0f32
-              compute_4[(cse_var_1 + 174)] = 0f32
-              compute_4[(cse_var_1 + 175)] = 0f32
-              compute_4[(cse_var_1 + 176)] = 0f32
-              compute_4[(cse_var_1 + 177)] = 0f32
-              compute_4[(cse_var_1 + 178)] = 0f32
-              compute_4[(cse_var_1 + 179)] = 0f32
-              compute_4[(cse_var_1 + 180)] = 0f32
-              compute_4[(cse_var_1 + 181)] = 0f32
-              compute_4[(cse_var_1 + 182)] = 0f32
-              compute_4[(cse_var_1 + 183)] = 0f32
-              compute_4[(cse_var_1 + 184)] = 0f32
-              compute_4[(cse_var_1 + 185)] = 0f32
-              compute_4[(cse_var_1 + 186)] = 0f32
-              compute_4[(cse_var_1 + 187)] = 0f32
-              compute_4[(cse_var_1 + 188)] = 0f32
-              compute_4[(cse_var_1 + 189)] = 0f32
-              compute_4[(cse_var_1 + 190)] = 0f32
-              compute_4[(cse_var_1 + 191)] = 0f32
-              compute_4[(cse_var_1 + 192)] = 0f32
-              compute_4[(cse_var_1 + 193)] = 0f32
-              compute_4[(cse_var_1 + 194)] = 0f32
-              compute_4[(cse_var_1 + 195)] = 0f32
-              compute_4[(cse_var_1 + 196)] = 0f32
-              compute_4[(cse_var_1 + 197)] = 0f32
-              compute_4[(cse_var_1 + 198)] = 0f32
-              compute_4[(cse_var_1 + 199)] = 0f32
-              compute_4[(cse_var_1 + 200)] = 0f32
-              compute_4[(cse_var_1 + 201)] = 0f32
-              compute_4[(cse_var_1 + 202)] = 0f32
-              compute_4[(cse_var_1 + 203)] = 0f32
-              compute_4[(cse_var_1 + 204)] = 0f32
-              compute_4[(cse_var_1 + 205)] = 0f32
-              compute_4[(cse_var_1 + 206)] = 0f32
-              compute_4[(cse_var_1 + 207)] = 0f32
-              compute_4[(cse_var_1 + 208)] = 0f32
-              compute_4[(cse_var_1 + 209)] = 0f32
-              compute_4[(cse_var_1 + 210)] = 0f32
-              compute_4[(cse_var_1 + 211)] = 0f32
-              compute_4[(cse_var_1 + 212)] = 0f32
-              compute_4[(cse_var_1 + 213)] = 0f32
-              compute_4[(cse_var_1 + 214)] = 0f32
-              compute_4[(cse_var_1 + 215)] = 0f32
-              compute_4[(cse_var_1 + 216)] = 0f32
-              compute_4[(cse_var_1 + 217)] = 0f32
-              compute_4[(cse_var_1 + 218)] = 0f32
-              compute_4[(cse_var_1 + 219)] = 0f32
-              compute_4[(cse_var_1 + 220)] = 0f32
-              compute_4[(cse_var_1 + 221)] = 0f32
-              compute_4[(cse_var_1 + 222)] = 0f32
-              compute_4[(cse_var_1 + 223)] = 0f32
-              compute_4[(cse_var_1 + 224)] = 0f32
-              compute_4[(cse_var_1 + 225)] = 0f32
-              compute_4[(cse_var_1 + 226)] = 0f32
-              compute_4[(cse_var_1 + 227)] = 0f32
-              compute_4[(cse_var_1 + 228)] = 0f32
-              compute_4[(cse_var_1 + 229)] = 0f32
-              compute_4[(cse_var_1 + 230)] = 0f32
-              compute_4[(cse_var_1 + 231)] = 0f32
-              compute_4[(cse_var_1 + 232)] = 0f32
-              compute_4[(cse_var_1 + 233)] = 0f32
-              compute_4[(cse_var_1 + 234)] = 0f32
-              compute_4[(cse_var_1 + 235)] = 0f32
-              compute_4[(cse_var_1 + 236)] = 0f32
-              compute_4[(cse_var_1 + 237)] = 0f32
-              compute_4[(cse_var_1 + 238)] = 0f32
-              compute_4[(cse_var_1 + 239)] = 0f32
-              compute_4[(cse_var_1 + 240)] = 0f32
-              compute_4[(cse_var_1 + 241)] = 0f32
-              compute_4[(cse_var_1 + 242)] = 0f32
-              compute_4[(cse_var_1 + 243)] = 0f32
-              compute_4[(cse_var_1 + 244)] = 0f32
-              compute_4[(cse_var_1 + 245)] = 0f32
-              compute_4[(cse_var_1 + 246)] = 0f32
-              compute_4[(cse_var_1 + 247)] = 0f32
-              compute_4[(cse_var_1 + 248)] = 0f32
-              compute_4[(cse_var_1 + 249)] = 0f32
-              compute_4[(cse_var_1 + 250)] = 0f32
-              compute_4[(cse_var_1 + 251)] = 0f32
-              compute_4[(cse_var_1 + 252)] = 0f32
-              compute_4[(cse_var_1 + 253)] = 0f32
-              compute_4[(cse_var_1 + 254)] = 0f32
-              compute_4[(cse_var_1 + 255)] = 0f32
-              for (elem_idx: int32, 0, (placeholder_3[(i1.outer + 1)] - placeholder_3[i1.outer])) {
-                let cse_var_258: int32 = (cse_var_1 + 184)
-                let cse_var_257: int32 = (cse_var_1 + 183)
-                let cse_var_256: int32 = (cse_var_1 + 182)
-                let cse_var_255: int32 = (cse_var_1 + 181)
-                let cse_var_254: int32 = (cse_var_1 + 180)
-                let cse_var_253: int32 = (cse_var_1 + 18)
-                let cse_var_252: int32 = (cse_var_1 + 179)
-                let cse_var_251: int32 = (cse_var_1 + 178)
-                let cse_var_250: int32 = (cse_var_1 + 177)
-                let cse_var_249: int32 = (cse_var_1 + 176)
-                let cse_var_248: int32 = (cse_var_1 + 175)
-                let cse_var_247: int32 = (cse_var_1 + 174)
-                let cse_var_246: int32 = (cse_var_1 + 173)
-                let cse_var_245: int32 = (cse_var_1 + 172)
-                let cse_var_244: int32 = (cse_var_1 + 171)
-                let cse_var_243: int32 = (cse_var_1 + 214)
-                let cse_var_242: int32 = (cse_var_1 + 17)
-                let cse_var_241: int32 = (cse_var_1 + 169)
-                let cse_var_240: int32 = (cse_var_1 + 168)
-                let cse_var_239: int32 = (cse_var_1 + 167)
-                let cse_var_238: int32 = (cse_var_1 + 166)
-                let cse_var_237: int32 = (cse_var_1 + 165)
-                let cse_var_236: int32 = (cse_var_1 + 164)
-                let cse_var_235: int32 = (cse_var_1 + 163)
-                let cse_var_234: int32 = (cse_var_1 + 162)
-                let cse_var_233: int32 = (cse_var_1 + 161)
-                let cse_var_232: int32 = (cse_var_1 + 160)
-                let cse_var_231: int32 = (cse_var_1 + 16)
-                let cse_var_230: int32 = (cse_var_1 + 159)
-                let cse_var_229: int32 = (cse_var_1 + 158)
-                let cse_var_228: int32 = (cse_var_1 + 157)
-                let cse_var_227: int32 = (cse_var_1 + 170)
-                let cse_var_226: int32 = (cse_var_1 + 212)
-                let cse_var_225: int32 = (cse_var_1 + 211)
-                let cse_var_224: int32 = (cse_var_1 + 210)
-                let cse_var_223: int32 = (cse_var_1 + 21)
-                let cse_var_222: int32 = (cse_var_1 + 209)
-                let cse_var_221: int32 = (cse_var_1 + 208)
-                let cse_var_220: int32 = (cse_var_1 + 207)
-                let cse_var_219: int32 = (cse_var_1 + 206)
-                let cse_var_218: int32 = (cse_var_1 + 205)
-                let cse_var_217: int32 = (cse_var_1 + 204)
-                let cse_var_216: int32 = (cse_var_1 + 203)
-                let cse_var_215: int32 = (cse_var_1 + 202)
-                let cse_var_214: int32 = (cse_var_1 + 201)
-                let cse_var_213: int32 = (cse_var_1 + 200)
-                let cse_var_212: int32 = (cse_var_1 + 20)
-                let cse_var_211: int32 = (cse_var_1 + 185)
-                let cse_var_210: int32 = (cse_var_1 + 199)
-                let cse_var_209: int32 = (cse_var_1 + 198)
-                let cse_var_208: int32 = (cse_var_1 + 197)
-                let cse_var_207: int32 = (cse_var_1 + 196)
-                let cse_var_206: int32 = (cse_var_1 + 195)
-                let cse_var_205: int32 = (cse_var_1 + 194)
-                let cse_var_204: int32 = (cse_var_1 + 193)
-                let cse_var_203: int32 = (cse_var_1 + 192)
-                let cse_var_202: int32 = (cse_var_1 + 191)
-                let cse_var_201: int32 = (cse_var_1 + 190)
-                let cse_var_200: int32 = (cse_var_1 + 19)
-                let cse_var_199: int32 = (cse_var_1 + 189)
-                let cse_var_198: int32 = (cse_var_1 + 188)
-                let cse_var_197: int32 = (cse_var_1 + 187)
-                let cse_var_196: int32 = (cse_var_1 + 186)
-                let cse_var_195: int32 = (cse_var_1 + 2)
-                let cse_var_194: int32 = (cse_var_1 + 126)
-                let cse_var_193: int32 = (cse_var_1 + 125)
-                let cse_var_192: int32 = (cse_var_1 + 124)
-                let cse_var_191: int32 = (cse_var_1 + 123)
-                let cse_var_190: int32 = (cse_var_1 + 122)
-                let cse_var_189: int32 = (cse_var_1 + 121)
-                let cse_var_188: int32 = (cse_var_1 + 120)
-                let cse_var_187: int32 = (cse_var_1 + 12)
-                let cse_var_186: int32 = (cse_var_1 + 119)
-                let cse_var_185: int32 = (cse_var_1 + 118)
-                let cse_var_184: int32 = (cse_var_1 + 117)
-                let cse_var_183: int32 = (cse_var_1 + 116)
-                let cse_var_182: int32 = (cse_var_1 + 115)
-                let cse_var_181: int32 = (cse_var_1 + 114)
-                let cse_var_180: int32 = (cse_var_1 + 113)
-                let cse_var_179: int32 = (cse_var_1 + 156)
-                let cse_var_178: int32 = (cse_var_1 + 111)
-                let cse_var_177: int32 = (cse_var_1 + 110)
-                let cse_var_176: int32 = (cse_var_1 + 11)
-                let cse_var_175: int32 = (cse_var_1 + 109)
-                let cse_var_174: int32 = (cse_var_1 + 108)
-                let cse_var_173: int32 = (cse_var_1 + 107)
-                let cse_var_172: int32 = (cse_var_1 + 106)
-                let cse_var_171: int32 = (cse_var_1 + 105)
-                let cse_var_170: int32 = (cse_var_1 + 104)
-                let cse_var_169: int32 = (cse_var_1 + 103)
-                let cse_var_168: int32 = (cse_var_1 + 102)
-                let cse_var_167: int32 = (cse_var_1 + 101)
-                let cse_var_166: int32 = (cse_var_1 + 100)
-                let cse_var_165: int32 = (cse_var_1 + 10)
-                let cse_var_164: int32 = (cse_var_1 + 1)
-                let cse_var_163: int32 = (cse_var_1 + 112)
-                let cse_var_162: int32 = (cse_var_1 + 155)
-                let cse_var_161: int32 = (cse_var_1 + 154)
-                let cse_var_160: int32 = (cse_var_1 + 153)
-                let cse_var_159: int32 = (cse_var_1 + 152)
-                let cse_var_158: int32 = (cse_var_1 + 151)
-                let cse_var_157: int32 = (cse_var_1 + 150)
-                let cse_var_156: int32 = (cse_var_1 + 15)
-                let cse_var_155: int32 = (cse_var_1 + 149)
-                let cse_var_154: int32 = (cse_var_1 + 148)
-                let cse_var_153: int32 = (cse_var_1 + 147)
-                let cse_var_152: int32 = (cse_var_1 + 146)
-                let cse_var_151: int32 = (cse_var_1 + 145)
-                let cse_var_150: int32 = (cse_var_1 + 144)
-                let cse_var_149: int32 = (cse_var_1 + 143)
-                let cse_var_148: int32 = (cse_var_1 + 142)
-                let cse_var_147: int32 = (cse_var_1 + 127)
-                let cse_var_146: int32 = (cse_var_1 + 140)
-                let cse_var_145: int32 = (cse_var_1 + 14)
-                let cse_var_144: int32 = (cse_var_1 + 139)
-                let cse_var_143: int32 = (cse_var_1 + 138)
-                let cse_var_142: int32 = (cse_var_1 + 137)
-                let cse_var_141: int32 = (cse_var_1 + 136)
-                let cse_var_140: int32 = (cse_var_1 + 135)
-                let cse_var_139: int32 = (cse_var_1 + 134)
-                let cse_var_138: int32 = (cse_var_1 + 133)
-                let cse_var_137: int32 = (cse_var_1 + 132)
-                let cse_var_136: int32 = (cse_var_1 + 131)
-                let cse_var_135: int32 = (cse_var_1 + 130)
-                let cse_var_134: int32 = (cse_var_1 + 13)
-                let cse_var_133: int32 = (cse_var_1 + 129)
-                let cse_var_132: int32 = (cse_var_1 + 128)
-                let cse_var_131: int32 = (cse_var_1 + 141)
-                let cse_var_130: int32 = (cse_var_1 + 70)
-                let cse_var_129: int32 = (cse_var_1 + 7)
-                let cse_var_128: int32 = (cse_var_1 + 69)
-                let cse_var_127: int32 = (cse_var_1 + 68)
-                let cse_var_126: int32 = (cse_var_1 + 67)
-                let cse_var_125: int32 = (cse_var_1 + 66)
-                let cse_var_124: int32 = (cse_var_1 + 65)
-                let cse_var_123: int32 = (cse_var_1 + 64)
-                let cse_var_122: int32 = (cse_var_1 + 63)
-                let cse_var_121: int32 = (cse_var_1 + 62)
-                let cse_var_120: int32 = (cse_var_1 + 61)
-                let cse_var_119: int32 = (cse_var_1 + 60)
-                let cse_var_118: int32 = (cse_var_1 + 6)
-                let cse_var_117: int32 = (cse_var_1 + 59)
-                let cse_var_116: int32 = (cse_var_1 + 58)
-                let cse_var_115: int32 = (cse_var_1 + 213)
-                let cse_var_114: int32 = (cse_var_1 + 56)
-                let cse_var_113: int32 = (cse_var_1 + 55)
-                let cse_var_112: int32 = (cse_var_1 + 54)
-                let cse_var_111: int32 = (cse_var_1 + 53)
-                let cse_var_110: int32 = (cse_var_1 + 52)
-                let cse_var_109: int32 = (cse_var_1 + 51)
-                let cse_var_108: int32 = (cse_var_1 + 50)
-                let cse_var_107: int32 = (cse_var_1 + 5)
-                let cse_var_106: int32 = (cse_var_1 + 49)
-                let cse_var_105: int32 = (cse_var_1 + 48)
-                let cse_var_104: int32 = (cse_var_1 + 47)
-                let cse_var_103: int32 = (cse_var_1 + 46)
-                let cse_var_102: int32 = (cse_var_1 + 45)
-                let cse_var_101: int32 = (cse_var_1 + 44)
-                let cse_var_100: int32 = (cse_var_1 + 43)
-                let cse_var_99: int32 = (cse_var_1 + 57)
-                let cse_var_98: int32 = (elem_idx*16)
-                let cse_var_97: int32 = (cse_var_1 + 99)
-                let cse_var_96: int32 = (cse_var_1 + 98)
-                let cse_var_95: int32 = (cse_var_1 + 97)
-                let cse_var_94: int32 = (cse_var_1 + 96)
-                let cse_var_93: int32 = (cse_var_1 + 95)
-                let cse_var_92: int32 = (cse_var_1 + 94)
-                let cse_var_91: int32 = (cse_var_1 + 93)
-                let cse_var_90: int32 = (cse_var_1 + 92)
-                let cse_var_89: int32 = (cse_var_1 + 91)
-                let cse_var_88: int32 = (cse_var_1 + 90)
-                let cse_var_87: int32 = (cse_var_1 + 9)
-                let cse_var_86: int32 = (cse_var_1 + 89)
-                let cse_var_85: int32 = (cse_var_1 + 88)
-                let cse_var_84: int32 = (cse_var_1 + 87)
-                let cse_var_83: int32 = (cse_var_1 + 71)
-                let cse_var_82: int32 = (cse_var_1 + 85)
-                let cse_var_81: int32 = (cse_var_1 + 84)
-                let cse_var_80: int32 = (cse_var_1 + 83)
-                let cse_var_79: int32 = (cse_var_1 + 82)
-                let cse_var_78: int32 = (cse_var_1 + 81)
-                let cse_var_77: int32 = (cse_var_1 + 80)
-                let cse_var_76: int32 = (cse_var_1 + 8)
-                let cse_var_75: int32 = (cse_var_1 + 79)
-                let cse_var_74: int32 = (cse_var_1 + 78)
-                let cse_var_73: int32 = (cse_var_1 + 77)
-                let cse_var_72: int32 = (cse_var_1 + 76)
-                let cse_var_71: int32 = (cse_var_1 + 75)
-                let cse_var_70: int32 = (cse_var_1 + 74)
-                let cse_var_69: int32 = (cse_var_1 + 73)
-                let cse_var_68: int32 = (cse_var_1 + 72)
-                let cse_var_67: int32 = (cse_var_1 + 86)
-                let cse_var_66: int32 = (cse_var_1 + 242)
-                let cse_var_65: int32 = (cse_var_1 + 241)
-                let cse_var_64: int32 = (cse_var_1 + 240)
-                let cse_var_63: int32 = (cse_var_1 + 24)
-                let cse_var_62: int32 = (cse_var_1 + 239)
-                let cse_var_61: int32 = (cse_var_1 + 238)
-                let cse_var_60: int32 = (cse_var_1 + 237)
-                let cse_var_59: int32 = (cse_var_1 + 236)
-                let cse_var_58: int32 = (cse_var_1 + 235)
-                let cse_var_57: int32 = (cse_var_1 + 234)
-                let cse_var_56: int32 = (cse_var_1 + 233)
-                let cse_var_55: int32 = (cse_var_1 + 232)
-                let cse_var_54: int32 = (cse_var_1 + 231)
-                let cse_var_53: int32 = (cse_var_1 + 230)
-                let cse_var_52: int32 = (cse_var_1 + 23)
-                let cse_var_51: int32 = (cse_var_1 + 243)
-                let cse_var_50: int32 = (cse_var_1 + 228)
-                let cse_var_49: int32 = (cse_var_1 + 227)
-                let cse_var_48: int32 = (cse_var_1 + 226)
-                let cse_var_47: int32 = (cse_var_1 + 225)
-                let cse_var_46: int32 = (cse_var_1 + 224)
-                let cse_var_45: int32 = (cse_var_1 + 223)
-                let cse_var_44: int32 = (cse_var_1 + 222)
-                let cse_var_43: int32 = (cse_var_1 + 221)
-                let cse_var_42: int32 = (cse_var_1 + 220)
-                let cse_var_41: int32 = (cse_var_1 + 22)
-                let cse_var_40: int32 = (cse_var_1 + 219)
-                let cse_var_39: int32 = (cse_var_1 + 218)
-                let cse_var_38: int32 = (cse_var_1 + 217)
-                let cse_var_37: int32 = (cse_var_1 + 216)
-                let cse_var_36: int32 = (cse_var_1 + 215)
-                let cse_var_35: int32 = (cse_var_1 + 229)
-                let cse_var_34: int32 = (cse_var_1 + 42)
-                let cse_var_33: int32 = (cse_var_1 + 40)
-                let cse_var_32: int32 = (cse_var_1 + 4)
-                let cse_var_31: int32 = (cse_var_1 + 39)
-                let cse_var_30: int32 = (cse_var_1 + 38)
-                let cse_var_29: int32 = (cse_var_1 + 37)
-                let cse_var_28: int32 = (cse_var_1 + 36)
-                let cse_var_27: int32 = (cse_var_1 + 35)
-                let cse_var_26: int32 = (cse_var_1 + 34)
-                let cse_var_25: int32 = (cse_var_1 + 33)
-                let cse_var_24: int32 = (cse_var_1 + 32)
-                let cse_var_23: int32 = (cse_var_1 + 31)
-                let cse_var_22: int32 = (cse_var_1 + 30)
-                let cse_var_21: int32 = (cse_var_1 + 3)
-                let cse_var_20: int32 = (cse_var_1 + 29)
-                let cse_var_19: int32 = (cse_var_1 + 28)
-                let cse_var_18: int32 = (cse_var_1 + 41)
-                let cse_var_17: int32 = (cse_var_1 + 245)
-                let cse_var_16: int32 = (cse_var_1 + 246)
-                let cse_var_15: int32 = (cse_var_1 + 247)
-                let cse_var_14: int32 = (cse_var_1 + 248)
-                let cse_var_13: int32 = (cse_var_1 + 249)
-                let cse_var_12: int32 = (cse_var_1 + 25)
-                let cse_var_11: int32 = (cse_var_1 + 250)
-                let cse_var_10: int32 = (cse_var_1 + 251)
-                let cse_var_9: int32 = (cse_var_1 + 252)
-                let cse_var_8: int32 = (cse_var_1 + 253)
-                let cse_var_7: int32 = (cse_var_1 + 254)
-                let cse_var_6: int32 = (cse_var_1 + 255)
-                let cse_var_5: int32 = (cse_var_1 + 26)
-                let cse_var_4: int32 = (cse_var_1 + 244)
-                let cse_var_3: int32 = (cse_var_1 + 27)
-                let cse_var_2: int32 = ((i0.outer*16384) + (i.outer.inner*4096))
+              compute_4: Buffer(compute_3, float32, [256], [])[cse_var_2] = 0f32
+              compute_4[(cse_var_2 + 1)] = 0f32
+              compute_4[(cse_var_2 + 2)] = 0f32
+              compute_4[(cse_var_2 + 3)] = 0f32
+              compute_4[(cse_var_2 + 4)] = 0f32
+              compute_4[(cse_var_2 + 5)] = 0f32
+              compute_4[(cse_var_2 + 6)] = 0f32
+              compute_4[(cse_var_2 + 7)] = 0f32
+              compute_4[(cse_var_2 + 8)] = 0f32
+              compute_4[(cse_var_2 + 9)] = 0f32
+              compute_4[(cse_var_2 + 10)] = 0f32
+              compute_4[(cse_var_2 + 11)] = 0f32
+              compute_4[(cse_var_2 + 12)] = 0f32
+              compute_4[(cse_var_2 + 13)] = 0f32
+              compute_4[(cse_var_2 + 14)] = 0f32
+              compute_4[(cse_var_2 + 15)] = 0f32
+              compute_4[(cse_var_2 + 32)] = 0f32
+              compute_4[(cse_var_2 + 33)] = 0f32
+              compute_4[(cse_var_2 + 34)] = 0f32
+              compute_4[(cse_var_2 + 35)] = 0f32
+              compute_4[(cse_var_2 + 36)] = 0f32
+              compute_4[(cse_var_2 + 37)] = 0f32
+              compute_4[(cse_var_2 + 38)] = 0f32
+              compute_4[(cse_var_2 + 39)] = 0f32
+              compute_4[(cse_var_2 + 40)] = 0f32
+              compute_4[(cse_var_2 + 41)] = 0f32
+              compute_4[(cse_var_2 + 42)] = 0f32
+              compute_4[(cse_var_2 + 43)] = 0f32
+              compute_4[(cse_var_2 + 44)] = 0f32
+              compute_4[(cse_var_2 + 45)] = 0f32
+              compute_4[(cse_var_2 + 46)] = 0f32
+              compute_4[(cse_var_2 + 47)] = 0f32
+              compute_4[(cse_var_2 + 64)] = 0f32
+              compute_4[(cse_var_2 + 65)] = 0f32
+              compute_4[(cse_var_2 + 66)] = 0f32
+              compute_4[(cse_var_2 + 67)] = 0f32
+              compute_4[(cse_var_2 + 68)] = 0f32
+              compute_4[(cse_var_2 + 69)] = 0f32
+              compute_4[(cse_var_2 + 70)] = 0f32
+              compute_4[(cse_var_2 + 71)] = 0f32
+              compute_4[(cse_var_2 + 72)] = 0f32
+              compute_4[(cse_var_2 + 73)] = 0f32
+              compute_4[(cse_var_2 + 74)] = 0f32
+              compute_4[(cse_var_2 + 75)] = 0f32
+              compute_4[(cse_var_2 + 76)] = 0f32
+              compute_4[(cse_var_2 + 77)] = 0f32
+              compute_4[(cse_var_2 + 78)] = 0f32
+              compute_4[(cse_var_2 + 79)] = 0f32
+              compute_4[(cse_var_2 + 96)] = 0f32
+              compute_4[(cse_var_2 + 97)] = 0f32
+              compute_4[(cse_var_2 + 98)] = 0f32
+              compute_4[(cse_var_2 + 99)] = 0f32
+              compute_4[(cse_var_2 + 100)] = 0f32
+              compute_4[(cse_var_2 + 101)] = 0f32
+              compute_4[(cse_var_2 + 102)] = 0f32
+              compute_4[(cse_var_2 + 103)] = 0f32
+              compute_4[(cse_var_2 + 104)] = 0f32
+              compute_4[(cse_var_2 + 105)] = 0f32
+              compute_4[(cse_var_2 + 106)] = 0f32
+              compute_4[(cse_var_2 + 107)] = 0f32
+              compute_4[(cse_var_2 + 108)] = 0f32
+              compute_4[(cse_var_2 + 109)] = 0f32
+              compute_4[(cse_var_2 + 110)] = 0f32
+              compute_4[(cse_var_2 + 111)] = 0f32
+              compute_4[(cse_var_2 + 128)] = 0f32
+              compute_4[(cse_var_2 + 129)] = 0f32
+              compute_4[(cse_var_2 + 130)] = 0f32
+              compute_4[(cse_var_2 + 131)] = 0f32
+              compute_4[(cse_var_2 + 132)] = 0f32
+              compute_4[(cse_var_2 + 133)] = 0f32
+              compute_4[(cse_var_2 + 134)] = 0f32
+              compute_4[(cse_var_2 + 135)] = 0f32
+              compute_4[(cse_var_2 + 136)] = 0f32
+              compute_4[(cse_var_2 + 137)] = 0f32
+              compute_4[(cse_var_2 + 138)] = 0f32
+              compute_4[(cse_var_2 + 139)] = 0f32
+              compute_4[(cse_var_2 + 140)] = 0f32
+              compute_4[(cse_var_2 + 141)] = 0f32
+              compute_4[(cse_var_2 + 142)] = 0f32
+              compute_4[(cse_var_2 + 143)] = 0f32
+              compute_4[(cse_var_2 + 160)] = 0f32
+              compute_4[(cse_var_2 + 161)] = 0f32
+              compute_4[(cse_var_2 + 162)] = 0f32
+              compute_4[(cse_var_2 + 163)] = 0f32
+              compute_4[(cse_var_2 + 164)] = 0f32
+              compute_4[(cse_var_2 + 165)] = 0f32
+              compute_4[(cse_var_2 + 166)] = 0f32
+              compute_4[(cse_var_2 + 167)] = 0f32
+              compute_4[(cse_var_2 + 168)] = 0f32
+              compute_4[(cse_var_2 + 169)] = 0f32
+              compute_4[(cse_var_2 + 170)] = 0f32
+              compute_4[(cse_var_2 + 171)] = 0f32
+              compute_4[(cse_var_2 + 172)] = 0f32
+              compute_4[(cse_var_2 + 173)] = 0f32
+              compute_4[(cse_var_2 + 174)] = 0f32
+              compute_4[(cse_var_2 + 175)] = 0f32
+              compute_4[(cse_var_2 + 192)] = 0f32
+              compute_4[(cse_var_2 + 193)] = 0f32
+              compute_4[(cse_var_2 + 194)] = 0f32
+              compute_4[(cse_var_2 + 195)] = 0f32
+              compute_4[(cse_var_2 + 196)] = 0f32
+              compute_4[(cse_var_2 + 197)] = 0f32
+              compute_4[(cse_var_2 + 198)] = 0f32
+              compute_4[(cse_var_2 + 199)] = 0f32
+              compute_4[(cse_var_2 + 200)] = 0f32
+              compute_4[(cse_var_2 + 201)] = 0f32
+              compute_4[(cse_var_2 + 202)] = 0f32
+              compute_4[(cse_var_2 + 203)] = 0f32
+              compute_4[(cse_var_2 + 204)] = 0f32
+              compute_4[(cse_var_2 + 205)] = 0f32
+              compute_4[(cse_var_2 + 206)] = 0f32
+              compute_4[(cse_var_2 + 207)] = 0f32
+              compute_4[(cse_var_2 + 224)] = 0f32
+              compute_4[(cse_var_2 + 225)] = 0f32
+              compute_4[(cse_var_2 + 226)] = 0f32
+              compute_4[(cse_var_2 + 227)] = 0f32
+              compute_4[(cse_var_2 + 228)] = 0f32
+              compute_4[(cse_var_2 + 229)] = 0f32
+              compute_4[(cse_var_2 + 230)] = 0f32
+              compute_4[(cse_var_2 + 231)] = 0f32
+              compute_4[(cse_var_2 + 232)] = 0f32
+              compute_4[(cse_var_2 + 233)] = 0f32
+              compute_4[(cse_var_2 + 234)] = 0f32
+              compute_4[(cse_var_2 + 235)] = 0f32
+              compute_4[(cse_var_2 + 236)] = 0f32
+              compute_4[(cse_var_2 + 237)] = 0f32
+              compute_4[(cse_var_2 + 238)] = 0f32
+              compute_4[(cse_var_2 + 239)] = 0f32
+              for (elem_idx: int32, 0, (placeholder_3[(cse_var_1 + 1)] - placeholder_3[cse_var_1])) {
+                let cse_var_131: int32 = (cse_var_2 + 143)
+                let cse_var_130: int32 = (cse_var_2 + 15)
+                let cse_var_129: int32 = (cse_var_2 + 160)
+                let cse_var_128: int32 = (cse_var_2 + 161)
+                let cse_var_127: int32 = (cse_var_2 + 162)
+                let cse_var_126: int32 = (cse_var_2 + 163)
+                let cse_var_125: int32 = (cse_var_2 + 164)
+                let cse_var_124: int32 = (cse_var_2 + 165)
+                let cse_var_123: int32 = (cse_var_2 + 166)
+                let cse_var_122: int32 = (cse_var_2 + 167)
+                let cse_var_121: int32 = (cse_var_2 + 168)
+                let cse_var_120: int32 = (cse_var_2 + 169)
+                let cse_var_119: int32 = (cse_var_2 + 170)
+                let cse_var_118: int32 = (cse_var_2 + 171)
+                let cse_var_117: int32 = (cse_var_2 + 172)
+                let cse_var_116: int32 = (cse_var_2 + 1)
+                let cse_var_115: int32 = (cse_var_2 + 174)
+                let cse_var_114: int32 = (cse_var_2 + 175)
+                let cse_var_113: int32 = (cse_var_2 + 192)
+                let cse_var_112: int32 = (cse_var_2 + 193)
+                let cse_var_111: int32 = (cse_var_2 + 194)
+                let cse_var_110: int32 = (cse_var_2 + 195)
+                let cse_var_109: int32 = (cse_var_2 + 196)
+                let cse_var_108: int32 = (cse_var_2 + 197)
+                let cse_var_107: int32 = (cse_var_2 + 198)
+                let cse_var_106: int32 = (cse_var_2 + 199)
+                let cse_var_105: int32 = (cse_var_2 + 2)
+                let cse_var_104: int32 = (cse_var_2 + 200)
+                let cse_var_103: int32 = (cse_var_2 + 201)
+                let cse_var_102: int32 = (cse_var_2 + 202)
+                let cse_var_101: int32 = (cse_var_2 + 203)
+                let cse_var_100: int32 = (cse_var_2 + 173)
+                let cse_var_99: int32 = (cse_var_2 + 10)
+                let cse_var_98: int32 = (cse_var_2 + 100)
+                let cse_var_97: int32 = (cse_var_2 + 101)
+                let cse_var_96: int32 = (cse_var_2 + 102)
+                let cse_var_95: int32 = (cse_var_2 + 103)
+                let cse_var_94: int32 = (cse_var_2 + 104)
+                let cse_var_93: int32 = (cse_var_2 + 105)
+                let cse_var_92: int32 = (cse_var_2 + 106)
+                let cse_var_91: int32 = (cse_var_2 + 107)
+                let cse_var_90: int32 = (cse_var_2 + 108)
+                let cse_var_89: int32 = (cse_var_2 + 109)
+                let cse_var_88: int32 = (cse_var_2 + 11)
+                let cse_var_87: int32 = (cse_var_2 + 110)
+                let cse_var_86: int32 = (cse_var_2 + 111)
+                let cse_var_85: int32 = (cse_var_2 + 12)
+                let cse_var_84: int32 = (cse_var_2 + 142)
+                let cse_var_83: int32 = (cse_var_2 + 129)
+                let cse_var_82: int32 = (cse_var_2 + 13)
+                let cse_var_81: int32 = (cse_var_2 + 130)
+                let cse_var_80: int32 = (cse_var_2 + 131)
+                let cse_var_79: int32 = (cse_var_2 + 132)
+                let cse_var_78: int32 = (cse_var_2 + 133)
+                let cse_var_77: int32 = (cse_var_2 + 134)
+                let cse_var_76: int32 = (cse_var_2 + 135)
+                let cse_var_75: int32 = (cse_var_2 + 136)
+                let cse_var_74: int32 = (cse_var_2 + 137)
+                let cse_var_73: int32 = (cse_var_2 + 138)
+                let cse_var_72: int32 = (cse_var_2 + 139)
+                let cse_var_71: int32 = (cse_var_2 + 14)
+                let cse_var_70: int32 = (cse_var_2 + 140)
+                let cse_var_69: int32 = (cse_var_2 + 141)
+                let cse_var_68: int32 = (cse_var_2 + 128)
+                let cse_var_67: int32 = (cse_var_2 + 44)
+                let cse_var_66: int32 = (cse_var_2 + 45)
+                let cse_var_65: int32 = (cse_var_2 + 46)
+                let cse_var_64: int32 = (cse_var_2 + 47)
+                let cse_var_63: int32 = (cse_var_2 + 5)
+                let cse_var_62: int32 = (cse_var_2 + 6)
+                let cse_var_61: int32 = (cse_var_2 + 64)
+                let cse_var_60: int32 = (cse_var_2 + 65)
+                let cse_var_59: int32 = (cse_var_2 + 66)
+                let cse_var_58: int32 = (cse_var_2 + 67)
+                let cse_var_57: int32 = (cse_var_2 + 68)
+                let cse_var_56: int32 = (cse_var_2 + 69)
+                let cse_var_55: int32 = (cse_var_2 + 7)
+                let cse_var_54: int32 = (cse_var_2 + 70)
+                let cse_var_53: int32 = (cse_var_2 + 71)
+                let cse_var_52: int32 = (cse_var_2 + 204)
+                let cse_var_51: int32 = (cse_var_2 + 73)
+                let cse_var_50: int32 = (cse_var_2 + 74)
+                let cse_var_49: int32 = (cse_var_2 + 75)
+                let cse_var_48: int32 = (cse_var_2 + 76)
+                let cse_var_47: int32 = (cse_var_2 + 77)
+                let cse_var_46: int32 = (cse_var_2 + 78)
+                let cse_var_45: int32 = (cse_var_2 + 79)
+                let cse_var_44: int32 = (cse_var_2 + 8)
+                let cse_var_43: int32 = (cse_var_2 + 9)
+                let cse_var_42: int32 = (cse_var_2 + 96)
+                let cse_var_41: int32 = (cse_var_2 + 97)
+                let cse_var_40: int32 = (cse_var_2 + 98)
+                let cse_var_39: int32 = (cse_var_2 + 99)
+                let cse_var_38: int32 = (elem_idx*16)
+                let cse_var_37: int32 = (i0.outer*2048)
+                let cse_var_36: int32 = (cse_var_2 + 72)
+                let cse_var_35: int32 = (cse_var_2 + 205)
+                let cse_var_34: int32 = (cse_var_2 + 206)
+                let cse_var_33: int32 = (cse_var_2 + 207)
+                let cse_var_32: int32 = (cse_var_2 + 224)
+                let cse_var_31: int32 = (cse_var_2 + 225)
+                let cse_var_30: int32 = (cse_var_2 + 226)
+                let cse_var_29: int32 = (cse_var_2 + 227)
+                let cse_var_28: int32 = (cse_var_2 + 228)
+                let cse_var_27: int32 = (cse_var_2 + 229)
+                let cse_var_26: int32 = (cse_var_2 + 230)
+                let cse_var_25: int32 = (cse_var_2 + 231)
+                let cse_var_24: int32 = (cse_var_2 + 232)
+                let cse_var_23: int32 = (cse_var_2 + 233)
+                let cse_var_22: int32 = (cse_var_2 + 234)
+                let cse_var_21: int32 = (cse_var_2 + 235)
+                let cse_var_20: int32 = (cse_var_2 + 43)
+                let cse_var_19: int32 = (cse_var_2 + 42)
+                let cse_var_18: int32 = (cse_var_2 + 41)
+                let cse_var_17: int32 = (cse_var_2 + 40)
+                let cse_var_16: int32 = (cse_var_2 + 4)
+                let cse_var_15: int32 = (cse_var_2 + 39)
+                let cse_var_14: int32 = (cse_var_2 + 38)
+                let cse_var_13: int32 = (cse_var_2 + 37)
+                let cse_var_12: int32 = (cse_var_2 + 236)
+                let cse_var_11: int32 = (cse_var_2 + 35)
+                let cse_var_10: int32 = (cse_var_2 + 34)
+                let cse_var_9: int32 = (cse_var_2 + 33)
+                let cse_var_8: int32 = (cse_var_2 + 32)
+                let cse_var_7: int32 = (cse_var_2 + 3)
+                let cse_var_6: int32 = (cse_var_2 + 239)
+                let cse_var_5: int32 = (cse_var_2 + 238)
+                let cse_var_4: int32 = (cse_var_2 + 237)
+                let cse_var_3: int32 = (cse_var_2 + 36)
                  {
-                  compute_4[cse_var_1] = (compute_4[cse_var_1] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_164] = (compute_4[cse_var_164] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_195] = (compute_4[cse_var_195] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_21] = (compute_4[cse_var_21] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_32] = (compute_4[cse_var_32] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_107] = (compute_4[cse_var_107] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_118] = (compute_4[cse_var_118] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_129] = (compute_4[cse_var_129] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_76] = (compute_4[cse_var_76] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_87] = (compute_4[cse_var_87] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_165] = (compute_4[cse_var_165] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_176] = (compute_4[cse_var_176] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_187] = (compute_4[cse_var_187] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_134] = (compute_4[cse_var_134] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_145] = (compute_4[cse_var_145] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_156] = (compute_4[cse_var_156] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_231] = (compute_4[cse_var_231] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_242] = (compute_4[cse_var_242] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_253] = (compute_4[cse_var_253] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_200] = (compute_4[cse_var_200] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_212] = (compute_4[cse_var_212] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_223] = (compute_4[cse_var_223] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_41] = (compute_4[cse_var_41] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_52] = (compute_4[cse_var_52] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_63] = (compute_4[cse_var_63] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_22] = (compute_4[cse_var_22] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_23] = (compute_4[cse_var_23] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-                  compute_4[cse_var_24] = (compute_4[cse_var_24] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_25] = (compute_4[cse_var_25] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_26] = (compute_4[cse_var_26] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_27] = (compute_4[cse_var_27] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_28] = (compute_4[cse_var_28] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_29] = (compute_4[cse_var_29] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_30] = (compute_4[cse_var_30] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_31] = (compute_4[cse_var_31] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_33] = (compute_4[cse_var_33] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_18] = (compute_4[cse_var_18] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_34] = (compute_4[cse_var_34] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_100] = (compute_4[cse_var_100] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_101] = (compute_4[cse_var_101] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_102] = (compute_4[cse_var_102] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_103] = (compute_4[cse_var_103] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_104] = (compute_4[cse_var_104] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-                  compute_4[cse_var_105] = (compute_4[cse_var_105] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_106] = (compute_4[cse_var_106] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_108] = (compute_4[cse_var_108] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_109] = (compute_4[cse_var_109] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_110] = (compute_4[cse_var_110] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_111] = (compute_4[cse_var_111] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_112] = (compute_4[cse_var_112] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_113] = (compute_4[cse_var_113] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_114] = (compute_4[cse_var_114] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_99] = (compute_4[cse_var_99] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_116] = (compute_4[cse_var_116] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_117] = (compute_4[cse_var_117] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_119] = (compute_4[cse_var_119] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_120] = (compute_4[cse_var_120] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_121] = (compute_4[cse_var_121] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_122] = (compute_4[cse_var_122] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-                  compute_4[cse_var_123] = (compute_4[cse_var_123] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_124] = (compute_4[cse_var_124] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_125] = (compute_4[cse_var_125] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_126] = (compute_4[cse_var_126] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_127] = (compute_4[cse_var_127] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_128] = (compute_4[cse_var_128] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_130] = (compute_4[cse_var_130] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_83] = (compute_4[cse_var_83] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_68] = (compute_4[cse_var_68] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_69] = (compute_4[cse_var_69] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_70] = (compute_4[cse_var_70] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_71] = (compute_4[cse_var_71] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_72] = (compute_4[cse_var_72] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_73] = (compute_4[cse_var_73] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_74] = (compute_4[cse_var_74] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_75] = (compute_4[cse_var_75] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-                  compute_4[cse_var_77] = (compute_4[cse_var_77] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_78] = (compute_4[cse_var_78] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_79] = (compute_4[cse_var_79] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_80] = (compute_4[cse_var_80] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_81] = (compute_4[cse_var_81] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_82] = (compute_4[cse_var_82] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_67] = (compute_4[cse_var_67] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_84] = (compute_4[cse_var_84] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_85] = (compute_4[cse_var_85] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_86] = (compute_4[cse_var_86] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_88] = (compute_4[cse_var_88] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_89] = (compute_4[cse_var_89] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_90] = (compute_4[cse_var_90] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_91] = (compute_4[cse_var_91] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_92] = (compute_4[cse_var_92] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_93] = (compute_4[cse_var_93] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-                  compute_4[cse_var_94] = (compute_4[cse_var_94] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_95] = (compute_4[cse_var_95] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_96] = (compute_4[cse_var_96] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_97] = (compute_4[cse_var_97] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_166] = (compute_4[cse_var_166] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_167] = (compute_4[cse_var_167] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_168] = (compute_4[cse_var_168] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_169] = (compute_4[cse_var_169] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_170] = (compute_4[cse_var_170] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_171] = (compute_4[cse_var_171] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_172] = (compute_4[cse_var_172] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_173] = (compute_4[cse_var_173] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_174] = (compute_4[cse_var_174] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_175] = (compute_4[cse_var_175] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_177] = (compute_4[cse_var_177] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_178] = (compute_4[cse_var_178] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-                  compute_4[cse_var_163] = (compute_4[cse_var_163] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_180] = (compute_4[cse_var_180] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_181] = (compute_4[cse_var_181] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_182] = (compute_4[cse_var_182] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_183] = (compute_4[cse_var_183] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_184] = (compute_4[cse_var_184] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_185] = (compute_4[cse_var_185] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_186] = (compute_4[cse_var_186] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_188] = (compute_4[cse_var_188] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_189] = (compute_4[cse_var_189] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_190] = (compute_4[cse_var_190] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_191] = (compute_4[cse_var_191] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_192] = (compute_4[cse_var_192] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_193] = (compute_4[cse_var_193] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_194] = (compute_4[cse_var_194] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_147] = (compute_4[cse_var_147] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-                  compute_4[cse_var_132] = (compute_4[cse_var_132] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_133] = (compute_4[cse_var_133] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_135] = (compute_4[cse_var_135] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_136] = (compute_4[cse_var_136] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_137] = (compute_4[cse_var_137] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_138] = (compute_4[cse_var_138] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_139] = (compute_4[cse_var_139] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_140] = (compute_4[cse_var_140] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_141] = (compute_4[cse_var_141] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_142] = (compute_4[cse_var_142] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_143] = (compute_4[cse_var_143] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_144] = (compute_4[cse_var_144] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_146] = (compute_4[cse_var_146] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_131] = (compute_4[cse_var_131] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_148] = (compute_4[cse_var_148] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_149] = (compute_4[cse_var_149] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-                  compute_4[cse_var_150] = (compute_4[cse_var_150] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_151] = (compute_4[cse_var_151] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_152] = (compute_4[cse_var_152] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_153] = (compute_4[cse_var_153] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_154] = (compute_4[cse_var_154] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_155] = (compute_4[cse_var_155] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_157] = (compute_4[cse_var_157] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_158] = (compute_4[cse_var_158] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_159] = (compute_4[cse_var_159] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_160] = (compute_4[cse_var_160] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_161] = (compute_4[cse_var_161] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_162] = (compute_4[cse_var_162] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_179] = (compute_4[cse_var_179] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_228] = (compute_4[cse_var_228] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_229] = (compute_4[cse_var_229] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_230] = (compute_4[cse_var_230] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-                  compute_4[cse_var_232] = (compute_4[cse_var_232] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_233] = (compute_4[cse_var_233] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_234] = (compute_4[cse_var_234] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_235] = (compute_4[cse_var_235] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_236] = (compute_4[cse_var_236] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_237] = (compute_4[cse_var_237] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_238] = (compute_4[cse_var_238] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_239] = (compute_4[cse_var_239] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_240] = (compute_4[cse_var_240] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_241] = (compute_4[cse_var_241] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_227] = (compute_4[cse_var_227] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_244] = (compute_4[cse_var_244] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_245] = (compute_4[cse_var_245] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_246] = (compute_4[cse_var_246] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_247] = (compute_4[cse_var_247] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_248] = (compute_4[cse_var_248] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-                  compute_4[cse_var_249] = (compute_4[cse_var_249] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_250] = (compute_4[cse_var_250] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_251] = (compute_4[cse_var_251] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_252] = (compute_4[cse_var_252] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_254] = (compute_4[cse_var_254] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_255] = (compute_4[cse_var_255] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_256] = (compute_4[cse_var_256] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_257] = (compute_4[cse_var_257] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_258] = (compute_4[cse_var_258] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_211] = (compute_4[cse_var_211] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_196] = (compute_4[cse_var_196] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_197] = (compute_4[cse_var_197] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_198] = (compute_4[cse_var_198] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_199] = (compute_4[cse_var_199] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_201] = (compute_4[cse_var_201] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_202] = (compute_4[cse_var_202] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-                  compute_4[cse_var_203] = (compute_4[cse_var_203] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_204] = (compute_4[cse_var_204] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_205] = (compute_4[cse_var_205] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_206] = (compute_4[cse_var_206] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_207] = (compute_4[cse_var_207] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_208] = (compute_4[cse_var_208] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_209] = (compute_4[cse_var_209] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_210] = (compute_4[cse_var_210] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_213] = (compute_4[cse_var_213] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_214] = (compute_4[cse_var_214] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_215] = (compute_4[cse_var_215] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_216] = (compute_4[cse_var_216] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_217] = (compute_4[cse_var_217] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_218] = (compute_4[cse_var_218] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_219] = (compute_4[cse_var_219] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_220] = (compute_4[cse_var_220] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-                  compute_4[cse_var_221] = (compute_4[cse_var_221] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_222] = (compute_4[cse_var_222] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_224] = (compute_4[cse_var_224] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_225] = (compute_4[cse_var_225] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_226] = (compute_4[cse_var_226] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_115] = (compute_4[cse_var_115] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_243] = (compute_4[cse_var_243] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_36] = (compute_4[cse_var_36] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_37] = (compute_4[cse_var_37] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_38] = (compute_4[cse_var_38] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_39] = (compute_4[cse_var_39] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_40] = (compute_4[cse_var_40] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_42] = (compute_4[cse_var_42] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_43] = (compute_4[cse_var_43] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_44] = (compute_4[cse_var_44] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_45] = (compute_4[cse_var_45] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-                  compute_4[cse_var_46] = (compute_4[cse_var_46] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_47] = (compute_4[cse_var_47] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_48] = (compute_4[cse_var_48] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_49] = (compute_4[cse_var_49] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_50] = (compute_4[cse_var_50] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_35] = (compute_4[cse_var_35] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_53] = (compute_4[cse_var_53] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_54] = (compute_4[cse_var_54] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_55] = (compute_4[cse_var_55] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_56] = (compute_4[cse_var_56] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_57] = (compute_4[cse_var_57] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_58] = (compute_4[cse_var_58] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_59] = (compute_4[cse_var_59] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_60] = (compute_4[cse_var_60] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_61] = (compute_4[cse_var_61] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_62] = (compute_4[cse_var_62] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-                  compute_4[cse_var_64] = (compute_4[cse_var_64] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_65] = (compute_4[cse_var_65] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_66] = (compute_4[cse_var_66] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_51] = (compute_4[cse_var_51] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-                  compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
+                  compute_4[cse_var_2] = (compute_4[cse_var_2] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_116] = (compute_4[cse_var_116] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_105] = (compute_4[cse_var_105] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_63] = (compute_4[cse_var_63] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_62] = (compute_4[cse_var_62] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_55] = (compute_4[cse_var_55] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_44] = (compute_4[cse_var_44] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_43] = (compute_4[cse_var_43] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_99] = (compute_4[cse_var_99] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_88] = (compute_4[cse_var_88] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_85] = (compute_4[cse_var_85] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_82] = (compute_4[cse_var_82] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_71] = (compute_4[cse_var_71] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_130] = (compute_4[cse_var_130] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+                  compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_18] = (compute_4[cse_var_18] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_67] = (compute_4[cse_var_67] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_66] = (compute_4[cse_var_66] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_65] = (compute_4[cse_var_65] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_64] = (compute_4[cse_var_64] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+                  compute_4[cse_var_61] = (compute_4[cse_var_61] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_60] = (compute_4[cse_var_60] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_59] = (compute_4[cse_var_59] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_58] = (compute_4[cse_var_58] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_57] = (compute_4[cse_var_57] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_56] = (compute_4[cse_var_56] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_54] = (compute_4[cse_var_54] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_53] = (compute_4[cse_var_53] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_36] = (compute_4[cse_var_36] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_51] = (compute_4[cse_var_51] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_50] = (compute_4[cse_var_50] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_49] = (compute_4[cse_var_49] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_48] = (compute_4[cse_var_48] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_47] = (compute_4[cse_var_47] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_46] = (compute_4[cse_var_46] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_45] = (compute_4[cse_var_45] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+                  compute_4[cse_var_42] = (compute_4[cse_var_42] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_41] = (compute_4[cse_var_41] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_40] = (compute_4[cse_var_40] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_39] = (compute_4[cse_var_39] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_98] = (compute_4[cse_var_98] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_97] = (compute_4[cse_var_97] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_96] = (compute_4[cse_var_96] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_95] = (compute_4[cse_var_95] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_94] = (compute_4[cse_var_94] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_93] = (compute_4[cse_var_93] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_92] = (compute_4[cse_var_92] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_91] = (compute_4[cse_var_91] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_90] = (compute_4[cse_var_90] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_89] = (compute_4[cse_var_89] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_87] = (compute_4[cse_var_87] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_86] = (compute_4[cse_var_86] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+                  compute_4[cse_var_68] = (compute_4[cse_var_68] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_83] = (compute_4[cse_var_83] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_81] = (compute_4[cse_var_81] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_80] = (compute_4[cse_var_80] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_79] = (compute_4[cse_var_79] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_78] = (compute_4[cse_var_78] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_77] = (compute_4[cse_var_77] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_76] = (compute_4[cse_var_76] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_75] = (compute_4[cse_var_75] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_74] = (compute_4[cse_var_74] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_73] = (compute_4[cse_var_73] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_72] = (compute_4[cse_var_72] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_70] = (compute_4[cse_var_70] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_69] = (compute_4[cse_var_69] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_84] = (compute_4[cse_var_84] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_131] = (compute_4[cse_var_131] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+                  compute_4[cse_var_129] = (compute_4[cse_var_129] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_128] = (compute_4[cse_var_128] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_127] = (compute_4[cse_var_127] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_126] = (compute_4[cse_var_126] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_125] = (compute_4[cse_var_125] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_124] = (compute_4[cse_var_124] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_123] = (compute_4[cse_var_123] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_122] = (compute_4[cse_var_122] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_121] = (compute_4[cse_var_121] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_120] = (compute_4[cse_var_120] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_119] = (compute_4[cse_var_119] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_118] = (compute_4[cse_var_118] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_117] = (compute_4[cse_var_117] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_100] = (compute_4[cse_var_100] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_115] = (compute_4[cse_var_115] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_114] = (compute_4[cse_var_114] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+                  compute_4[cse_var_113] = (compute_4[cse_var_113] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_112] = (compute_4[cse_var_112] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_111] = (compute_4[cse_var_111] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_110] = (compute_4[cse_var_110] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_109] = (compute_4[cse_var_109] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_108] = (compute_4[cse_var_108] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_107] = (compute_4[cse_var_107] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_106] = (compute_4[cse_var_106] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_104] = (compute_4[cse_var_104] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_103] = (compute_4[cse_var_103] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_102] = (compute_4[cse_var_102] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_101] = (compute_4[cse_var_101] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_52] = (compute_4[cse_var_52] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_35] = (compute_4[cse_var_35] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_34] = (compute_4[cse_var_34] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_33] = (compute_4[cse_var_33] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+                  compute_4[cse_var_32] = (compute_4[cse_var_32] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_31] = (compute_4[cse_var_31] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_30] = (compute_4[cse_var_30] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_29] = (compute_4[cse_var_29] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_28] = (compute_4[cse_var_28] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_27] = (compute_4[cse_var_27] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_26] = (compute_4[cse_var_26] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_25] = (compute_4[cse_var_25] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_24] = (compute_4[cse_var_24] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_23] = (compute_4[cse_var_23] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_22] = (compute_4[cse_var_22] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_21] = (compute_4[cse_var_21] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+                  compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
                 }
               }
             }
           }
-          for (i0.inner: int32, 0, 64) {
-            let cse_var_259: int32 = (((i0.outer*32768) + (i0.inner*512)) + (i1.outer*16))
-            compute[ramp(cse_var_259, 1, 16)] = max((compute_4[ramp((i0.inner*16), 1, 16)] + placeholder_4[ramp(cse_var_259, 1, 16)]), broadcast(0f32, 16))
+          for (i0.inner: int32, 0, 8) {
+            let cse_var_132: int32 = (((i0.outer*4096) + (i0.inner*512)) + (i1.outer*32))
+            compute[ramp(cse_var_132, 1, 32)] = max((compute_4[ramp((i0.inner*32), 1, 32)] + placeholder_4[ramp(cse_var_132, 1, 32)]), broadcast(0f32, 32))
           }
         }
       }
@@ -1199,7 +816,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 3.337 ms
+    Execution time of this operator: 2.706 ms
 
 
 
diff --git a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
index f51ba922e..4f5f0a1a9 100644
--- a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**00:43.418** total execution time for **how_to_tune_with_autotvm** files:
+**00:43.955** total execution time for **how_to_tune_with_autotvm** files:
 
-- **00:42.622**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)
-- **00:00.209**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)
-- **00:00.202**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``)
-- **00:00.192**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)
-- **00:00.192**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``)
+- **00:43.085**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)
+- **00:00.228**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)
+- **00:00.216**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)
+- **00:00.213**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``)
+- **00:00.213**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``)
diff --git a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
index 1cca2eb4c..f28f2ec32 100644
--- a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
@@ -859,8 +859,8 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 4, 4, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2885496
-    No: 6   GFLOPS: 103.50/103.50   result: MeasureResult(costs=(0.002236654708333333,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.592435359954834, timestamp=1650063815.6699162)        [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
-    No: 7   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 6   GFLOPS: 103.76/103.76   result: MeasureResult(costs=(0.0022312065416666667,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5933806896209717, timestamp=1650065132.1649663)      [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
+    No: 7   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -983,7 +983,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 16, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6225319
-    No: 8   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 8   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1106,7 +1106,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 1, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,943546
-    No: 9   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 9   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1229,7 +1229,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 4, 16, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2868708
-    No: 10  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 10  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 142, in build
         res = future.result()
       File "/usr/lib/python3.7/concurrent/futures/_base.py", line 435, in result
@@ -1247,7 +1247,7 @@ for this template
     TimeoutError
 
             [('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4691833
-    No: 11  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 11  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1370,7 +1370,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 2, 64]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1042124
-    No: 12  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 12  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1493,7 +1493,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 32, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10013405
-    No: 13  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 13  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1616,7 +1616,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6732082
-    No: 14  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 14  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1739,7 +1739,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7536735
-    No: 15  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 15  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1862,7 +1862,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,482121
-    No: 16  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 16  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1985,7 +1985,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 1, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2824525
-    No: 17  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 17  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2108,7 +2108,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 64, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4559286
-    No: 18  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 18  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2231,7 +2231,7 @@ for this template
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
     tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 32, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 512]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9677544
-    No: 19  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+    No: 19  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 721, in __call__
         yield remote, remote.load_module(os.path.split(build_result.filename)[1])
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 685, in run_through_rpc
@@ -2319,7 +2319,7 @@ for this template
       15: _PyEval_EvalFrameDefault
       14: 0x0000000000537c30
       13: _PyObject_FastCallKeywords
-      12: 0x00007fc4d562cfa2
+      12: 0x00007f0bfab37fa2
       11: _ctypes_callproc
       10: ffi_call
       9: ffi_call_unix64
@@ -2384,7 +2384,7 @@ for this template
       21: _PyFunction_FastCallKeywords
       20: _PyEval_EvalFrameDefault
       19: _PyFunction_FastCall      [('tile_f', [-1, 8, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6390073
-    No: 20  GFLOPS: 144.98/144.98   result: MeasureResult(costs=(0.0015968331,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4312255382537842, timestamp=1650063841.3437583)       [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
+    No: 20  GFLOPS: 144.55/144.55   result: MeasureResult(costs=(0.0016015775600000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4185152053833008, timestamp=1650065158.492134)       [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
 
 
 
@@ -2437,7 +2437,7 @@ and measure running time.
 
     Best config:
     [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
-    Time cost of this operator: 0.001952
+    Time cost of this operator: 0.002011
 
 
 
diff --git a/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt b/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt
index 0e6828c55..ccd06cdb0 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt
@@ -292,10 +292,10 @@ Timing the untuned program
     ########## Build without Autotuning ##########
     Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs  
     ---------                                     ---                                           --------  -------  -----              ------  -------  
-    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  309.7     98.733   (1, 2, 10, 10, 3)  2       1        
-    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.067     0.978    (1, 6, 10, 10)     1       1        
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.906     0.289    (1, 1, 10, 10, 3)  1       1        
-    Total_time                                    -                                             313.673   -        -                  -       -        
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  313.0     98.744   (1, 2, 10, 10, 3)  2       1        
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.073     0.969    (1, 6, 10, 10)     1       1        
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.907     0.286    (1, 1, 10, 10, 3)  1       1        
+    Total_time                                    -                                             316.98    -        -                  -       -        
 
 
 
@@ -357,10 +357,10 @@ Timing the tuned program
     ########## Build with Autotuning ##########
     Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs  
     ---------                                     ---                                           --------  -------  -----              ------  -------  
-    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  89.1      97.148   (1, 6, 10, 10, 1)  2       1        
-    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.715     1.87     (1, 6, 10, 10)     1       1        
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.901     0.982    (1, 1, 10, 10, 3)  1       1        
-    Total_time                                    -                                             91.716    -        -                  -       -        
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  133.7     98.082   (1, 6, 10, 10, 1)  2       1        
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.702     1.248    (1, 6, 10, 10)     1       1        
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.913     0.67     (1, 1, 10, 10, 3)  1       1        
+    Total_time                                    -                                             136.314   -        -                  -       -        
 
 
 
diff --git a/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt b/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt
index 696f424f0..d8594a20a 100644
--- a/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**00:43.326** total execution time for **how_to_work_with_microtvm** files:
+**00:43.330** total execution time for **how_to_work_with_microtvm** files:
 
-- **00:39.410**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)
-- **00:03.367**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)
-- **00:00.188**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)
-- **00:00.186**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``)
+- **00:39.365**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)
+- **00:03.415**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)
+- **00:00.190**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``)
+- **00:00.184**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)
 - **00:00.176**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``)
diff --git a/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt b/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
index 175ebd546..cf2f8d765 100644
--- a/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
@@ -5,8 +5,8 @@
 
 Computation times
 =================
-**00:05.817** total execution time for **how_to_work_with_relay** files:
+**00:09.148** total execution time for **how_to_work_with_relay** files:
 
-- **00:04.046**: :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)
-- **00:01.566**: :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)
-- **00:00.205**: :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)
+- **00:07.078**: :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)
+- **00:01.882**: :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)
+- **00:00.188**: :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)
diff --git a/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt b/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
index a89063969..d7df8dfd7 100644
--- a/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
@@ -5,13 +5,13 @@
 
 Computation times
 =================
-**00:05.195** total execution time for **how_to_work_with_schedules** files:
+**00:05.384** total execution time for **how_to_work_with_schedules** files:
 
-- **00:01.939**: :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)
-- **00:00.986**: :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)
-- **00:00.687**: :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)
-- **00:00.656**: :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)
-- **00:00.289**: :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)
-- **00:00.227**: :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``)
-- **00:00.213**: :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)
-- **00:00.198**: :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)
+- **00:02.025**: :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)
+- **00:01.059**: :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)
+- **00:00.701**: :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)
+- **00:00.690**: :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)
+- **00:00.277**: :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)
+- **00:00.231**: :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``)
+- **00:00.207**: :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)
+- **00:00.194**: :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)
diff --git a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
index a3275fdc0..2f539e4b1 100644
--- a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
@@ -314,7 +314,7 @@ The importing needs to happen before the tensorized GEMV being executed.
                  B: Buffer(B_2: Pointer(float32), float32, [32768], []),
                  C: Buffer(C_2: Pointer(float32), float32, [524288], [])}
       buffer_map = {A_1: A, B_1: B, C_1: C} {
-      attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmphcc2ycl2/input0.cc'\nsource_filename = \"/tmp/tmphcc2ycl2/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n  %7 = alloca float*, align 8\n  %8 = alloca float*, align 8\n  %9 = alloca floa [...]
+      attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmp5asqara1/input0.cc'\nsource_filename = \"/tmp/tmp5asqara1/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n  %7 = alloca float*, align 8\n  %8 = alloca float*, align 8\n  %9 = alloca floa [...]
       for (i, 0, 1024) {
         for (j.outer: int32, 0, 32) {
           @tir.call_extern("gemv_update", @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), C_2, ((i*512) + (j.outer*16)), 16, 2, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), A_2, (i*64), 64, 1, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), B_2, (j.outer*1024), 1024, 1, dtype=handle), 16, 64, 64, dtype=int32)
diff --git a/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
index cbd03957f..fb29ef755 100644
--- a/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
@@ -5,7 +5,7 @@
 
 Computation times
 =================
-**00:20.128** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:20.420** total execution time for **topic_vta_tutorials_autotvm** files:
 
-- **00:19.945**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``)
-- **00:00.184**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)
+- **00:20.224**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``)
+- **00:00.196**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)
diff --git a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
index c5483c007..6dcc4582a 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
@@ -265,7 +265,7 @@ The compilation steps are:
       DeprecationWarning,
     /workspace/vta/tutorials/frontend/deploy_classification.py:213: DeprecationWarning: legacy graph executor behavior of producing json / lib / params will be removed in the next release. Please see documents of tvm.contrib.graph_executor.GraphModule for the  new recommended usage.
       relay_prog, target=tvm.target.Target(target, host=env.target_host), params=params
-    resnet18_v1 inference graph built in 21.18s!
+    resnet18_v1 inference graph built in 21.20s!
 
 
 
diff --git a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
index 2ca6dbcdd..57b07b313 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
@@ -301,7 +301,7 @@ The compilation steps are:
 
     /workspace/python/tvm/relay/build_module.py:439: DeprecationWarning: Please use input parameter mod (tvm.IRModule) instead of deprecated parameter mod (tvm.relay.function.Function)
       DeprecationWarning,
-    yolov3-tiny inference graph built in 14.74s!
+    yolov3-tiny inference graph built in 14.79s!
 
 
 
diff --git a/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt
index cb0aa80e6..961a680b5 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt
@@ -5,7 +5,7 @@
 
 Computation times
 =================
-**01:27.605** total execution time for **topic_vta_tutorials_frontend** files:
+**01:27.783** total execution time for **topic_vta_tutorials_frontend** files:
 
-- **00:46.557**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)
-- **00:41.048**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``)
+- **00:46.643**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)
+- **00:41.140**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``)
diff --git a/docs/_sources/topic/vta/tutorials/optimize/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/optimize/sg_execution_times.rst.txt
index 7ae7e7968..0ac418b84 100644
--- a/docs/_sources/topic/vta/tutorials/optimize/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/optimize/sg_execution_times.rst.txt
@@ -5,7 +5,7 @@
 
 Computation times
 =================
-**00:03.467** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.500** total execution time for **topic_vta_tutorials_optimize** files:
 
-- **00:02.970**: :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)
-- **00:00.497**: :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``)
+- **00:02.969**: :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)
+- **00:00.531**: :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``)
diff --git a/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
index 706574f47..2e18a3e4e 100644
--- a/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
@@ -5,7 +5,7 @@
 
 Computation times
 =================
-**00:00.907** total execution time for **topic_vta_tutorials** files:
+**00:00.927** total execution time for **topic_vta_tutorials** files:
 
-- **00:00.463**: :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``)
-- **00:00.444**: :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``)
+- **00:00.471**: :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``)
+- **00:00.456**: :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``)
diff --git a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
index af00ae3aa..a34c5e975 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -305,7 +305,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 92.791 ms
+    Execution time of this operator: 93.274 ms
 
 
 
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index bf2e9df60..cf84a334d 100644
--- a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
@@ -268,7 +268,7 @@ standard deviation.
 
  .. code-block:: none
 
-    {'mean': 490.400963600041, 'median': 490.2133211000546, 'std': 0.7120891974454429}
+    {'mean': 492.5690862900001, 'median': 492.4631355000031, 'std': 1.3102262374601348}
 
 
 
@@ -482,31 +482,31 @@ the tuning data to.
 
  .. code-block:: none
 
-
    [Task  1/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  1/25]  Current/Best:   17.88/  17.88 GFLOPS | Progress: (4/10) | 5.42 s
    [Task  1/25]  Current/Best:   16.82/  18.32 GFLOPS | Progress: (8/10) | 9.04 s
    [Task  1/25]  Current/Best:   15.25/  18.32 GFLOPS | Progress: (10/10) | 10.29 s Done.
-
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  2/25]  Current/Best:    5.85/  16.32 GFLOPS | Progress: (4/10) | 2.41 s
    [Task  2/25]  Current/Best:   15.50/  16.85 GFLOPS | Progress: (8/10) | 3.77 s
    [Task  2/25]  Current/Best:   17.75/  20.05 GFLOPS | Progress: (10/10) | 4.39 s Done.
-
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  3/25]  Current/Best:    8.11/  14.04 GFLOPS | Progress: (4/10) | 3.37 s
    [Task  3/25]  Current/Best:   12.35/  17.13 GFLOPS | Progress: (8/10) | 4.97 s
    [Task  3/25]  Current/Best:   13.29/  17.13 GFLOPS | Progress: (10/10) | 6.06 s Done.
-
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  4/25]  Current/Best:   18.37/  18.37 GFLOPS | Progress: (4/10) | 2.65 s
    [Task  4/25]  Current/Best:    4.72/  20.28 GFLOPS | Progress: (8/10) | 7.83 s
    [Task  4/25]  Current/Best:   12.64/  20.28 GFLOPS | Progress: (10/10) | 9.09 s Done.
-
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  5/25]  Current/Best:   21.00/  21.46 GFLOPS | Progress: (4/10) | 2.46 s
    [Task  5/25]  Current/Best:    7.65/  21.46 GFLOPS | Progress: (8/10) | 3.98 s
    [Task  5/25]  Current/Best:   16.54/  21.46 GFLOPS | Progress: (10/10) | 4.89 s Done.
-
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  6/25]  Current/Best:   11.72/  18.02 GFLOPS | Progress: (4/10) | 4.66 s
    [Task  6/25]  Current/Best:   14.79/  18.02 GFLOPS | Progress: (8/10) | 7.22 s
    [Task  6/25]  Current/Best:   16.45/  18.02 GFLOPS | Progress: (10/10) | 8.08 s Done.
-
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  7/25]  Current/Best:    6.33/  15.80 GFLOPS | Progress: (4/10) | 3.70 s
    [Task  7/25]  Current/Best:    9.05/  15.80 GFLOPS | Progress: (8/10) | 6.42 s
    [Task  7/25]  Current/Best:   17.16/  17.16 GFLOPS | Progress: (10/10) | 7.20 s Done.
-
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  8/25]  Current/Best:    5.38/  12.06 GFLOPS | Progress: (4/10) | 7.40 s
    [Task  8/25]  Current/Best:    9.89/  17.30 GFLOPS | Progress: (8/10) | 12.55 s
    [Task  8/25]  Current/Best:   10.60/  17.30 GFLOPS | Progress: (10/10) | 15.53 s Done.
-
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  9/25]  Current/Best:   12.41/  14.43 GFLOPS | Progress: (4/10) | 10.41 s
    [Task  9/25]  Current/Best:   17.91/  22.62 GFLOPS | Progress: (8/10) | 11.70 s
    [Task  9/25]  Current/Best:   17.47/  22.62 GFLOPS | Progress: (10/10) | 12.76 s Done.
-
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 10/25]  Current/Best:    3.00/  18.14 GFLOPS | Progress: (4/10) | 3.02 s
    [Task 10/25]  Current/Best:   19.79/  19.79 GFLOPS | Progress: (8/10) | 5.23 s
    [Task 10/25]  Current/Best:    9.55/  19.79 GFLOPS | Progress: (10/10) | 5.87 s Done.
-
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 11/25]  Current/Best:    6.20/  16.52 GFLOPS | Progress: (4/10) | 4.90 s
    [Task 11/25]  Current/Best:   15.74/  17.53 GFLOPS | Progress: (8/10) | 6.84 s
    [Task 11/25]  Current/Best:   18.58/  18.58 GFLOPS | Progress: (10/10) | 7.95 s Done.
-
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 12/25]  Current/Best:   10.68/  14.09 GFLOPS | Progress: (4/10) | 3.16 s
    [Task 12/25]  Current/Best:   15.10/  15.32 GFLOPS | Progress: (8/10) | 4.78 s
    [Task 12/25]  Current/Best:   10.08/  15.32 GFLOPS | Progress: (10/10) | 6.55 s Done.
-
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 13/25]  Current/Best:   18.01/  18.01 GFLOPS | Progress: (4/10) | 5.41 s
    [Task 13/25]  Current/Best:    3.11/  18.01 GFLOPS | Progress: (8/10) | 8.19 s
    [Task 13/25]  Current/Best:   14.66/  20.23 GFLOPS | Progress: (10/10) | 9.10 s Done.
-
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 14/25]  Current/Best:   12.97/  12.97 GFLOPS | Progress: (4/10) | 7.04 s
    [Task 14/25]  Current/Best:   13.36/  15.10 GFLOPS | Progress: (8/10) | 9.23 s
    [Task 14/25]  Current/Best:    9.44/  15.10 GFLOPS | Progress: (10/10) | 10.40 s
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 15/25]  Current/Best:   15.71/  24.03 GFLOPS | Progress: (4/10) | 3.67 s
    [Task 15/25]  Current/Best:    5.99/  24.03 GFLOPS | Progress: (8/10) | 4.86 s
    [Task 15/25]  Current/Best:   10.84/  24.03 GFLOPS | Progress: (10/10) | 6.22 s Done.
-
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 16/25]  Current/Best:    6.30/  15.06 GFLOPS | Progress: (4/10) | 2.83 s
    [Task 16/25]  Current/Best:   14.64/  18.77 GFLOPS | Progress: (8/10) | 4.36 s
    [Task 16/25]  Current/Best:   16.19/  18.77 GFLOPS | Progress: (10/10) | 5.69 s Done.
-
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 17/25]  Current/Best:   18.16/  18.16 GFLOPS | Progress: (4/10) | 3.94 s
    [Task 17/25]  Current/Best:   21.02/  21.02 GFLOPS | Progress: (8/10) | 5.37 s
    [Task 17/25]  Current/Best:   15.53/  23.93 GFLOPS | Progress: (10/10) | 6.77 s Done.
-
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 18/25]  Current/Best:   11.67/  19.96 GFLOPS | Progress: (4/10) | 2.43 s
    [Task 18/25]  Current/Best:   11.79/  20.65 GFLOPS | Progress: (8/10) | 8.02 s
    [Task 18/25]  Current/Best:   14.76/  20.65 GFLOPS | Progress: (10/10) | 9.38 s Done.
-
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 19/25]  Current/Best:   10.68/  10.68 GFLOPS | Progress: (4/10) | 5.60 s
    [Task 19/25]  Current/Best:   16.43/  20.13 GFLOPS | Progress: (8/10) | 8.72 s
    [Task 19/25]  Current/Best:   10.97/  20.13 GFLOPS | Progress: (10/10) | 10.19 s Done.
-
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 20/25]  Current/Best:   10.14/  10.28 GFLOPS | Progress: (4/10) | 4.21 s
    [Task 20/25]  Current/Best:    4.00/  21.25 GFLOPS | Progress: (8/10) | 6.61 s
    [Task 20/25]  Current/Best:   20.10/  21.25 GFLOPS | Progress: (10/10) | 7.63 s Done.
-
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 21/25]  Current/Best:   14.90/  21.97 GFLOPS | Progress: (4/10) | 2.66 s
    [Task 21/25]  Current/Best:    1.63/  21.97 GFLOPS | Progress: (8/10) | 4.66 s
    [Task 21/25]  Current/Best:    9.70/  21.97 GFLOPS | Progress: (10/10) | 5.74 s
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
+
    [Task  1/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  1/25]  Current/Best:   10.28/  10.28 GFLOPS | Progress: (4/10) | 6.02 s
    [Task  1/25]  Current/Best:    6.49/  23.64 GFLOPS | Progress: (8/10) | 8.43 s
    [Task  1/25]  Current/Best:    4.69/  23.73 GFLOPS | Progress: (10/10) | 10.78 s Done.
+
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  2/25]  Current/Best:    8.29/  14.64 GFLOPS | Progress: (4/10) | 2.48 s
    [Task  2/25]  Current/Best:   16.06/  18.47 GFLOPS | Progress: (8/10) | 5.16 s
    [Task  2/25]  Current/Best:   15.88/  18.47 GFLOPS | Progress: (10/10) | 5.78 s Done.
+
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  3/25]  Current/Best:   10.82/  10.82 GFLOPS | Progress: (4/10) | 3.50 s
    [Task  3/25]  Current/Best:   20.96/  23.00 GFLOPS | Progress: (8/10) | 5.06 s
    [Task  3/25]  Current/Best:   16.66/  23.00 GFLOPS | Progress: (10/10) | 5.80 s Done.
+
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  4/25]  Current/Best:    7.86/  13.29 GFLOPS | Progress: (4/10) | 3.30 s
    [Task  4/25]  Current/Best:   13.72/  16.33 GFLOPS | Progress: (8/10) | 4.98 s
    [Task  4/25]  Current/Best:   20.22/  20.22 GFLOPS | Progress: (10/10) | 6.12 s Done.
+
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  5/25]  Current/Best:   12.54/  16.00 GFLOPS | Progress: (4/10) | 2.67 s
    [Task  5/25]  Current/Best:    5.44/  17.62 GFLOPS | Progress: (8/10) | 4.55 s
    [Task  5/25]  Current/Best:   17.35/  17.62 GFLOPS | Progress: (10/10) | 5.43 s Done.
+
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  6/25]  Current/Best:   19.37/  19.37 GFLOPS | Progress: (4/10) | 2.92 s
    [Task  6/25]  Current/Best:   10.08/  19.37 GFLOPS | Progress: (8/10) | 4.92 s
    [Task  6/25]  Current/Best:    9.78/  19.37 GFLOPS | Progress: (10/10) | 6.24 s Done.
+
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  7/25]  Current/Best:   15.97/  16.63 GFLOPS | Progress: (4/10) | 2.73 s
    [Task  7/25]  Current/Best:   17.82/  17.82 GFLOPS | Progress: (8/10) | 4.52 s
    [Task  7/25]  Current/Best:   14.28/  19.29 GFLOPS | Progress: (10/10) | 5.29 s Done.
+
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  8/25]  Current/Best:   12.66/  12.73 GFLOPS | Progress: (4/10) | 3.66 s
    [Task  8/25]  Current/Best:   10.49/  15.92 GFLOPS | Progress: (8/10) | 11.18 s
    [Task  8/25]  Current/Best:   12.64/  15.92 GFLOPS | Progress: (10/10) | 12.10 s Done.
+
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task  9/25]  Current/Best:    6.93/  20.88 GFLOPS | Progress: (4/10) | 5.73 s
    [Task  9/25]  Current/Best:    9.11/  20.88 GFLOPS | Progress: (8/10) | 16.96 s
    [Task  9/25]  Current/Best:   15.31/  20.88 GFLOPS | Progress: (10/10) | 17.60 s
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 10/25]  Current/Best:   12.11/  12.11 GFLOPS | Progress: (4/10) | 2.95 s
    [Task 10/25]  Current/Best:   14.79/  14.87 GFLOPS | Progress: (8/10) | 5.34 s
    [Task 10/25]  Current/Best:   13.11/  14.87 GFLOPS | Progress: (10/10) | 6.37 s Done.
+
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 11/25]  Current/Best:   14.40/  20.34 GFLOPS | Progress: (4/10) | 3.58 s
    [Task 11/25]  Current/Best:   18.13/  20.67 GFLOPS | Progress: (8/10) | 5.20 s
    [Task 11/25]  Current/Best:   20.56/  20.67 GFLOPS | Progress: (10/10) | 6.31 s Done.
+
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 12/25]  Current/Best:   15.05/  18.33 GFLOPS | Progress: (4/10) | 2.65 s
    [Task 12/25]  Current/Best:   10.02/  20.13 GFLOPS | Progress: (8/10) | 5.26 s
    [Task 12/25]  Current/Best:   17.95/  20.13 GFLOPS | Progress: (10/10) | 6.60 s Done.
+
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 13/25]  Current/Best:   10.66/  18.87 GFLOPS | Progress: (4/10) | 3.95 s
    [Task 13/25]  Current/Best:   15.68/  18.87 GFLOPS | Progress: (8/10) | 7.03 s
    [Task 13/25]  Current/Best:   15.29/  18.87 GFLOPS | Progress: (10/10) | 7.98 s Done.
+
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 14/25]  Current/Best:   14.55/  22.67 GFLOPS | Progress: (4/10) | 2.59 s
    [Task 14/25]  Current/Best:   12.84/  22.67 GFLOPS | Progress: (8/10) | 4.76 s
    [Task 14/25]  Current/Best:   11.20/  22.67 GFLOPS | Progress: (10/10) | 9.04 s Done.
+
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 15/25]  Current/Best:    3.13/  19.24 GFLOPS | Progress: (4/10) | 2.81 s
    [Task 15/25]  Current/Best:    7.10/  20.90 GFLOPS | Progress: (8/10) | 4.53 s
    [Task 15/25]  Current/Best:   14.32/  20.90 GFLOPS | Progress: (10/10) | 5.16 s
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
      Done.
-
    [Task 22/25]  Current/Best:    9.84/  16.59 GFLOPS | Progress: (4/10) | 2.95 s
    [Task 22/25]  Current/Best:    7.24/  16.79 GFLOPS | Progress: (8/10) | 4.53 s
    [Task 22/25]  Current/Best:   19.38/  19.38 GFLOPS | Progress: (10/10) | 6.70 s Done.
-
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 23/25]  Current/Best:   18.45/  23.04 GFLOPS | Progress: (4/10) | 3.22 s
    [Task 23/25]  Current/Best:    3.09/  23.04 GFLOPS | Progress: (8/10) | 5.78 s
    [Task 23/25]  Current/Best:    9.02/  23.04 GFLOPS | Progress: (10/10) | 7.11 s Done.
-
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 24/25]  Current/Best:    6.01/   6.16 GFLOPS | Progress: (4/10) | 3.65 s
    [Task 24/25]  Current/Best:    4.02/  10.00 GFLOPS | Progress: (8/10) | 5.83 s
    [Task 24/25]  Current/Best:    7.13/  10.00 GFLOPS | Progress: (10/10) | 7.21 s Done.
-
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 25/25]  Current/Best:    9.67/   9.67 GFLOPS | Progress: (4/10) | 2.34 s
    [Task 25/25]  Current/Best:    8.43/  10.39 GFLOPS | Progress: (8/10) | 34.56 s
    [Task 25/25]  Current/Best:    1.53/  10.39 GFLOPS | Progress: (10/10) | 50.42 s
+
    [Task 16/25]  Current/Best:    9.49/  15.85 GFLOPS | Progress: (4/10) | 2.90 s
    [Task 16/25]  Current/Best:   13.13/  20.91 GFLOPS | Progress: (8/10) | 4.53 s
    [Task 16/25]  Current/Best:    6.03/  20.91 GFLOPS | Progress: (10/10) | 5.70 s Done.
+
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 17/25]  Current/Best:    3.10/  11.93 GFLOPS | Progress: (4/10) | 4.81 s
    [Task 17/25]  Current/Best:    3.11/  24.29 GFLOPS | Progress: (8/10) | 7.42 s
    [Task 17/25]  Current/Best:   12.75/  24.29 GFLOPS | Progress: (10/10) | 8.59 s Done.
+
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 18/25]  Current/Best:   12.43/  16.49 GFLOPS | Progress: (4/10) | 3.55 s
    [Task 18/25]  Current/Best:   12.53/  18.23 GFLOPS | Progress: (8/10) | 6.10 s
    [Task 18/25]  Current/Best:   11.47/  18.23 GFLOPS | Progress: (10/10) | 6.86 s Done.
+
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 19/25]  Current/Best:   14.29/  14.29 GFLOPS | Progress: (4/10) | 5.71 s
    [Task 19/25]  Current/Best:   15.07/  18.40 GFLOPS | Progress: (8/10) | 7.84 s
    [Task 19/25]  Current/Best:    6.19/  23.47 GFLOPS | Progress: (10/10) | 9.90 s Done.
+
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 20/25]  Current/Best:   21.85/  21.85 GFLOPS | Progress: (4/10) | 3.56 s
    [Task 20/25]  Current/Best:    6.04/  21.85 GFLOPS | Progress: (8/10) | 7.38 s
    [Task 20/25]  Current/Best:    2.71/  21.85 GFLOPS | Progress: (10/10) | 9.16 s Done.
+
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 21/25]  Current/Best:   13.66/  16.36 GFLOPS | Progress: (4/10) | 2.66 s
    [Task 21/25]  Current/Best:   23.76/  23.76 GFLOPS | Progress: (8/10) | 3.94 s
    [Task 21/25]  Current/Best:    1.63/  23.76 GFLOPS | Progress: (10/10) | 5.16 s
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 22/25]  Current/Best:   18.89/  18.89 GFLOPS | Progress: (4/10) | 2.68 s
    [Task 22/25]  Current/Best:    6.02/  18.89 GFLOPS | Progress: (8/10) | 5.57 s
    [Task 22/25]  Current/Best:   18.12/  18.89 GFLOPS | Progress: (10/10) | 6.25 s Done.
+
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 23/25]  Current/Best:    5.39/  19.67 GFLOPS | Progress: (4/10) | 10.13 s
    [Task 23/25]  Current/Best:   11.44/  20.19 GFLOPS | Progress: (8/10) | 12.75 s
    [Task 23/25]  Current/Best:   20.62/  20.62 GFLOPS | Progress: (10/10) | 13.57 s Done.
+
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s
    [Task 24/25]  Current/Best:    4.41/   7.86 GFLOPS | Progress: (4/10) | 3.22 s
    [Task 24/25]  Current/Best:    4.36/   8.15 GFLOPS | Progress: (8/10) | 15.45 s
    [Task 24/25]  Current/Best:    4.05/   8.15 GFLOPS | Progress: (10/10) | 30.63 s
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
+     Done.
+
    [Task 25/25]  Current/Best:    1.55/   1.55 GFLOPS | Progress: (4/10) | 34.45 s
    [Task 25/25]  Current/Best:    6.07/   8.23 GFLOPS | Progress: (8/10) | 51.29 s
    [Task 25/25]  Current/Best:    0.00/   8.23 GFLOPS | Progress: (10/10) | 62.50 s
 
 
 The output from this tuning process will look something like this:
@@ -656,8 +656,8 @@ improvement in comparing the optimized model to the unoptimized model.
 
  .. code-block:: none
 
-    optimized: {'mean': 443.517296980026, 'median': 443.01024564992986, 'std': 1.563976552930658}
-    unoptimized: {'mean': 490.400963600041, 'median': 490.2133211000546, 'std': 0.7120891974454429}
+    optimized: {'mean': 431.6472741200005, 'median': 431.64223475000085, 'std': 0.9637452451115474}
+    unoptimized: {'mean': 492.5690862900001, 'median': 492.4631355000031, 'std': 1.3102262374601348}
 
 
 
@@ -677,7 +677,7 @@ profiling/benchmarking.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 7 minutes  15.308 seconds)
+   **Total running time of the script:** ( 7 minutes  48.916 seconds)
 
 
 .. _sphx_glr_download_tutorial_autotvm_relay_x86.py:
diff --git a/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt b/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
index a01177d45..4b52d23c5 100644
--- a/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
+++ b/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
@@ -235,7 +235,7 @@ device and returns the measured cost. Network overhead is excluded.
 
  .. code-block:: none
 
-    1.233e-07 secs/op
+    1.344e-07 secs/op
 
 
 
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index b46ea6db0..8b799fb12 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -230,7 +230,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
 
  .. code-block:: none
 
-    [stage(a, placeholder(a, 0x2342b560)), stage(b, placeholder(b, 0x28d18ad0)), stage(T_add, compute(T_add, body=[(a[ax0, ax1, ax2] + b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(min=0, ext=10))], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[(a[ax0, ax1, ax2]*b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(mi [...]
+    [stage(a, placeholder(a, 0x107bad80)), stage(b, placeholder(b, 0x229654d0)), stage(T_add, compute(T_add, body=[(a[ax0, ax1, ax2] + b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(min=0, ext=10))], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[(a[ax0, ax1, ax2]*b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(mi [...]
 
 
 
diff --git a/docs/_sources/tutorial/sg_execution_times.rst.txt b/docs/_sources/tutorial/sg_execution_times.rst.txt
index 2113ed072..5b8235d10 100644
--- a/docs/_sources/tutorial/sg_execution_times.rst.txt
+++ b/docs/_sources/tutorial/sg_execution_times.rst.txt
@@ -5,17 +5,17 @@
 
 Computation times
 =================
-**09:47.352** total execution time for **tutorial** files:
+**10:34.152** total execution time for **tutorial** files:
 
-- **07:15.308**: :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)
-- **00:58.544**: :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)
-- **00:49.313**: :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``)
-- **00:25.816**: :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)
-- **00:16.820**: :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)
-- **00:00.700**: :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)
-- **00:00.539**: :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)
-- **00:00.190**: :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``)
-- **00:00.033**: :ref:`sphx_glr_tutorial_install.py` (``install.py``)
-- **00:00.032**: :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)
-- **00:00.029**: :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)
-- **00:00.028**: :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)
+- **07:48.916**: :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)
+- **01:00.960**: :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)
+- **00:49.020**: :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``)
+- **00:26.868**: :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)
+- **00:26.069**: :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)
+- **00:01.264**: :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)
+- **00:00.701**: :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)
+- **00:00.216**: :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``)
+- **00:00.042**: :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)
+- **00:00.035**: :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)
+- **00:00.032**: :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)
+- **00:00.030**: :ref:`sphx_glr_tutorial_install.py` (``install.py``)
diff --git a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
index c402939c3..d7770a2fe 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -243,8 +243,8 @@ helper function to run a profile of the TVM generated code.
 
  .. code-block:: none
 
-    Numpy running time: 0.000008
-    naive: 0.000006
+    Numpy running time: 0.000009
+    naive: 0.000007
 
 
 
@@ -334,7 +334,7 @@ compile and run this new schedule with the parallel operation applied:
 
  .. code-block:: none
 
-    parallel: 0.000006
+    parallel: 0.000007
 
 
 
@@ -436,10 +436,10 @@ We can now compare the different schedules
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                   numpy    8.072379969235045e-06                    1.0
-                   naive              5.8386e-06      0.7232811168765236
-                parallel    6.2619999999999995e-06    0.7757315715892149
-                  vector    2.4645900000000004e-05    3.0531144586762435
+                   numpy    8.90588999936881e-06                     1.0
+                   naive              6.8638e-06      0.7707034334004194
+                parallel              7.0911e-06      0.7962258685546947
+                  vector             2.45479e-05      2.7563668540415156
 
 
 
@@ -828,7 +828,7 @@ matrix multiplication.
 
  .. code-block:: none
 
-    Numpy running time: 0.018119
+    Numpy running time: 0.018910
 
 
 
@@ -884,7 +884,7 @@ optimizations.
 
  .. code-block:: none
 
-    none: 3.231856
+    none: 3.389160
 
 
 
@@ -982,7 +982,7 @@ schedule.
 
  .. code-block:: none
 
-    blocking: 0.307331
+    blocking: 0.315400
 
 
 
@@ -1073,7 +1073,7 @@ already cache friendly from our previous optimizations.
 
  .. code-block:: none
 
-    vectorization: 0.337173
+    vectorization: 0.346792
     @main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
       attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
       buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1144,7 +1144,7 @@ more cache friendly.
 
  .. code-block:: none
 
-    loop permutation: 0.112953
+    loop permutation: 0.115190
     @main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
       attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
       buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1240,7 +1240,7 @@ optimized schedule.
 
  .. code-block:: none
 
-    array packing: 0.108143
+    array packing: 0.110624
     @main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
       attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
       buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1330,7 +1330,7 @@ to `C` when all the block results are ready.
 
  .. code-block:: none
 
-    block caching: 0.111776
+    block caching: 0.110961
     @main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
       attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
       buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1413,7 +1413,7 @@ of thread-level parallelization.
 
  .. code-block:: none
 
-    parallelization: 0.143978
+    parallelization: 0.144058
     @main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
       attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
       buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1491,13 +1491,13 @@ working, we can compare the results.
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                    none            3.2318558358                     1.0
-                blocking            0.3073305725      0.0950941465567955
-           vectorization     0.33717297030000004     0.10432797359494152
-        loop permutation             0.112953436      0.0349500230637732
-           array packing            0.1081428521    0.033461533432920215
-           block caching     0.11177630140000001     0.03458579437913925
-         parallelization            0.1439777031     0.04454954379620722
+                    none            3.3891596221                     1.0
+                blocking            0.3153996863      0.0930613253631799
+           vectorization     0.34679244389999997     0.10232402204919445
+        loop permutation     0.11519021680000001     0.03398784053984024
+           array packing     0.11062351150000001     0.03264039580155719
+           block caching     0.11096075920000001     0.03273990356678633
+         parallelization            0.1440581698    0.042505572431769466
 
 
 
@@ -1532,6 +1532,11 @@ operations with tunable parameters that allows you to automatically optimize
 the computation for specific platforms.
 
 
+.. rst-class:: sphx-glr-timing
+
+   **Total running time of the script:** ( 1 minutes  0.960 seconds)
+
+
 .. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
 
 
diff --git a/docs/commit_hash b/docs/commit_hash
index 4e0cd7761..6f0b64a70 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-351f31b51cd85648b66f2b344b96a7460052760b
+fafabc96c1ba1a5f987c2402fcc2ce4d1bad5cc8
diff --git a/docs/how_to/compile_models/from_mxnet.html b/docs/how_to/compile_models/from_mxnet.html
index 9bc23ec4a..c4a2be8ed 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -400,7 +400,7 @@
 </div>
 <img alt="../../_images/sphx_glr_from_mxnet_001.png" class="sphx-glr-single-img" src="../../_images/sphx_glr_from_mxnet_001.png" />
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip67db9577-e96a-4c5c-8273-04062e07c5aa from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip6d3e2761-0eae-4f2a-aa04-e3b1d5eca3da from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
 x (1, 3, 224, 224)
 </pre></div>
 </div>
diff --git a/docs/how_to/compile_models/from_paddle.html b/docs/how_to/compile_models/from_paddle.html
index eaa92c64f..f335c8a34 100644
--- a/docs/how_to/compile_models/from_paddle.html
+++ b/docs/how_to/compile_models/from_paddle.html
@@ -463,7 +463,7 @@ A quick solution is</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>TVM prediction top-1 id: 282, class name:  282: &#39;tiger cat&#39;,
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  4.500 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  3.900 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-paddle-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/16269b77359771348d507395692524cf/from_paddle.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">from_paddle.py</span></code></a></p>
diff --git a/docs/how_to/compile_models/from_pytorch.html b/docs/how_to/compile_models/from_pytorch.html
index 24722fb34..958308e02 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -386,10 +386,10 @@ be unstable.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading: &quot;https://download.pytorch.org/models/resnet18-f37072fd.pth&quot; to /workspace/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
 
   0%|          | 0.00/44.7M [00:00&lt;?, ?B/s]
-  1%|1         | 472k/44.7M [00:00&lt;00:09, 4.80MB/s]
- 10%|#         | 4.68M/44.7M [00:00&lt;00:01, 27.9MB/s]
- 70%|#######   | 31.4M/44.7M [00:00&lt;00:00, 143MB/s]
-100%|##########| 44.7M/44.7M [00:00&lt;00:00, 135MB/s]
+  7%|6         | 3.05M/44.7M [00:00&lt;00:01, 32.0MB/s]
+ 17%|#7        | 7.61M/44.7M [00:00&lt;00:00, 41.0MB/s]
+ 69%|######8   | 30.7M/44.7M [00:00&lt;00:00, 133MB/s]
+100%|##########| 44.7M/44.7M [00:00&lt;00:00, 131MB/s]
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/compile_models/sg_execution_times.html b/docs/how_to/compile_models/sg_execution_times.html
index a1761f0a1..6d83df9b5 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -300,17 +300,17 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-compile-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>04:39.864</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>04:38.468</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
 <ul class="simple">
-<li><p><strong>01:04.500</strong>: <a class="reference internal" href="from_paddle.html#sphx-glr-how-to-compile-models-from-paddle-py"><span class="std std-ref">Compile PaddlePaddle Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_paddle.py</span></code>)</p></li>
-<li><p><strong>00:59.553</strong>: <a class="reference internal" href="from_tensorflow.html#sphx-glr-how-to-compile-models-from-tensorflow-py"><span class="std std-ref">Compile Tensorflow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tensorflow.py</span></code>)</p></li>
-<li><p><strong>00:55.423</strong>: <a class="reference internal" href="from_darknet.html#sphx-glr-how-to-compile-models-from-darknet-py"><span class="std std-ref">Compile YOLO-V2 and YOLO-V3 in DarkNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_darknet.py</span></code>)</p></li>
-<li><p><strong>00:25.357</strong>: <a class="reference internal" href="from_tflite.html#sphx-glr-how-to-compile-models-from-tflite-py"><span class="std std-ref">Compile TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tflite.py</span></code>)</p></li>
-<li><p><strong>00:20.788</strong>: <a class="reference internal" href="from_mxnet.html#sphx-glr-how-to-compile-models-from-mxnet-py"><span class="std std-ref">Compile MXNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_mxnet.py</span></code>)</p></li>
-<li><p><strong>00:20.742</strong>: <a class="reference internal" href="from_coreml.html#sphx-glr-how-to-compile-models-from-coreml-py"><span class="std std-ref">Compile CoreML Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_coreml.py</span></code>)</p></li>
-<li><p><strong>00:18.969</strong>: <a class="reference internal" href="from_pytorch.html#sphx-glr-how-to-compile-models-from-pytorch-py"><span class="std std-ref">Compile PyTorch Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_pytorch.py</span></code>)</p></li>
-<li><p><strong>00:12.078</strong>: <a class="reference internal" href="from_keras.html#sphx-glr-how-to-compile-models-from-keras-py"><span class="std std-ref">Compile Keras Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_keras.py</span></code>)</p></li>
-<li><p><strong>00:02.454</strong>: <a class="reference internal" href="from_onnx.html#sphx-glr-how-to-compile-models-from-onnx-py"><span class="std std-ref">Compile ONNX Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_onnx.py</span></code>)</p></li>
+<li><p><strong>01:03.900</strong>: <a class="reference internal" href="from_paddle.html#sphx-glr-how-to-compile-models-from-paddle-py"><span class="std std-ref">Compile PaddlePaddle Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_paddle.py</span></code>)</p></li>
+<li><p><strong>00:59.085</strong>: <a class="reference internal" href="from_tensorflow.html#sphx-glr-how-to-compile-models-from-tensorflow-py"><span class="std std-ref">Compile Tensorflow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tensorflow.py</span></code>)</p></li>
+<li><p><strong>00:55.277</strong>: <a class="reference internal" href="from_darknet.html#sphx-glr-how-to-compile-models-from-darknet-py"><span class="std std-ref">Compile YOLO-V2 and YOLO-V3 in DarkNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_darknet.py</span></code>)</p></li>
+<li><p><strong>00:25.181</strong>: <a class="reference internal" href="from_tflite.html#sphx-glr-how-to-compile-models-from-tflite-py"><span class="std std-ref">Compile TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tflite.py</span></code>)</p></li>
+<li><p><strong>00:20.886</strong>: <a class="reference internal" href="from_coreml.html#sphx-glr-how-to-compile-models-from-coreml-py"><span class="std std-ref">Compile CoreML Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_coreml.py</span></code>)</p></li>
+<li><p><strong>00:20.807</strong>: <a class="reference internal" href="from_mxnet.html#sphx-glr-how-to-compile-models-from-mxnet-py"><span class="std std-ref">Compile MXNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_mxnet.py</span></code>)</p></li>
+<li><p><strong>00:18.207</strong>: <a class="reference internal" href="from_pytorch.html#sphx-glr-how-to-compile-models-from-pytorch-py"><span class="std std-ref">Compile PyTorch Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_pytorch.py</span></code>)</p></li>
+<li><p><strong>00:12.435</strong>: <a class="reference internal" href="from_keras.html#sphx-glr-how-to-compile-models-from-keras-py"><span class="std std-ref">Compile Keras Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_keras.py</span></code>)</p></li>
+<li><p><strong>00:02.691</strong>: <a class="reference internal" href="from_onnx.html#sphx-glr-how-to-compile-models-from-onnx-py"><span class="std std-ref">Compile ONNX Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_onnx.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/deploy_models/deploy_model_on_android.html b/docs/how_to/deploy_models/deploy_model_on_android.html
index 02348dda1..63da494c4 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -622,7 +622,7 @@ to the remote android device.</p>
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  15.6234      15.6349      15.7135      15.5296       0.0604
+  15.9951      15.9650      16.4304      15.8805       0.1509
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
index 01b20dbe3..4d6adf963 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -409,14 +409,15 @@ be unstable.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading: &quot;https://download.pytorch.org/models/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth&quot; to /workspace/.cache/torch/hub/checkpoints/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth
 
   0%|          | 0.00/170M [00:00&lt;?, ?B/s]
- 10%|#         | 17.8M/170M [00:00&lt;00:00, 186MB/s]
- 25%|##4       | 41.7M/170M [00:00&lt;00:00, 224MB/s]
- 39%|###8      | 65.6M/170M [00:00&lt;00:00, 236MB/s]
- 52%|#####2    | 88.3M/170M [00:00&lt;00:00, 236MB/s]
- 65%|######5   | 111M/170M [00:00&lt;00:00, 166MB/s]
- 78%|#######8  | 133M/170M [00:00&lt;00:00, 183MB/s]
- 92%|#########2| 157M/170M [00:00&lt;00:00, 202MB/s]
-100%|##########| 170M/170M [00:00&lt;00:00, 204MB/s]
+  3%|2         | 4.52M/170M [00:00&lt;00:03, 47.3MB/s]
+  6%|5         | 9.66M/170M [00:00&lt;00:03, 51.2MB/s]
+ 20%|##        | 34.1M/170M [00:00&lt;00:00, 145MB/s]
+ 36%|###5      | 60.7M/170M [00:00&lt;00:00, 198MB/s]
+ 51%|#####1    | 87.4M/170M [00:00&lt;00:00, 227MB/s]
+ 67%|######6   | 114M/170M [00:00&lt;00:00, 243MB/s]
+ 82%|########2 | 139M/170M [00:00&lt;00:00, 252MB/s]
+ 98%|#########7| 166M/170M [00:00&lt;00:00, 261MB/s]
+100%|##########| 170M/170M [00:00&lt;00:00, 217MB/s]
 /usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:3878: UserWarning: To copy construct from a tensor, it is recommended to use sourceTensor.clone().detach() or sourceTensor.clone().detach().requires_grad_(True), rather than torch.tensor(sourceTensor).
   for i in range(dim)
 /usr/local/lib/python3.7/dist-packages/torchvision/models/detection/anchor_utils.py:127: UserWarning: __floordiv__ is deprecated, and its behavior will change in a future version of pytorch. It currently rounds toward 0 (like the &#39;trunc&#39; function NOT &#39;floor&#39;). This results in incorrect rounding for negative values. To keep the current behavior, use torch.div(a, b, rounding_mode=&#39;trunc&#39;), or for actual floor division, use torch.div(a, b, rounding_mode=&#39;floor&#39;).
@@ -509,7 +510,7 @@ torchvision rcnn models.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Get 9 valid boxes
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  58.290 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes  0.840 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-object-detection-pytorch-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/7795da4b258c8feff986668b95ef57ad/deploy_object_detection_pytorch.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_object_detection_pytorch.py</span></code></a></p>
diff --git a/docs/how_to/deploy_models/deploy_prequantized.html b/docs/how_to/deploy_models/deploy_prequantized.html
index 534a3f3e5..7cb06b79e 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -450,7 +450,7 @@ training. Other models require a full post training calibration.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading: &quot;https://download.pytorch.org/models/mobilenet_v2-b0353104.pth&quot; to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2-b0353104.pth
 
   0%|          | 0.00/13.6M [00:00&lt;?, ?B/s]
-100%|##########| 13.6M/13.6M [00:00&lt;00:00, 149MB/s]
+100%|##########| 13.6M/13.6M [00:00&lt;00:00, 185MB/s]
 </pre></div>
 </div>
 </div>
@@ -539,7 +539,7 @@ output values are identical out of 1000 outputs from mobilenet v2.</p>
 <p class="sphx-glr-script-out">Out:</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  90.1806      90.1495      91.6179      89.9756       0.1878
+  90.1193      90.0943      91.0812      89.8953       0.1627
 </pre></div>
 </div>
 <div class="admonition note">
@@ -578,7 +578,7 @@ This includes support for the VNNI 8 bit dot product instruction (CascadeLake or
 <div class="section" id="deploy-a-quantized-tflite-model">
 <h2>Deploy a quantized TFLite Model<a class="headerlink" href="#deploy-a-quantized-tflite-model" title="Permalink to this headline">¶</a></h2>
 <p>TODO</p>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  3.795 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  4.586 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/fb8217c13f4351224c6cf3aacf1a87fc/deploy_prequantized.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_prequantized.py</span></code></a></p>
diff --git a/docs/how_to/deploy_models/deploy_prequantized_tflite.html b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
index f4da12a4b..33ad6bafb 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -540,7 +540,7 @@ TFLite Top-5 labels: [387 102 386 341 349]
 <p class="sphx-glr-script-out">Out:</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  118.9514     118.9186     121.8738     118.0731      0.4844
+  119.4266     119.4160     120.6976     118.3389      0.4110
 </pre></div>
 </div>
 <div class="admonition note">
@@ -568,7 +568,7 @@ network for ARM CPU</span></a>.</p></li>
 </ul>
 </div></blockquote>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  57.873 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  54.100 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-tflite-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/56691c7a27d45da61d112276334640d3/deploy_prequantized_tflite.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_prequantized_tflite.py</span></code></a></p>
diff --git a/docs/how_to/deploy_models/deploy_quantized.html b/docs/how_to/deploy_models/deploy_quantized.html
index a7b863220..f0c135cdf 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -480,7 +480,7 @@ for calibration. But the accuracy might be impacted.</p>
   DeprecationWarning,
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  9.391 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  9.426 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-quantized-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/7810ecf51bfc05f7d5e8a400ac3e815d/deploy_quantized.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_quantized.py</span></code></a></p>
diff --git a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
index 00958a576..b4682eeea 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -415,25 +415,26 @@ to your device.</p>
 Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
 
   0%|          | 0/132723 [00:00&lt;?, ?KB/s]
-  2%|1         | 2122/132723 [00:00&lt;00:06, 21216.88KB/s]
-  5%|5         | 7018/132723 [00:00&lt;00:03, 37533.45KB/s]
- 11%|#1        | 14657/132723 [00:00&lt;00:02, 55273.91KB/s]
- 17%|#6        | 22292/132723 [00:00&lt;00:01, 63591.79KB/s]
- 23%|##2       | 29989/132723 [00:00&lt;00:01, 68413.06KB/s]
- 28%|##7       | 36831/132723 [00:00&lt;00:01, 63368.68KB/s]
- 34%|###3      | 44568/132723 [00:00&lt;00:01, 67707.28KB/s]
- 39%|###9      | 52285/132723 [00:00&lt;00:01, 70612.24KB/s]
- 45%|####5     | 60150/132723 [00:00&lt;00:00, 73061.44KB/s]
- 51%|#####1    | 67967/132723 [00:01&lt;00:00, 74605.57KB/s]
- 57%|#####7    | 75785/132723 [00:01&lt;00:00, 75683.36KB/s]
- 63%|######2   | 83613/132723 [00:01&lt;00:00, 76462.65KB/s]
- 69%|######8   | 91424/132723 [00:01&lt;00:00, 76956.22KB/s]
- 75%|#######4  | 99307/132723 [00:01&lt;00:00, 77518.30KB/s]
- 81%|########  | 107068/132723 [00:01&lt;00:00, 76879.52KB/s]
- 86%|########6 | 114764/132723 [00:01&lt;00:00, 65402.06KB/s]
- 92%|#########2| 122373/132723 [00:01&lt;00:00, 68244.07KB/s]
- 98%|#########8| 130274/132723 [00:01&lt;00:00, 71208.53KB/s]
-100%|##########| 132723/132723 [00:01&lt;00:00, 69214.61KB/s]
+  3%|3         | 4011/132723 [00:00&lt;00:03, 40104.97KB/s]
+  9%|8         | 11818/132723 [00:00&lt;00:01, 62429.31KB/s]
+ 15%|#4        | 19487/132723 [00:00&lt;00:01, 68937.87KB/s]
+ 20%|#9        | 26381/132723 [00:00&lt;00:01, 64125.14KB/s]
+ 25%|##4       | 32841/132723 [00:00&lt;00:01, 51078.33KB/s]
+ 30%|##9       | 39514/132723 [00:00&lt;00:01, 55473.10KB/s]
+ 36%|###5      | 47266/132723 [00:00&lt;00:01, 61784.39KB/s]
+ 41%|####      | 54021/132723 [00:00&lt;00:01, 63440.57KB/s]
+ 47%|####6     | 62075/132723 [00:00&lt;00:01, 68450.30KB/s]
+ 52%|#####2    | 69109/132723 [00:01&lt;00:00, 67496.95KB/s]
+ 57%|#####7    | 75990/132723 [00:01&lt;00:01, 53175.22KB/s]
+ 62%|######1   | 81847/132723 [00:01&lt;00:00, 51028.31KB/s]
+ 66%|######5   | 87318/132723 [00:01&lt;00:00, 46698.56KB/s]
+ 72%|#######1  | 95496/132723 [00:01&lt;00:00, 55166.11KB/s]
+ 76%|#######6  | 101424/132723 [00:01&lt;00:00, 55132.32KB/s]
+ 81%|########  | 107224/132723 [00:01&lt;00:00, 51928.00KB/s]
+ 86%|########6 | 114680/132723 [00:02&lt;00:00, 50506.20KB/s]
+ 90%|######### | 119903/132723 [00:02&lt;00:00, 46107.30KB/s]
+ 96%|#########6| 128043/132723 [00:02&lt;00:00, 54593.47KB/s]
+100%|##########| 132723/132723 [00:02&lt;00:00, 56199.73KB/s]
 </pre></div>
 </div>
 <p>Create TVM runtime and do inference
@@ -473,7 +474,7 @@ Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from h
 </pre></div>
 </div>
 <img alt="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" class="sphx-glr-single-img" src="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" />
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  21.103 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  21.536 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-ssd-gluoncv-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/cccb17d28e5e8b2e94ea8cd5ec59f6ed/deploy_ssd_gluoncv.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">deploy_ssd_gluoncv.py</span></code></a></p>
diff --git a/docs/how_to/deploy_models/sg_execution_times.html b/docs/how_to/deploy_models/sg_execution_times.html
index db368dbab..2d28d8d69 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -300,16 +300,16 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-deploy-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>10:19.589</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>10:19.507</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
 <ul class="simple">
-<li><p><strong>02:58.290</strong>: <a class="reference internal" href="deploy_object_detection_pytorch.html#sphx-glr-how-to-deploy-models-deploy-object-detection-pytorch-py"><span class="std std-ref">Compile PyTorch Object Detection Models</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_object_detection_pytorch.py</span></code>)</p></li>
-<li><p><strong>02:21.103</strong>: <a class="reference internal" href="deploy_ssd_gluoncv.html#sphx-glr-how-to-deploy-models-deploy-ssd-gluoncv-py"><span class="std std-ref">Deploy Single Shot Multibox Detector(SSD) model</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_ssd_gluoncv.py</span></code>)</p></li>
-<li><p><strong>01:57.873</strong>: <a class="reference internal" href="deploy_prequantized_tflite.html#sphx-glr-how-to-deploy-models-deploy-prequantized-tflite-py"><span class="std std-ref">Deploy a Framework-prequantized Model with TVM - Part 3 (TFLite)</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_prequantized_tflite.py</span></code>)</p></li>
-<li><p><strong>01:09.391</strong>: <a class="reference internal" href="deploy_quantized.html#sphx-glr-how-to-deploy-models-deploy-quantized-py"><span class="std std-ref">Deploy a Quantized Model on Cuda</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_quantized.py</span></code>)</p></li>
-<li><p><strong>01:03.795</strong>: <a class="reference internal" href="deploy_prequantized.html#sphx-glr-how-to-deploy-models-deploy-prequantized-py"><span class="std std-ref">Deploy a Framework-prequantized Model with TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_prequantized.py</span></code>)</p></li>
-<li><p><strong>00:27.182</strong>: <a class="reference internal" href="deploy_model_on_android.html#sphx-glr-how-to-deploy-models-deploy-model-on-android-py"><span class="std std-ref">Deploy the Pretrained Model on Android</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_android.py</span></code>)</p></li>
-<li><p><strong>00:21.776</strong>: <a class="reference internal" href="deploy_model_on_rasp.html#sphx-glr-how-to-deploy-models-deploy-model-on-rasp-py"><span class="std std-ref">Deploy the Pretrained Model on Raspberry Pi</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_rasp.py</span></code>)</p></li>
-<li><p><strong>00:00.179</strong>: <a class="reference internal" href="deploy_sparse.html#sphx-glr-how-to-deploy-models-deploy-sparse-py"><span class="std std-ref">Deploy a Hugging Face Pruned Model on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_sparse.py</span></code>)</p></li>
+<li><p><strong>03:00.840</strong>: <a class="reference internal" href="deploy_object_detection_pytorch.html#sphx-glr-how-to-deploy-models-deploy-object-detection-pytorch-py"><span class="std std-ref">Compile PyTorch Object Detection Models</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_object_detection_pytorch.py</span></code>)</p></li>
+<li><p><strong>02:21.536</strong>: <a class="reference internal" href="deploy_ssd_gluoncv.html#sphx-glr-how-to-deploy-models-deploy-ssd-gluoncv-py"><span class="std std-ref">Deploy Single Shot Multibox Detector(SSD) model</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_ssd_gluoncv.py</span></code>)</p></li>
+<li><p><strong>01:54.100</strong>: <a class="reference internal" href="deploy_prequantized_tflite.html#sphx-glr-how-to-deploy-models-deploy-prequantized-tflite-py"><span class="std std-ref">Deploy a Framework-prequantized Model with TVM - Part 3 (TFLite)</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_prequantized_tflite.py</span></code>)</p></li>
+<li><p><strong>01:09.426</strong>: <a class="reference internal" href="deploy_quantized.html#sphx-glr-how-to-deploy-models-deploy-quantized-py"><span class="std std-ref">Deploy a Quantized Model on Cuda</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_quantized.py</span></code>)</p></li>
+<li><p><strong>01:04.586</strong>: <a class="reference internal" href="deploy_prequantized.html#sphx-glr-how-to-deploy-models-deploy-prequantized-py"><span class="std std-ref">Deploy a Framework-prequantized Model with TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_prequantized.py</span></code>)</p></li>
+<li><p><strong>00:27.722</strong>: <a class="reference internal" href="deploy_model_on_android.html#sphx-glr-how-to-deploy-models-deploy-model-on-android-py"><span class="std std-ref">Deploy the Pretrained Model on Android</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_android.py</span></code>)</p></li>
+<li><p><strong>00:21.110</strong>: <a class="reference internal" href="deploy_model_on_rasp.html#sphx-glr-how-to-deploy-models-deploy-model-on-rasp-py"><span class="std std-ref">Deploy the Pretrained Model on Raspberry Pi</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_rasp.py</span></code>)</p></li>
+<li><p><strong>00:00.188</strong>: <a class="reference internal" href="deploy_sparse.html#sphx-glr-how-to-deploy-models-deploy-sparse-py"><span class="std std-ref">Deploy a Hugging Face Pruned Model on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_sparse.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/extend_tvm/bring_your_own_datatypes.html b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
index 8c71c419b..ab35f1c7b 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -588,7 +588,7 @@ In this alpha state of the Bring Your Own Datatypes framework, we have not imple
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip87c00332-7f47-42da-8a75-970612d26974 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip1c0edb4c-2721-4485-a82f-207087b6065e from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
 </pre></div>
 </div>
 <p>It’s easy to execute MobileNet with native TVM:</p>
diff --git a/docs/how_to/extend_tvm/sg_execution_times.html b/docs/how_to/extend_tvm/sg_execution_times.html
index f51cb4fd5..539da0f14 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -300,12 +300,12 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-extend-tvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:37.555</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:37.772</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:34.169</strong>: <a class="reference internal" href="bring_your_own_datatypes.html#sphx-glr-how-to-extend-tvm-bring-your-own-datatypes-py"><span class="std std-ref">Bring Your Own Datatypes to TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">bring_your_own_datatypes.py</span></code>)</p></li>
-<li><p><strong>00:02.187</strong>: <a class="reference internal" href="use_pass_instrument.html#sphx-glr-how-to-extend-tvm-use-pass-instrument-py"><span class="std std-ref">How to Use TVM Pass Instrument</span></a> (<code class="docutils literal notranslate"><span class="pre">use_pass_instrument.py</span></code>)</p></li>
-<li><p><strong>00:01.007</strong>: <a class="reference internal" href="use_pass_infra.html#sphx-glr-how-to-extend-tvm-use-pass-infra-py"><span class="std std-ref">How to Use TVM Pass Infra</span></a> (<code class="docutils literal notranslate"><span class="pre">use_pass_infra.py</span></code>)</p></li>
-<li><p><strong>00:00.192</strong>: <a class="reference internal" href="low_level_custom_pass.html#sphx-glr-how-to-extend-tvm-low-level-custom-pass-py"><span class="std std-ref">Writing a Customized Pass</span></a> (<code class="docutils literal notranslate"><span class="pre">low_level_custom_pass.py</span></code>)</p></li>
+<li><p><strong>00:34.314</strong>: <a class="reference internal" href="bring_your_own_datatypes.html#sphx-glr-how-to-extend-tvm-bring-your-own-datatypes-py"><span class="std std-ref">Bring Your Own Datatypes to TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">bring_your_own_datatypes.py</span></code>)</p></li>
+<li><p><strong>00:02.210</strong>: <a class="reference internal" href="use_pass_instrument.html#sphx-glr-how-to-extend-tvm-use-pass-instrument-py"><span class="std std-ref">How to Use TVM Pass Instrument</span></a> (<code class="docutils literal notranslate"><span class="pre">use_pass_instrument.py</span></code>)</p></li>
+<li><p><strong>00:01.048</strong>: <a class="reference internal" href="use_pass_infra.html#sphx-glr-how-to-extend-tvm-use-pass-infra-py"><span class="std std-ref">How to Use TVM Pass Infra</span></a> (<code class="docutils literal notranslate"><span class="pre">use_pass_infra.py</span></code>)</p></li>
+<li><p><strong>00:00.200</strong>: <a class="reference internal" href="low_level_custom_pass.html#sphx-glr-how-to-extend-tvm-low-level-custom-pass-py"><span class="std std-ref">Writing a Customized Pass</span></a> (<code class="docutils literal notranslate"><span class="pre">low_level_custom_pass.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index 6e0c92fc0..c55c07a65 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -486,10 +486,10 @@ profile the execution time of each passes.</p>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 5881us [5881us] (44.91%; 44.91%)
-FoldScaleAxis: 7213us [2us] (55.09%; 55.09%)
-        FoldConstant: 7211us [1499us] (55.07%; 99.97%)
-                InferType: 5712us [5712us] (43.62%; 79.21%)
+InferType: 5825us [5825us] (44.98%; 44.98%)
+FoldScaleAxis: 7124us [2us] (55.02%; 55.02%)
+        FoldConstant: 7122us [1502us] (55.00%; 99.97%)
+                InferType: 5620us [5620us] (43.40%; 78.92%)
 </pre></div>
 </div>
 </div>
@@ -512,10 +512,10 @@ Refer to following sections and <a class="reference internal" href="../../refere
 </div>
 <p class="sphx-glr-script-out">Out:</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 5735us [5735us] (44.33%; 44.33%)
-FoldScaleAxis: 7201us [2us] (55.67%; 55.67%)
-        FoldConstant: 7199us [1533us] (55.65%; 99.97%)
-                InferType: 5666us [5666us] (43.80%; 78.70%)
+InferType: 5729us [5729us] (44.57%; 44.57%)
+FoldScaleAxis: 7124us [2us] (55.43%; 55.43%)
+        FoldConstant: 7122us [1488us] (55.41%; 99.98%)
+                InferType: 5634us [5634us] (43.83%; 79.10%)
 </pre></div>
 </div>
 <p>Register empty list to clear existing instruments.</p>
diff --git a/docs/how_to/optimize_operators/opt_conv_cuda.html b/docs/how_to/optimize_operators/opt_conv_cuda.html
index e51998719..54acaef0b 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -534,7 +534,7 @@ latency of convolution.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 33.608237 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 44.464928 ms
 </pre></div>
 </div>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-optimize-operators-opt-conv-cuda-py">
diff --git a/docs/how_to/optimize_operators/opt_conv_tensorcore.html b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
index 1d81eab1a..8e3a25c41 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -876,7 +876,7 @@ be able to run on our build server</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 8.191855 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 6.923151 ms
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/optimize_operators/opt_gemm.html b/docs/how_to/optimize_operators/opt_gemm.html
index 2baabde11..7d80ed69f 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -431,8 +431,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.018082
-Baseline: 3.226418
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.018355
+Baseline: 3.388145
 </pre></div>
 </div>
 <p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -493,7 +493,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.302598
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.304776
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -561,7 +561,7 @@ vastly.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.333762
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.339592
 </pre></div>
 </div>
 <p>Here is the generated IR after vectorization.</p>
@@ -623,7 +623,7 @@ the access pattern for A matrix is more cache friendly.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.115119
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.116929
 </pre></div>
 </div>
 <p>Here is the generated IR after loop permutation.</p>
@@ -707,7 +707,7 @@ flattening.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.112020
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.110666
 </pre></div>
 </div>
 <p>Here is the generated IR after array packing.</p>
@@ -794,7 +794,7 @@ write to C when all the block results are ready.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.111046
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.110533
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -885,7 +885,7 @@ write to C when all the block results are ready.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.144175
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.145002
 </pre></div>
 </div>
 <p>Here is the generated IR after parallelization.</p>
diff --git a/docs/how_to/optimize_operators/sg_execution_times.html b/docs/how_to/optimize_operators/sg_execution_times.html
index 0b7335022..ab36ebd38 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -300,11 +300,11 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-optimize-operators-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:34.203</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:34.973</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:31.709</strong>: <a class="reference internal" href="opt_gemm.html#sphx-glr-how-to-optimize-operators-opt-gemm-py"><span class="std std-ref">How to optimize GEMM on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_gemm.py</span></code>)</p></li>
-<li><p><strong>00:01.346</strong>: <a class="reference internal" href="opt_conv_tensorcore.html#sphx-glr-how-to-optimize-operators-opt-conv-tensorcore-py"><span class="std std-ref">How to optimize convolution using TensorCores</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_conv_tensorcore.py</span></code>)</p></li>
-<li><p><strong>00:01.148</strong>: <a class="reference internal" href="opt_conv_cuda.html#sphx-glr-how-to-optimize-operators-opt-conv-cuda-py"><span class="std std-ref">How to optimize convolution on GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_conv_cuda.py</span></code>)</p></li>
+<li><p><strong>00:32.353</strong>: <a class="reference internal" href="opt_gemm.html#sphx-glr-how-to-optimize-operators-opt-gemm-py"><span class="std std-ref">How to optimize GEMM on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_gemm.py</span></code>)</p></li>
+<li><p><strong>00:01.398</strong>: <a class="reference internal" href="opt_conv_tensorcore.html#sphx-glr-how-to-optimize-operators-opt-conv-tensorcore-py"><span class="std std-ref">How to optimize convolution using TensorCores</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_conv_tensorcore.py</span></code>)</p></li>
+<li><p><strong>00:01.221</strong>: <a class="reference internal" href="opt_conv_cuda.html#sphx-glr-how-to-optimize-operators-opt-conv-cuda-py"><span class="std std-ref">How to optimize convolution on GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_conv_cuda.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
index 17d403de3..35d4f2054 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -300,14 +300,14 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-tune-with-autoscheduler-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>04:56.539</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>04:53.427</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
 <ul class="simple">
-<li><p><strong>02:17.771</strong>: <a class="reference internal" href="tune_conv2d_layer_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-conv2d-layer-cuda-py"><span class="std std-ref">Auto-scheduling a Convolution Layer for GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_layer_cuda.py</span></code>)</p></li>
-<li><p><strong>01:19.538</strong>: <a class="reference internal" href="tune_network_x86.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-x86-py"><span class="std std-ref">Auto-scheduling a Neural Network for x86 CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_x86.py</span></code>)</p></li>
-<li><p><strong>00:39.1000</strong>: <a class="reference internal" href="tune_network_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-cuda-py"><span class="std std-ref">Auto-scheduling a Neural Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_cuda.py</span></code>)</p></li>
-<li><p><strong>00:22.188</strong>: <a class="reference internal" href="tune_sparse_x86.html#sphx-glr-how-to-tune-with-autoscheduler-tune-sparse-x86-py"><span class="std std-ref">Auto-scheduling Sparse Matrix Multiplication on CPU with Custom Sketch Rule</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_sparse_x86.py</span></code>)</p></li>
-<li><p><strong>00:08.630</strong>: <a class="reference internal" href="tune_network_mali.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-mali-py"><span class="std std-ref">Auto-scheduling a Neural Network for mali GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_mali.py</span></code>)</p></li>
-<li><p><strong>00:08.413</strong>: <a class="reference internal" href="tune_network_arm.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-arm-py"><span class="std std-ref">Auto-scheduling a Neural Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_arm.py</span></code>)</p></li>
+<li><p><strong>02:20.154</strong>: <a class="reference internal" href="tune_conv2d_layer_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-conv2d-layer-cuda-py"><span class="std std-ref">Auto-scheduling a Convolution Layer for GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_layer_cuda.py</span></code>)</p></li>
+<li><p><strong>01:18.957</strong>: <a class="reference internal" href="tune_network_x86.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-x86-py"><span class="std std-ref">Auto-scheduling a Neural Network for x86 CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_x86.py</span></code>)</p></li>
+<li><p><strong>00:40.191</strong>: <a class="reference internal" href="tune_network_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-cuda-py"><span class="std std-ref">Auto-scheduling a Neural Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_cuda.py</span></code>)</p></li>
+<li><p><strong>00:17.223</strong>: <a class="reference internal" href="tune_sparse_x86.html#sphx-glr-how-to-tune-with-autoscheduler-tune-sparse-x86-py"><span class="std std-ref">Auto-scheduling Sparse Matrix Multiplication on CPU with Custom Sketch Rule</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_sparse_x86.py</span></code>)</p></li>
+<li><p><strong>00:08.638</strong>: <a class="reference internal" href="tune_network_mali.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-mali-py"><span class="std std-ref">Auto-scheduling a Neural Network for mali GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_mali.py</span></code>)</p></li>
+<li><p><strong>00:08.264</strong>: <a class="reference internal" href="tune_network_arm.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-arm-py"><span class="std std-ref">Auto-scheduling a Neural Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_arm.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.html b/docs/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.html
index a08906be1..e6a04fc92 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_conv2d_layer_cuda.html
@@ -471,116 +471,347 @@ cooperative fetching, unrolling and operator fusion.</p>
   buffer_map = {data_1: data, kernel_1: kernel, bias_1: bias, compute_1: compute} {
   attr [IterVar(blockIdx.x: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;blockIdx.x&quot;)] &quot;thread_extent&quot; = 32;
   allocate(conv2d_nchw: Pointer(local float32), float32, [28]), storage_scope = local;
-  allocate(pad_temp.shared: Pointer(shared float32), float32, [252]), storage_scope = shared;
-  allocate(kernel.shared: Pointer(shared float32), float32, [192]), storage_scope = shared;
+  allocate(pad_temp.shared: Pointer(shared float32), float32, [504]), storage_scope = shared;
+  allocate(kernel.shared: Pointer(shared float32), float32, [384]), storage_scope = shared;
   attr [IterVar(threadIdx.x: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28 {
-    conv2d_nchw_1: Buffer(conv2d_nchw, float32, [49], [], scope=&quot;local&quot;, align=16)[0] = 0f32
-    conv2d_nchw_1[7] = 0f32
-    conv2d_nchw_1[14] = 0f32
-    conv2d_nchw_1[21] = 0f32
-    conv2d_nchw_1[1] = 0f32
+    conv2d_nchw_1: Buffer(conv2d_nchw, float32, [16], [], scope=&quot;local&quot;, align=16)[0] = 0f32
+    conv2d_nchw_1[4] = 0f32
     conv2d_nchw_1[8] = 0f32
-    conv2d_nchw_1[15] = 0f32
-    conv2d_nchw_1[22] = 0f32
-    conv2d_nchw_1[2] = 0f32
-    conv2d_nchw_1[9] = 0f32
+    conv2d_nchw_1[12] = 0f32
     conv2d_nchw_1[16] = 0f32
-    conv2d_nchw_1[23] = 0f32
-    conv2d_nchw_1[3] = 0f32
-    conv2d_nchw_1[10] = 0f32
-    conv2d_nchw_1[17] = 0f32
+    conv2d_nchw_1[20] = 0f32
     conv2d_nchw_1[24] = 0f32
-    conv2d_nchw_1[4] = 0f32
-    conv2d_nchw_1[11] = 0f32
-    conv2d_nchw_1[18] = 0f32
-    conv2d_nchw_1[25] = 0f32
+    conv2d_nchw_1[1] = 0f32
     conv2d_nchw_1[5] = 0f32
-    conv2d_nchw_1[12] = 0f32
-    conv2d_nchw_1[19] = 0f32
-    conv2d_nchw_1[26] = 0f32
-    conv2d_nchw_1[6] = 0f32
+    conv2d_nchw_1[9] = 0f32
     conv2d_nchw_1[13] = 0f32
-    conv2d_nchw_1[20] = 0f32
+    conv2d_nchw_1[17] = 0f32
+    conv2d_nchw_1[21] = 0f32
+    conv2d_nchw_1[25] = 0f32
+    conv2d_nchw_1[2] = 0f32
+    conv2d_nchw_1[6] = 0f32
+    conv2d_nchw_1[10] = 0f32
+    conv2d_nchw_1[14] = 0f32
+    conv2d_nchw_1[18] = 0f32
+    conv2d_nchw_1[22] = 0f32
+    conv2d_nchw_1[26] = 0f32
+    conv2d_nchw_1[3] = 0f32
+    conv2d_nchw_1[7] = 0f32
+    conv2d_nchw_1[11] = 0f32
+    conv2d_nchw_1[15] = 0f32
+    conv2d_nchw_1[19] = 0f32
+    conv2d_nchw_1[23] = 0f32
     conv2d_nchw_1[27] = 0f32
-    for (rc.outer.outer: int32, 0, 128) {
-      for (rx.outer.outer: int32, 0, 3) {
-        let cse_var_2: int32 = (rc.outer.outer*196)
-        let cse_var_1: int32 = (rc.outer.outer*36)
+    for (rc.outer.outer: int32, 0, 64) {
+      for (ry.outer.outer: int32, 0, 3) {
+        let cse_var_4: int32 = (rc.outer.outer*392)
+        let cse_var_3: int32 = (ry.outer.outer*7)
+        let cse_var_2: int32 = (rc.outer.outer*72)
+        let cse_var_1: int32 = (ry.outer.outer*3)
          {
           attr [IterVar(threadIdx.x_1: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1: Buffer(pad_temp.shared, float32, [252], [], scope=&quot;shared&quot;)[threadIdx.x_1] = @tir.if_then_else((((7 &lt;= threadIdx.x_1) &amp;&amp; (1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((cse_var_2 + threadIdx.x_1) + rx.outer.outer) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1: Buffer(pad_temp.shared, float32, [504], [], scope=&quot;shared&quot;)[threadIdx.x_1] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 9) + ry.outer.outer)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 9))) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data[((((cse_var_4 + (floordiv(threadIdx.x_1, 9)*7)) + cse_var_3) + floormod(threadIdx.x_1, 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 28)] = @tir.if_then_else(((((floordiv((threadIdx.x_1 + 28), 9) + ry.outer.outer) &lt; 8) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 1), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 1), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 28), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 1), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 56)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 56), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 56), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 2), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 2), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 56), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 84)] = @tir.if_then_else(((1 &lt;= floormod((threadIdx.x_1 + 3), 9)) &amp;&amp; (floormod((threadIdx.x_1 + 3), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 84), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 3), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 112)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 112), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 112), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 4), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 4), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 112), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 4), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 140)] = @tir.if_then_else(((1 &lt;= floormod((threadIdx.x_1 + 5), 9)) &amp;&amp; (floormod((threadIdx.x_1 + 5), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 140), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 5), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 168)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 168), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 168), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 6), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 6), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 168), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 6), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 196), 63), 9) + ry.outer.outer)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 7), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 7), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 196), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 7), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 224), 63), 9) + ry.outer.outer) &lt; 8) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 8), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 8), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 224), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 8), 9)) - 8)], 0f32, dtype=float32)
+          attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          pad_temp.shared_1[(threadIdx.x_1 + 252)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 9) + ry.outer.outer)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 9))) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data[((((cse_var_4 + (floordiv(threadIdx.x_1, 9)*7)) + cse_var_3) + floormod(threadIdx.x_1, 9)) + 188)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 28)] = @tir.if_then_else(((1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[((((cse_var_2 + ((floordiv(threadIdx.x_1, 7) + 4)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1[(threadIdx.x_1 + 280)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 280), 63), 9) + ry.outer.outer) &lt; 8) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 1), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 1), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 280), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 1), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 56)] = @tir.if_then_else(((((1 &lt;= floormod((floordiv(threadIdx.x_1, 7) + 8), 9)) &amp;&amp; (floormod((floordiv(threadIdx.x_1, 7) + 8), 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 8), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 8), 9)*7)) + rx.outer.outer) + floormod(thre [...]
+          pad_temp.shared_1[(threadIdx.x_1 + 308)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 308), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 308), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 2), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 2), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 308), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 84)] = @tir.if_then_else(((1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 12), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 3), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1[(threadIdx.x_1 + 336)] = @tir.if_then_else(((1 &lt;= floormod((threadIdx.x_1 + 3), 9)) &amp;&amp; (floormod((threadIdx.x_1 + 3), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 336), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 3), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 112)] = @tir.if_then_else(((((1 &lt;= floormod((floordiv(threadIdx.x_1, 7) + 7), 9)) &amp;&amp; (floormod((floordiv(threadIdx.x_1, 7) + 7), 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 16), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 7), 9)*7)) + rx.outer.outer) + floormod(th [...]
+          pad_temp.shared_1[(threadIdx.x_1 + 364)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 364), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 364), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 4), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 4), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 364), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 4), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 140)] = @tir.if_then_else(((1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 20), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 2), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else(((1 &lt;= floormod((threadIdx.x_1 + 5), 9)) &amp;&amp; (floormod((threadIdx.x_1 + 5), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 392), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 5), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 168)] = @tir.if_then_else(((((1 &lt;= floormod((floordiv(threadIdx.x_1, 7) + 6), 9)) &amp;&amp; (floormod((floordiv(threadIdx.x_1, 7) + 6), 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 24), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 6), 9)*7)) + rx.outer.outer) + floormod(th [...]
+          pad_temp.shared_1[(threadIdx.x_1 + 420)] = @tir.if_then_else(((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 420), 63), 9) + ry.outer.outer)) &amp;&amp; ((floordiv(floormod((threadIdx.x_1 + 420), 63), 9) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 6), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 6), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 420), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 6), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else(((1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 28), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 1), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1[(threadIdx.x_1 + 448)] = @tir.if_then_else((((1 &lt;= (floordiv(floormod((threadIdx.x_1 + 448), 63), 9) + ry.outer.outer)) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 7), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 7), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 448), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 7), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else((((floormod((floordiv(threadIdx.x_1, 7) + 5), 9) &lt; 8) &amp;&amp; (1 &lt;= (rx.outer.outer + floormod(threadIdx.x_1, 7)))) &amp;&amp; ((rx.outer.outer + floormod(threadIdx.x_1, 7)) &lt; 8)), data[(((((cse_var_2 + (floordiv((floordiv(threadIdx.x_1, 7) + 32), 9)*49)) + (floormod((floordiv(threadIdx.x_1, 7) + 5), 9)*7)) + rx.outer.outer) + floormod(threadIdx.x_1, 7)) - 8)], 0f32, dtype=float32)
+          pad_temp.shared_1[(threadIdx.x_1 + 476)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 476), 63), 9) + ry.outer.outer) &lt; 8) &amp;&amp; (1 &lt;= floormod((threadIdx.x_1 + 8), 9))) &amp;&amp; (floormod((threadIdx.x_1 + 8), 9) &lt; 8)), data[((((cse_var_4 + (floordiv((threadIdx.x_1 + 476), 9)*7)) + cse_var_3) + floormod((threadIdx.x_1 + 8), 9)) - 8)], 0f32, dtype=float32)
           attr [IterVar(threadIdx.x_2: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1: Buffer(kernel.shared, float32, [192], [], scope=&quot;shared&quot;)[threadIdx.x_2] = kernel[(((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 12)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer)]
+          kernel.shared_1: Buffer(kernel.shared, float32, [384], [], scope=&quot;shared&quot;)[threadIdx.x_2] = kernel[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 24)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1[(threadIdx.x_2 + 28)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 7), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 4), 12)*3)) + rx.outer.outer)]
+          kernel.shared_1[(threadIdx.x_2 + 28)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 7), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 28), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1[(threadIdx.x_2 + 56)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 14), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 8), 12)*3)) + rx.outer.outer)]
+          kernel.shared_1[(threadIdx.x_2 + 56)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 14), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 56), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1[(threadIdx.x_2 + 84)] = kernel[((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 3)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer) + 32256)]
+          kernel.shared_1[(threadIdx.x_2 + 84)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 21), 6)*4608)) + cse_var_2) + (floormod((floordiv(threadIdx.x_2, 3) + 4), 8)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1[(threadIdx.x_2 + 112)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 28), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 4), 12)*3)) + rx.outer.outer)]
+          kernel.shared_1[(threadIdx.x_2 + 112)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 28), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 112), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          kernel.shared_1[(threadIdx.x_2 + 140)] = kernel[(((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 35), 3)*4608)) + cse_var_1) + (floormod((threadIdx.x_2 + 8), 12)*3)) + rx.outer.outer)]
+          kernel.shared_1[(threadIdx.x_2 + 140)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 35), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 140), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
           attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
-          if @tir.likely((threadIdx.x_2 &lt; 24), dtype=bool) {
-            kernel.shared_1[(threadIdx.x_2 + 168)] = kernel[((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 3)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 12)*3)) + rx.outer.outer) + 64512)]
+          kernel.shared_1[(threadIdx.x_2 + 168)] = kernel[(((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 6)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 32256)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 196)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 49), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 196), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 224)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 56), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 224), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 252)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 63), 6)*4608)) + cse_var_2) + (floormod((floordiv(threadIdx.x_2, 3) + 4), 8)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 280)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 70), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 280), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 308)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 77), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 308), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          kernel.shared_1[(threadIdx.x_2 + 336)] = kernel[(((((((blockIdx.x*73728) + (floordiv(floordiv(threadIdx.x_2, 4), 6)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 64512)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 28;
+          if @tir.likely((threadIdx.x_2 &lt; 20), dtype=bool) {
+            kernel.shared_1[(threadIdx.x_2 + 364)] = kernel[((((((blockIdx.x*73728) + (floordiv((floordiv(threadIdx.x_2, 4) + 91), 6)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 364), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
           }
-          for (ry.outer.inner: int32, 0, 3) {
-            for (rc.inner: int32, 0, 4) {
-              conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[(((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7))]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-              conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner)]))
-              conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 48)]))
-              conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 96)]))
-              conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((((rc.inner*63) + (ry.outer.inner*7)) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((((floordiv(threadIdx.x, 7)*12) + (rc.inner*3)) + ry.outer.inner) + 144)]))
-            }
+          for (rx.outer.inner: int32, 0, 3) {
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + rx.outer.inner)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 3)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 6)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 9)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 12)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 15)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 18)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[16] = (conv2d_nchw_1[16] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[20] = (conv2d_nchw_1[20] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[24] = (conv2d_nchw_1[24] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 21)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 24)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 27)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 30)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 33)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 36)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 39)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 42)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[17] = (conv2d_nchw_1[17] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[21] = (conv2d_nchw_1[21] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[25] = (conv2d_nchw_1[25] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 45)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 48)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 51)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 54)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 57)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 60)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 63)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 66)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[18] = (conv2d_nchw_1[18] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[22] = (conv2d_nchw_1[22] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[26] = (conv2d_nchw_1[26] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 69)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(rx.outer.inner + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 9)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 18)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 27)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 36)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 45)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 54)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 72)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 63)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 72)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 81)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 90)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 99)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 108)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 117)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 75)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 126)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 135)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 144)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 153)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 162)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 171)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 180)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 78)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 189)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 198)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 207)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 216)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 225)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 234)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 243)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 81)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 252)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 261)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 270)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 279)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 288)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 297)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 306)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 84)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 315)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 324)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 333)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 342)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 351)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 360)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 369)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 87)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 378)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 387)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 396)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 405)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 414)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 423)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 432)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 90)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 441)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 450)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 459)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 468)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[19] = (conv2d_nchw_1[19] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 477)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[23] = (conv2d_nchw_1[23] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 486)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
+            conv2d_nchw_1[27] = (conv2d_nchw_1[27] + (pad_temp.shared_1[((rx.outer.inner + floormod(threadIdx.x, 7)) + 495)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + rx.outer.inner) + 93)]))
           }
         }
       }
     }
-    for (i2.inner: int32, 0, 7) {
-      compute[((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[i2.inner] + bias[((blockIdx.x*16) + floordiv(threadIdx.x, 7))]), 0f32)
-      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 196)] = max((conv2d_nchw_1[(i2.inner + 7)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 4)]), 0f32)
-      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 392)] = max((conv2d_nchw_1[(i2.inner + 14)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 8)]), 0f32)
-      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7)) + 588)] = max((conv2d_nchw_1[(i2.inner + 21)] + bias[(((blockIdx.x*16) + floordiv(threadIdx.x, 7)) + 12)]), 0f32)
+    for (i1.inner: int32, 0, 4) {
+      compute[((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[i1.inner] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 7)] = max((conv2d_nchw_1[(i1.inner + 4)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 14)] = max((conv2d_nchw_1[(i1.inner + 8)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 21)] = max((conv2d_nchw_1[(i1.inner + 12)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 28)] = max((conv2d_nchw_1[(i1.inner + 16)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 35)] = max((conv2d_nchw_1[(i1.inner + 20)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
+      compute[(((((blockIdx.x*784) + (floordiv(threadIdx.x, 7)*196)) + (i1.inner*49)) + floormod(threadIdx.x, 7)) + 42)] = max((conv2d_nchw_1[(i1.inner + 24)] + bias[(((blockIdx.x*16) + (floordiv(threadIdx.x, 7)*4)) + i1.inner)]), 0f32)
     }
   }
 }
@@ -618,7 +849,7 @@ cooperative fetching, unrolling and operator fusion.</p>
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.351 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.315 ms
 </pre></div>
 </div>
 </div>
@@ -649,33 +880,33 @@ conv2d_nchw_nn_o_o_i, conv2d_nchw_nn_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o
 conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
 conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
 conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
-conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
+conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=4)
 conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=4)
-conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=4)
-conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=7)
+conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=1)
+conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=1)
 conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
 conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
-conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
+conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=7)
 conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
 conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
 conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=7)
 conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
-conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
+conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=8)
 conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=1)
 conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=1)
-conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
+conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=1)
 conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=1)
-conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
 s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2d_nc [...]
 compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
 compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
 compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
+compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=4)
 compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=4)
-compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=4)
-compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=7)
+compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
+compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
 compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
-compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
+compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=7)
 compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
 compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=7)
 compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
@@ -704,7 +935,7 @@ pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, pad_temp_shared_ax0_ax1_fus
 s[pad_temp_shared].vectorize(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i)
 pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=28)
 s[pad_temp_shared].bind(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i, te.thread_axis(&quot;threadIdx.x&quot;))
-s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, &quot;auto_unroll_max_step&quot;, 64)
+s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, &quot;auto_unroll_max_step&quot;, 512)
 s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, &quot;unroll_explicit&quot;, True)
 
 CUDA source code:
@@ -724,97 +955,310 @@ CUDA source code:
 #endif
 extern &quot;C&quot; __global__ void __launch_bounds__(28) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
   float conv2d_nchw[28];
-  __shared__ float pad_temp_shared[252];
-  __shared__ float kernel_shared[192];
+  __shared__ float pad_temp_shared[504];
+  __shared__ float kernel_shared[384];
   conv2d_nchw[0] = 0.000000e+00f;
-  conv2d_nchw[7] = 0.000000e+00f;
-  conv2d_nchw[14] = 0.000000e+00f;
-  conv2d_nchw[21] = 0.000000e+00f;
-  conv2d_nchw[1] = 0.000000e+00f;
+  conv2d_nchw[4] = 0.000000e+00f;
   conv2d_nchw[8] = 0.000000e+00f;
-  conv2d_nchw[15] = 0.000000e+00f;
-  conv2d_nchw[22] = 0.000000e+00f;
-  conv2d_nchw[2] = 0.000000e+00f;
-  conv2d_nchw[9] = 0.000000e+00f;
+  conv2d_nchw[12] = 0.000000e+00f;
   conv2d_nchw[16] = 0.000000e+00f;
-  conv2d_nchw[23] = 0.000000e+00f;
-  conv2d_nchw[3] = 0.000000e+00f;
-  conv2d_nchw[10] = 0.000000e+00f;
-  conv2d_nchw[17] = 0.000000e+00f;
+  conv2d_nchw[20] = 0.000000e+00f;
   conv2d_nchw[24] = 0.000000e+00f;
-  conv2d_nchw[4] = 0.000000e+00f;
-  conv2d_nchw[11] = 0.000000e+00f;
-  conv2d_nchw[18] = 0.000000e+00f;
-  conv2d_nchw[25] = 0.000000e+00f;
+  conv2d_nchw[1] = 0.000000e+00f;
   conv2d_nchw[5] = 0.000000e+00f;
-  conv2d_nchw[12] = 0.000000e+00f;
-  conv2d_nchw[19] = 0.000000e+00f;
-  conv2d_nchw[26] = 0.000000e+00f;
-  conv2d_nchw[6] = 0.000000e+00f;
+  conv2d_nchw[9] = 0.000000e+00f;
   conv2d_nchw[13] = 0.000000e+00f;
-  conv2d_nchw[20] = 0.000000e+00f;
+  conv2d_nchw[17] = 0.000000e+00f;
+  conv2d_nchw[21] = 0.000000e+00f;
+  conv2d_nchw[25] = 0.000000e+00f;
+  conv2d_nchw[2] = 0.000000e+00f;
+  conv2d_nchw[6] = 0.000000e+00f;
+  conv2d_nchw[10] = 0.000000e+00f;
+  conv2d_nchw[14] = 0.000000e+00f;
+  conv2d_nchw[18] = 0.000000e+00f;
+  conv2d_nchw[22] = 0.000000e+00f;
+  conv2d_nchw[26] = 0.000000e+00f;
+  conv2d_nchw[3] = 0.000000e+00f;
+  conv2d_nchw[7] = 0.000000e+00f;
+  conv2d_nchw[11] = 0.000000e+00f;
+  conv2d_nchw[15] = 0.000000e+00f;
+  conv2d_nchw[19] = 0.000000e+00f;
+  conv2d_nchw[23] = 0.000000e+00f;
   conv2d_nchw[27] = 0.000000e+00f;
-  for (int rc_outer_outer = 0; rc_outer_outer &lt; 128; ++rc_outer_outer) {
-    for (int rx_outer_outer = 0; rx_outer_outer &lt; 3; ++rx_outer_outer) {
+  for (int rc_outer_outer = 0; rc_outer_outer &lt; 64; ++rc_outer_outer) {
+    for (int ry_outer_outer = 0; ry_outer_outer &lt; 3; ++ry_outer_outer) {
       __syncthreads();
-      pad_temp_shared[((int)threadIdx.x)] = ((((7 &lt;= ((int)threadIdx.x)) &amp;&amp; (1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7)))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((rc_outer_outer * 196) + ((int)threadIdx.x)) + rx_outer_outer) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 28)] = (((1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((rc_outer_outer * 196) + ((int)threadIdx.x)) + rx_outer_outer) + 20)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 56)] = (((((1 &lt;= (((((int)threadIdx.x) / 7) + 8) % 9)) &amp;&amp; ((((((int)threadIdx.x) / 7) + 8) % 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7)))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 56) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 84)] = (((1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 84) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 3) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 &lt;= (((((int)threadIdx.x) / 7) + 7) % 9)) &amp;&amp; ((((((int)threadIdx.x) / 7) + 7) % 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7)))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 112) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 140)] = (((1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 140) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 2) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 168)] = (((((1 &lt;= (((((int)threadIdx.x) / 7) + 6) % 9)) &amp;&amp; ((((((int)threadIdx.x) / 7) + 6) % 9) &lt; 8)) &amp;&amp; (1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7)))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 168) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 196)] = (((1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 196) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 1) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 224)] = ((((((int)threadIdx.x) &lt; 21) &amp;&amp; (1 &lt;= (rx_outer_outer + (((int)threadIdx.x) % 7)))) &amp;&amp; ((rx_outer_outer + (((int)threadIdx.x) % 7)) &lt; 8)) ? data[((((((rc_outer_outer * 196) + (((((int)threadIdx.x) + 224) / 63) * 49)) + (((((int)threadIdx.x) / 7) + 5) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
-      kernel_shared[((int)threadIdx.x)] = kernel[(((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer)];
-      kernel_shared[(((int)threadIdx.x) + 28)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 28) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 4) % 12) * 3)) + rx_outer_outer)];
-      kernel_shared[(((int)threadIdx.x) + 56)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 56) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 8) % 12) * 3)) + rx_outer_outer)];
-      kernel_shared[(((int)threadIdx.x) + 84)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer) + 32256)];
-      kernel_shared[(((int)threadIdx.x) + 112)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 112) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 4) % 12) * 3)) + rx_outer_outer)];
-      kernel_shared[(((int)threadIdx.x) + 140)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 140) / 12) * 4608)) + (rc_outer_outer * 36)) + (((((int)threadIdx.x) + 8) % 12) * 3)) + rx_outer_outer)];
-      if (((int)threadIdx.x) &lt; 24) {
-        kernel_shared[(((int)threadIdx.x) + 168)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 12) * 4608)) + (rc_outer_outer * 36)) + ((((int)threadIdx.x) % 12) * 3)) + rx_outer_outer) + 64512)];
+      pad_temp_shared[((int)threadIdx.x)] = ((((1 &lt;= ((((int)threadIdx.x) / 9) + ry_outer_outer)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 9))) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 28)] = (((((((((int)threadIdx.x) + 28) / 9) + ry_outer_outer) &lt; 8) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 1) % 9))) &amp;&amp; (((((int)threadIdx.x) + 1) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 28) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 56)] = (((((1 &lt;= ((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 2) % 9))) &amp;&amp; (((((int)threadIdx.x) + 2) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 56) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 84)] = (((1 &lt;= ((((int)threadIdx.x) + 3) % 9)) &amp;&amp; (((((int)threadIdx.x) + 3) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 84) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 &lt;= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 4) % 9))) &amp;&amp; (((((int)threadIdx.x) + 4) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 112) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 140)] = (((1 &lt;= ((((int)threadIdx.x) + 5) % 9)) &amp;&amp; (((((int)threadIdx.x) + 5) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 140) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 168)] = (((((1 &lt;= ((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 6) % 9))) &amp;&amp; (((((int)threadIdx.x) + 6) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 168) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 &lt;= ((((((int)threadIdx.x) + 7) % 63) / 9) + ry_outer_outer)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 7) % 9))) &amp;&amp; (((((int)threadIdx.x) + 7) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 196) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 224)] = ((((((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) &lt; 8) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 8) % 9))) &amp;&amp; (((((int)threadIdx.x) + 8) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 224) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 252)] = ((((1 &lt;= ((((int)threadIdx.x) / 9) + ry_outer_outer)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 9))) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) + 188)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 280)] = ((((((((((int)threadIdx.x) + 28) % 63) / 9) + ry_outer_outer) &lt; 8) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 1) % 9))) &amp;&amp; (((((int)threadIdx.x) + 1) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 280) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 308)] = (((((1 &lt;= ((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 56) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 2) % 9))) &amp;&amp; (((((int)threadIdx.x) + 2) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 308) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 336)] = (((1 &lt;= ((((int)threadIdx.x) + 3) % 9)) &amp;&amp; (((((int)threadIdx.x) + 3) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 336) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 364)] = (((((1 &lt;= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 4) % 9))) &amp;&amp; (((((int)threadIdx.x) + 4) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 364) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 392)] = (((1 &lt;= ((((int)threadIdx.x) + 5) % 9)) &amp;&amp; (((((int)threadIdx.x) + 5) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 392) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 420)] = (((((1 &lt;= ((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer)) &amp;&amp; (((((((int)threadIdx.x) + 42) % 63) / 9) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 6) % 9))) &amp;&amp; (((((int)threadIdx.x) + 6) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 420) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 448)] = ((((1 &lt;= ((((((int)threadIdx.x) + 7) % 63) / 9) + ry_outer_outer)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 7) % 9))) &amp;&amp; (((((int)threadIdx.x) + 7) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 448) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+      pad_temp_shared[(((int)threadIdx.x) + 476)] = ((((((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) &lt; 8) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) + 8) % 9))) &amp;&amp; (((((int)threadIdx.x) + 8) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 476) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
+      kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 28)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 28) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 56)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 56) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 84)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 84) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) / 3) + 4) &amp; 7) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 112)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 112) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 140)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 140) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 20) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 168)] = kernel[(((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 32256)];
+      kernel_shared[(((int)threadIdx.x) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 196) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 224) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 252)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 252) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) / 3) + 4) &amp; 7) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 280)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 308)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 308) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 20) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 64512)];
+      if (((int)threadIdx.x) &lt; 20) {
+        kernel_shared[(((int)threadIdx.x) + 364)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 364) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 4) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
       }
       __syncthreads();
-      for (int ry_outer_inner = 0; ry_outer_inner &lt; 3; ++ry_outer_inner) {
-        for (int rc_inner = 0; rc_inner &lt; 4; ++rc_inner) {
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner)]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 48)]));
-          conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 96)]));
-          conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((((rc_inner * 63) + (ry_outer_inner * 7)) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((((int)threadIdx.x) / 7) * 12) + (rc_inner * 3)) + ry_outer_inner) + 144)]));
-        }
+      for (int rx_outer_inner = 0; rx_outer_inner &lt; 3; ++rx_outer_inner) {
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + rx_outer_inner)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 3)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 6)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 9)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 12)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 15)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 18)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 21)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 24)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 27)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 30)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 33)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 36)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 39)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 42)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 45)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 48)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 51)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 54)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 57)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 60)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 63)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 66)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 69)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(rx_outer_inner + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 9)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 18)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 27)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 36)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 45)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 54)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 72)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 63)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 72)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 81)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 90)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 99)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 108)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 117)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 75)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 126)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 135)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 144)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 153)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 162)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 171)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 180)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 78)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 189)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 198)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 207)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 216)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 225)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 234)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 243)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 81)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 252)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 261)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 270)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 279)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 288)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 297)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 306)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 84)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 315)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 324)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 333)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 342)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 351)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 360)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 369)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 87)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 378)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 387)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 396)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 405)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 414)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 423)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 432)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 90)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 441)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 450)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 459)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 468)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 477)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 486)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
+        conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[((rx_outer_inner + (((int)threadIdx.x) % 7)) + 495)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + rx_outer_inner) + 93)]));
       }
     }
   }
-  for (int i2_inner = 0; i2_inner &lt; 7; ++i2_inner) {
-    compute[((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[i2_inner] + bias[((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
-    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 196)] = max((conv2d_nchw[(i2_inner + 7)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 4)]), 0.000000e+00f);
-    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 392)] = max((conv2d_nchw[(i2_inner + 14)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 8)]), 0.000000e+00f);
-    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7)) + 588)] = max((conv2d_nchw[(i2_inner + 21)] + bias[(((((int)blockIdx.x) * 16) + (((int)threadIdx.x) / 7)) + 12)]), 0.000000e+00f);
+  for (int i1_inner = 0; i1_inner &lt; 4; ++i1_inner) {
+    compute[((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[i1_inner] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 7)] = max((conv2d_nchw[(i1_inner + 4)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 14)] = max((conv2d_nchw[(i1_inner + 8)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 21)] = max((conv2d_nchw[(i1_inner + 12)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 28)] = max((conv2d_nchw[(i1_inner + 16)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 35)] = max((conv2d_nchw[(i1_inner + 20)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+    compute[(((((((int)blockIdx.x) * 784) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + (((int)threadIdx.x) % 7)) + 42)] = max((conv2d_nchw[(i1_inner + 24)] + bias[(((((int)blockIdx.x) * 16) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
   }
 }
 </pre></div>
@@ -852,7 +1296,7 @@ In the example below we resume the status and do more 5 trials.</p>
 Get devices for measurement successfully!
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  17.771 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  20.154 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autoscheduler-tune-conv2d-layer-cuda-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/e3e540f3b477c0c52d8eb73e674e8ffd/tune_conv2d_layer_cuda.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">tune_conv2d_layer_cuda.py</span></code></a></p>
diff --git a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
index 1609842cb..1ca7f93fb 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -876,7 +876,7 @@ so we can read the log file and load the best schedules.</p>
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-   9.6715       9.6668       9.6970       9.6506       0.0193
+   9.8233       9.8322       9.8639       9.7739       0.0372
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
index 1e113a707..16fa477ad 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -895,7 +895,7 @@ so we can read the log file and load the best schedules.</p>
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  762.0476     761.2159     765.8830     759.0438      2.8534
+  755.6295     751.7929     764.4012     750.6943      6.2188
 </pre></div>
 </div>
 </div>
@@ -917,7 +917,7 @@ to learn how to use the RPC Tracker and RPC Server.
 To use the RPC Tracker in auto-scheduler, replace the runner in <code class="code docutils literal notranslate"><span class="pre">TuningOptions</span></code>
 with <a class="reference internal" href="../../reference/api/python/auto_scheduler.html#tvm.auto_scheduler.RPCRunner" title="tvm.auto_scheduler.RPCRunner"><code class="xref any py py-class docutils literal notranslate"><span class="pre">auto_scheduler.RPCRunner</span></code></a>.</p></li>
 </ol>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  19.538 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  18.957 seconds)</p>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autoscheduler-tune-network-x86-py">
 <div class="sphx-glr-download docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/e416b94ca1090b0897c0f6e0df95b911/tune_network_x86.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">tune_network_x86.py</span></code></a></p>
diff --git a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
index 83dc60614..e85558411 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -600,790 +600,407 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
              placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [65536], []),
              compute: Buffer(compute_2: Pointer(float32), float32, [65536], [])}
   buffer_map = {placeholder_5: placeholder, placeholder_6: placeholder_1, placeholder_7: placeholder_2, placeholder_8: placeholder_3, placeholder_9: placeholder_4, compute_1: compute} {
-  for (i0.outer: int32, 0, 2) &quot;parallel&quot; {
-    allocate(compute_3: Pointer(global float32), float32, [1024]), storage_scope = global;
-    for (i1.outer: int32, 0, 32) {
-      for (i.outer.inner: int32, 0, 4) {
-        let cse_var_1: int32 = (i.outer.inner*256)
+  for (i0.outer: int32, 0, 16) &quot;parallel&quot; {
+    allocate(compute_3: Pointer(global float32), float32, [256]), storage_scope = global;
+    for (i1.outer: int32, 0, 16) {
+      for (nb_j.inner: int32, 0, 2) {
+        let cse_var_2: int32 = (nb_j.inner*16)
+        let cse_var_1: int32 = ((i1.outer*2) + nb_j.inner)
          {
-          compute_4: Buffer(compute_3, float32, [1024], [])[cse_var_1] = 0f32
-          compute_4[(cse_var_1 + 1)] = 0f32
-          compute_4[(cse_var_1 + 2)] = 0f32
-          compute_4[(cse_var_1 + 3)] = 0f32
-          compute_4[(cse_var_1 + 4)] = 0f32
-          compute_4[(cse_var_1 + 5)] = 0f32
-          compute_4[(cse_var_1 + 6)] = 0f32
-          compute_4[(cse_var_1 + 7)] = 0f32
-          compute_4[(cse_var_1 + 8)] = 0f32
-          compute_4[(cse_var_1 + 9)] = 0f32
-          compute_4[(cse_var_1 + 10)] = 0f32
-          compute_4[(cse_var_1 + 11)] = 0f32
-          compute_4[(cse_var_1 + 12)] = 0f32
-          compute_4[(cse_var_1 + 13)] = 0f32
-          compute_4[(cse_var_1 + 14)] = 0f32
-          compute_4[(cse_var_1 + 15)] = 0f32
-          compute_4[(cse_var_1 + 16)] = 0f32
-          compute_4[(cse_var_1 + 17)] = 0f32
-          compute_4[(cse_var_1 + 18)] = 0f32
-          compute_4[(cse_var_1 + 19)] = 0f32
-          compute_4[(cse_var_1 + 20)] = 0f32
-          compute_4[(cse_var_1 + 21)] = 0f32
-          compute_4[(cse_var_1 + 22)] = 0f32
-          compute_4[(cse_var_1 + 23)] = 0f32
-          compute_4[(cse_var_1 + 24)] = 0f32
-          compute_4[(cse_var_1 + 25)] = 0f32
-          compute_4[(cse_var_1 + 26)] = 0f32
-          compute_4[(cse_var_1 + 27)] = 0f32
-          compute_4[(cse_var_1 + 28)] = 0f32
-          compute_4[(cse_var_1 + 29)] = 0f32
-          compute_4[(cse_var_1 + 30)] = 0f32
-          compute_4[(cse_var_1 + 31)] = 0f32
-          compute_4[(cse_var_1 + 32)] = 0f32
-          compute_4[(cse_var_1 + 33)] = 0f32
-          compute_4[(cse_var_1 + 34)] = 0f32
-          compute_4[(cse_var_1 + 35)] = 0f32
-          compute_4[(cse_var_1 + 36)] = 0f32
-          compute_4[(cse_var_1 + 37)] = 0f32
-          compute_4[(cse_var_1 + 38)] = 0f32
-          compute_4[(cse_var_1 + 39)] = 0f32
-          compute_4[(cse_var_1 + 40)] = 0f32
-          compute_4[(cse_var_1 + 41)] = 0f32
-          compute_4[(cse_var_1 + 42)] = 0f32
-          compute_4[(cse_var_1 + 43)] = 0f32
-          compute_4[(cse_var_1 + 44)] = 0f32
-          compute_4[(cse_var_1 + 45)] = 0f32
-          compute_4[(cse_var_1 + 46)] = 0f32
-          compute_4[(cse_var_1 + 47)] = 0f32
-          compute_4[(cse_var_1 + 48)] = 0f32
-          compute_4[(cse_var_1 + 49)] = 0f32
-          compute_4[(cse_var_1 + 50)] = 0f32
-          compute_4[(cse_var_1 + 51)] = 0f32
-          compute_4[(cse_var_1 + 52)] = 0f32
-          compute_4[(cse_var_1 + 53)] = 0f32
-          compute_4[(cse_var_1 + 54)] = 0f32
-          compute_4[(cse_var_1 + 55)] = 0f32
-          compute_4[(cse_var_1 + 56)] = 0f32
-          compute_4[(cse_var_1 + 57)] = 0f32
-          compute_4[(cse_var_1 + 58)] = 0f32
-          compute_4[(cse_var_1 + 59)] = 0f32
-          compute_4[(cse_var_1 + 60)] = 0f32
-          compute_4[(cse_var_1 + 61)] = 0f32
-          compute_4[(cse_var_1 + 62)] = 0f32
-          compute_4[(cse_var_1 + 63)] = 0f32
-          compute_4[(cse_var_1 + 64)] = 0f32
-          compute_4[(cse_var_1 + 65)] = 0f32
-          compute_4[(cse_var_1 + 66)] = 0f32
-          compute_4[(cse_var_1 + 67)] = 0f32
-          compute_4[(cse_var_1 + 68)] = 0f32
-          compute_4[(cse_var_1 + 69)] = 0f32
-          compute_4[(cse_var_1 + 70)] = 0f32
-          compute_4[(cse_var_1 + 71)] = 0f32
-          compute_4[(cse_var_1 + 72)] = 0f32
-          compute_4[(cse_var_1 + 73)] = 0f32
-          compute_4[(cse_var_1 + 74)] = 0f32
-          compute_4[(cse_var_1 + 75)] = 0f32
-          compute_4[(cse_var_1 + 76)] = 0f32
-          compute_4[(cse_var_1 + 77)] = 0f32
-          compute_4[(cse_var_1 + 78)] = 0f32
-          compute_4[(cse_var_1 + 79)] = 0f32
-          compute_4[(cse_var_1 + 80)] = 0f32
-          compute_4[(cse_var_1 + 81)] = 0f32
-          compute_4[(cse_var_1 + 82)] = 0f32
-          compute_4[(cse_var_1 + 83)] = 0f32
-          compute_4[(cse_var_1 + 84)] = 0f32
-          compute_4[(cse_var_1 + 85)] = 0f32
-          compute_4[(cse_var_1 + 86)] = 0f32
-          compute_4[(cse_var_1 + 87)] = 0f32
-          compute_4[(cse_var_1 + 88)] = 0f32
-          compute_4[(cse_var_1 + 89)] = 0f32
-          compute_4[(cse_var_1 + 90)] = 0f32
-          compute_4[(cse_var_1 + 91)] = 0f32
-          compute_4[(cse_var_1 + 92)] = 0f32
-          compute_4[(cse_var_1 + 93)] = 0f32
-          compute_4[(cse_var_1 + 94)] = 0f32
-          compute_4[(cse_var_1 + 95)] = 0f32
-          compute_4[(cse_var_1 + 96)] = 0f32
-          compute_4[(cse_var_1 + 97)] = 0f32
-          compute_4[(cse_var_1 + 98)] = 0f32
-          compute_4[(cse_var_1 + 99)] = 0f32
-          compute_4[(cse_var_1 + 100)] = 0f32
-          compute_4[(cse_var_1 + 101)] = 0f32
-          compute_4[(cse_var_1 + 102)] = 0f32
-          compute_4[(cse_var_1 + 103)] = 0f32
-          compute_4[(cse_var_1 + 104)] = 0f32
-          compute_4[(cse_var_1 + 105)] = 0f32
-          compute_4[(cse_var_1 + 106)] = 0f32
-          compute_4[(cse_var_1 + 107)] = 0f32
-          compute_4[(cse_var_1 + 108)] = 0f32
-          compute_4[(cse_var_1 + 109)] = 0f32
-          compute_4[(cse_var_1 + 110)] = 0f32
-          compute_4[(cse_var_1 + 111)] = 0f32
-          compute_4[(cse_var_1 + 112)] = 0f32
-          compute_4[(cse_var_1 + 113)] = 0f32
-          compute_4[(cse_var_1 + 114)] = 0f32
-          compute_4[(cse_var_1 + 115)] = 0f32
-          compute_4[(cse_var_1 + 116)] = 0f32
-          compute_4[(cse_var_1 + 117)] = 0f32
-          compute_4[(cse_var_1 + 118)] = 0f32
-          compute_4[(cse_var_1 + 119)] = 0f32
-          compute_4[(cse_var_1 + 120)] = 0f32
-          compute_4[(cse_var_1 + 121)] = 0f32
-          compute_4[(cse_var_1 + 122)] = 0f32
-          compute_4[(cse_var_1 + 123)] = 0f32
-          compute_4[(cse_var_1 + 124)] = 0f32
-          compute_4[(cse_var_1 + 125)] = 0f32
-          compute_4[(cse_var_1 + 126)] = 0f32
-          compute_4[(cse_var_1 + 127)] = 0f32
-          compute_4[(cse_var_1 + 128)] = 0f32
-          compute_4[(cse_var_1 + 129)] = 0f32
-          compute_4[(cse_var_1 + 130)] = 0f32
-          compute_4[(cse_var_1 + 131)] = 0f32
-          compute_4[(cse_var_1 + 132)] = 0f32
-          compute_4[(cse_var_1 + 133)] = 0f32
-          compute_4[(cse_var_1 + 134)] = 0f32
-          compute_4[(cse_var_1 + 135)] = 0f32
-          compute_4[(cse_var_1 + 136)] = 0f32
-          compute_4[(cse_var_1 + 137)] = 0f32
-          compute_4[(cse_var_1 + 138)] = 0f32
-          compute_4[(cse_var_1 + 139)] = 0f32
-          compute_4[(cse_var_1 + 140)] = 0f32
-          compute_4[(cse_var_1 + 141)] = 0f32
-          compute_4[(cse_var_1 + 142)] = 0f32
-          compute_4[(cse_var_1 + 143)] = 0f32
-          compute_4[(cse_var_1 + 144)] = 0f32
-          compute_4[(cse_var_1 + 145)] = 0f32
-          compute_4[(cse_var_1 + 146)] = 0f32
-          compute_4[(cse_var_1 + 147)] = 0f32
-          compute_4[(cse_var_1 + 148)] = 0f32
-          compute_4[(cse_var_1 + 149)] = 0f32
-          compute_4[(cse_var_1 + 150)] = 0f32
-          compute_4[(cse_var_1 + 151)] = 0f32
-          compute_4[(cse_var_1 + 152)] = 0f32
-          compute_4[(cse_var_1 + 153)] = 0f32
-          compute_4[(cse_var_1 + 154)] = 0f32
-          compute_4[(cse_var_1 + 155)] = 0f32
-          compute_4[(cse_var_1 + 156)] = 0f32
-          compute_4[(cse_var_1 + 157)] = 0f32
-          compute_4[(cse_var_1 + 158)] = 0f32
-          compute_4[(cse_var_1 + 159)] = 0f32
-          compute_4[(cse_var_1 + 160)] = 0f32
-          compute_4[(cse_var_1 + 161)] = 0f32
-          compute_4[(cse_var_1 + 162)] = 0f32
-          compute_4[(cse_var_1 + 163)] = 0f32
-          compute_4[(cse_var_1 + 164)] = 0f32
-          compute_4[(cse_var_1 + 165)] = 0f32
-          compute_4[(cse_var_1 + 166)] = 0f32
-          compute_4[(cse_var_1 + 167)] = 0f32
-          compute_4[(cse_var_1 + 168)] = 0f32
-          compute_4[(cse_var_1 + 169)] = 0f32
-          compute_4[(cse_var_1 + 170)] = 0f32
-          compute_4[(cse_var_1 + 171)] = 0f32
-          compute_4[(cse_var_1 + 172)] = 0f32
-          compute_4[(cse_var_1 + 173)] = 0f32
-          compute_4[(cse_var_1 + 174)] = 0f32
-          compute_4[(cse_var_1 + 175)] = 0f32
-          compute_4[(cse_var_1 + 176)] = 0f32
-          compute_4[(cse_var_1 + 177)] = 0f32
-          compute_4[(cse_var_1 + 178)] = 0f32
-          compute_4[(cse_var_1 + 179)] = 0f32
-          compute_4[(cse_var_1 + 180)] = 0f32
-          compute_4[(cse_var_1 + 181)] = 0f32
-          compute_4[(cse_var_1 + 182)] = 0f32
-          compute_4[(cse_var_1 + 183)] = 0f32
-          compute_4[(cse_var_1 + 184)] = 0f32
-          compute_4[(cse_var_1 + 185)] = 0f32
-          compute_4[(cse_var_1 + 186)] = 0f32
-          compute_4[(cse_var_1 + 187)] = 0f32
-          compute_4[(cse_var_1 + 188)] = 0f32
-          compute_4[(cse_var_1 + 189)] = 0f32
-          compute_4[(cse_var_1 + 190)] = 0f32
-          compute_4[(cse_var_1 + 191)] = 0f32
-          compute_4[(cse_var_1 + 192)] = 0f32
-          compute_4[(cse_var_1 + 193)] = 0f32
-          compute_4[(cse_var_1 + 194)] = 0f32
-          compute_4[(cse_var_1 + 195)] = 0f32
-          compute_4[(cse_var_1 + 196)] = 0f32
-          compute_4[(cse_var_1 + 197)] = 0f32
-          compute_4[(cse_var_1 + 198)] = 0f32
-          compute_4[(cse_var_1 + 199)] = 0f32
-          compute_4[(cse_var_1 + 200)] = 0f32
-          compute_4[(cse_var_1 + 201)] = 0f32
-          compute_4[(cse_var_1 + 202)] = 0f32
-          compute_4[(cse_var_1 + 203)] = 0f32
-          compute_4[(cse_var_1 + 204)] = 0f32
-          compute_4[(cse_var_1 + 205)] = 0f32
-          compute_4[(cse_var_1 + 206)] = 0f32
-          compute_4[(cse_var_1 + 207)] = 0f32
-          compute_4[(cse_var_1 + 208)] = 0f32
-          compute_4[(cse_var_1 + 209)] = 0f32
-          compute_4[(cse_var_1 + 210)] = 0f32
-          compute_4[(cse_var_1 + 211)] = 0f32
-          compute_4[(cse_var_1 + 212)] = 0f32
-          compute_4[(cse_var_1 + 213)] = 0f32
-          compute_4[(cse_var_1 + 214)] = 0f32
-          compute_4[(cse_var_1 + 215)] = 0f32
-          compute_4[(cse_var_1 + 216)] = 0f32
-          compute_4[(cse_var_1 + 217)] = 0f32
-          compute_4[(cse_var_1 + 218)] = 0f32
-          compute_4[(cse_var_1 + 219)] = 0f32
-          compute_4[(cse_var_1 + 220)] = 0f32
-          compute_4[(cse_var_1 + 221)] = 0f32
-          compute_4[(cse_var_1 + 222)] = 0f32
-          compute_4[(cse_var_1 + 223)] = 0f32
-          compute_4[(cse_var_1 + 224)] = 0f32
-          compute_4[(cse_var_1 + 225)] = 0f32
-          compute_4[(cse_var_1 + 226)] = 0f32
-          compute_4[(cse_var_1 + 227)] = 0f32
-          compute_4[(cse_var_1 + 228)] = 0f32
-          compute_4[(cse_var_1 + 229)] = 0f32
-          compute_4[(cse_var_1 + 230)] = 0f32
-          compute_4[(cse_var_1 + 231)] = 0f32
-          compute_4[(cse_var_1 + 232)] = 0f32
-          compute_4[(cse_var_1 + 233)] = 0f32
-          compute_4[(cse_var_1 + 234)] = 0f32
-          compute_4[(cse_var_1 + 235)] = 0f32
-          compute_4[(cse_var_1 + 236)] = 0f32
-          compute_4[(cse_var_1 + 237)] = 0f32
-          compute_4[(cse_var_1 + 238)] = 0f32
-          compute_4[(cse_var_1 + 239)] = 0f32
-          compute_4[(cse_var_1 + 240)] = 0f32
-          compute_4[(cse_var_1 + 241)] = 0f32
-          compute_4[(cse_var_1 + 242)] = 0f32
-          compute_4[(cse_var_1 + 243)] = 0f32
-          compute_4[(cse_var_1 + 244)] = 0f32
-          compute_4[(cse_var_1 + 245)] = 0f32
-          compute_4[(cse_var_1 + 246)] = 0f32
-          compute_4[(cse_var_1 + 247)] = 0f32
-          compute_4[(cse_var_1 + 248)] = 0f32
-          compute_4[(cse_var_1 + 249)] = 0f32
-          compute_4[(cse_var_1 + 250)] = 0f32
-          compute_4[(cse_var_1 + 251)] = 0f32
-          compute_4[(cse_var_1 + 252)] = 0f32
-          compute_4[(cse_var_1 + 253)] = 0f32
-          compute_4[(cse_var_1 + 254)] = 0f32
-          compute_4[(cse_var_1 + 255)] = 0f32
-          for (elem_idx: int32, 0, (placeholder_3[(i1.outer + 1)] - placeholder_3[i1.outer])) {
-            let cse_var_258: int32 = (cse_var_1 + 184)
-            let cse_var_257: int32 = (cse_var_1 + 183)
-            let cse_var_256: int32 = (cse_var_1 + 182)
-            let cse_var_255: int32 = (cse_var_1 + 181)
-            let cse_var_254: int32 = (cse_var_1 + 180)
-            let cse_var_253: int32 = (cse_var_1 + 18)
-            let cse_var_252: int32 = (cse_var_1 + 179)
-            let cse_var_251: int32 = (cse_var_1 + 178)
-            let cse_var_250: int32 = (cse_var_1 + 177)
-            let cse_var_249: int32 = (cse_var_1 + 176)
-            let cse_var_248: int32 = (cse_var_1 + 175)
-            let cse_var_247: int32 = (cse_var_1 + 174)
-            let cse_var_246: int32 = (cse_var_1 + 173)
-            let cse_var_245: int32 = (cse_var_1 + 172)
-            let cse_var_244: int32 = (cse_var_1 + 171)
-            let cse_var_243: int32 = (cse_var_1 + 214)
-            let cse_var_242: int32 = (cse_var_1 + 17)
-            let cse_var_241: int32 = (cse_var_1 + 169)
-            let cse_var_240: int32 = (cse_var_1 + 168)
-            let cse_var_239: int32 = (cse_var_1 + 167)
-            let cse_var_238: int32 = (cse_var_1 + 166)
-            let cse_var_237: int32 = (cse_var_1 + 165)
-            let cse_var_236: int32 = (cse_var_1 + 164)
-            let cse_var_235: int32 = (cse_var_1 + 163)
-            let cse_var_234: int32 = (cse_var_1 + 162)
-            let cse_var_233: int32 = (cse_var_1 + 161)
-            let cse_var_232: int32 = (cse_var_1 + 160)
-            let cse_var_231: int32 = (cse_var_1 + 16)
-            let cse_var_230: int32 = (cse_var_1 + 159)
-            let cse_var_229: int32 = (cse_var_1 + 158)
-            let cse_var_228: int32 = (cse_var_1 + 157)
-            let cse_var_227: int32 = (cse_var_1 + 170)
-            let cse_var_226: int32 = (cse_var_1 + 212)
-            let cse_var_225: int32 = (cse_var_1 + 211)
-            let cse_var_224: int32 = (cse_var_1 + 210)
-            let cse_var_223: int32 = (cse_var_1 + 21)
-            let cse_var_222: int32 = (cse_var_1 + 209)
-            let cse_var_221: int32 = (cse_var_1 + 208)
-            let cse_var_220: int32 = (cse_var_1 + 207)
-            let cse_var_219: int32 = (cse_var_1 + 206)
-            let cse_var_218: int32 = (cse_var_1 + 205)
-            let cse_var_217: int32 = (cse_var_1 + 204)
-            let cse_var_216: int32 = (cse_var_1 + 203)
-            let cse_var_215: int32 = (cse_var_1 + 202)
-            let cse_var_214: int32 = (cse_var_1 + 201)
-            let cse_var_213: int32 = (cse_var_1 + 200)
-            let cse_var_212: int32 = (cse_var_1 + 20)
-            let cse_var_211: int32 = (cse_var_1 + 185)
-            let cse_var_210: int32 = (cse_var_1 + 199)
-            let cse_var_209: int32 = (cse_var_1 + 198)
-            let cse_var_208: int32 = (cse_var_1 + 197)
-            let cse_var_207: int32 = (cse_var_1 + 196)
-            let cse_var_206: int32 = (cse_var_1 + 195)
-            let cse_var_205: int32 = (cse_var_1 + 194)
-            let cse_var_204: int32 = (cse_var_1 + 193)
-            let cse_var_203: int32 = (cse_var_1 + 192)
-            let cse_var_202: int32 = (cse_var_1 + 191)
-            let cse_var_201: int32 = (cse_var_1 + 190)
-            let cse_var_200: int32 = (cse_var_1 + 19)
-            let cse_var_199: int32 = (cse_var_1 + 189)
-            let cse_var_198: int32 = (cse_var_1 + 188)
-            let cse_var_197: int32 = (cse_var_1 + 187)
-            let cse_var_196: int32 = (cse_var_1 + 186)
-            let cse_var_195: int32 = (cse_var_1 + 2)
-            let cse_var_194: int32 = (cse_var_1 + 126)
-            let cse_var_193: int32 = (cse_var_1 + 125)
-            let cse_var_192: int32 = (cse_var_1 + 124)
-            let cse_var_191: int32 = (cse_var_1 + 123)
-            let cse_var_190: int32 = (cse_var_1 + 122)
-            let cse_var_189: int32 = (cse_var_1 + 121)
-            let cse_var_188: int32 = (cse_var_1 + 120)
-            let cse_var_187: int32 = (cse_var_1 + 12)
-            let cse_var_186: int32 = (cse_var_1 + 119)
-            let cse_var_185: int32 = (cse_var_1 + 118)
-            let cse_var_184: int32 = (cse_var_1 + 117)
-            let cse_var_183: int32 = (cse_var_1 + 116)
-            let cse_var_182: int32 = (cse_var_1 + 115)
-            let cse_var_181: int32 = (cse_var_1 + 114)
-            let cse_var_180: int32 = (cse_var_1 + 113)
-            let cse_var_179: int32 = (cse_var_1 + 156)
-            let cse_var_178: int32 = (cse_var_1 + 111)
-            let cse_var_177: int32 = (cse_var_1 + 110)
-            let cse_var_176: int32 = (cse_var_1 + 11)
-            let cse_var_175: int32 = (cse_var_1 + 109)
-            let cse_var_174: int32 = (cse_var_1 + 108)
-            let cse_var_173: int32 = (cse_var_1 + 107)
-            let cse_var_172: int32 = (cse_var_1 + 106)
-            let cse_var_171: int32 = (cse_var_1 + 105)
-            let cse_var_170: int32 = (cse_var_1 + 104)
-            let cse_var_169: int32 = (cse_var_1 + 103)
-            let cse_var_168: int32 = (cse_var_1 + 102)
-            let cse_var_167: int32 = (cse_var_1 + 101)
-            let cse_var_166: int32 = (cse_var_1 + 100)
-            let cse_var_165: int32 = (cse_var_1 + 10)
-            let cse_var_164: int32 = (cse_var_1 + 1)
-            let cse_var_163: int32 = (cse_var_1 + 112)
-            let cse_var_162: int32 = (cse_var_1 + 155)
-            let cse_var_161: int32 = (cse_var_1 + 154)
-            let cse_var_160: int32 = (cse_var_1 + 153)
-            let cse_var_159: int32 = (cse_var_1 + 152)
-            let cse_var_158: int32 = (cse_var_1 + 151)
-            let cse_var_157: int32 = (cse_var_1 + 150)
-            let cse_var_156: int32 = (cse_var_1 + 15)
-            let cse_var_155: int32 = (cse_var_1 + 149)
-            let cse_var_154: int32 = (cse_var_1 + 148)
-            let cse_var_153: int32 = (cse_var_1 + 147)
-            let cse_var_152: int32 = (cse_var_1 + 146)
-            let cse_var_151: int32 = (cse_var_1 + 145)
-            let cse_var_150: int32 = (cse_var_1 + 144)
-            let cse_var_149: int32 = (cse_var_1 + 143)
-            let cse_var_148: int32 = (cse_var_1 + 142)
-            let cse_var_147: int32 = (cse_var_1 + 127)
-            let cse_var_146: int32 = (cse_var_1 + 140)
-            let cse_var_145: int32 = (cse_var_1 + 14)
-            let cse_var_144: int32 = (cse_var_1 + 139)
-            let cse_var_143: int32 = (cse_var_1 + 138)
-            let cse_var_142: int32 = (cse_var_1 + 137)
-            let cse_var_141: int32 = (cse_var_1 + 136)
-            let cse_var_140: int32 = (cse_var_1 + 135)
-            let cse_var_139: int32 = (cse_var_1 + 134)
-            let cse_var_138: int32 = (cse_var_1 + 133)
-            let cse_var_137: int32 = (cse_var_1 + 132)
-            let cse_var_136: int32 = (cse_var_1 + 131)
-            let cse_var_135: int32 = (cse_var_1 + 130)
-            let cse_var_134: int32 = (cse_var_1 + 13)
-            let cse_var_133: int32 = (cse_var_1 + 129)
-            let cse_var_132: int32 = (cse_var_1 + 128)
-            let cse_var_131: int32 = (cse_var_1 + 141)
-            let cse_var_130: int32 = (cse_var_1 + 70)
-            let cse_var_129: int32 = (cse_var_1 + 7)
-            let cse_var_128: int32 = (cse_var_1 + 69)
-            let cse_var_127: int32 = (cse_var_1 + 68)
-            let cse_var_126: int32 = (cse_var_1 + 67)
-            let cse_var_125: int32 = (cse_var_1 + 66)
-            let cse_var_124: int32 = (cse_var_1 + 65)
-            let cse_var_123: int32 = (cse_var_1 + 64)
-            let cse_var_122: int32 = (cse_var_1 + 63)
-            let cse_var_121: int32 = (cse_var_1 + 62)
-            let cse_var_120: int32 = (cse_var_1 + 61)
-            let cse_var_119: int32 = (cse_var_1 + 60)
-            let cse_var_118: int32 = (cse_var_1 + 6)
-            let cse_var_117: int32 = (cse_var_1 + 59)
-            let cse_var_116: int32 = (cse_var_1 + 58)
-            let cse_var_115: int32 = (cse_var_1 + 213)
-            let cse_var_114: int32 = (cse_var_1 + 56)
-            let cse_var_113: int32 = (cse_var_1 + 55)
-            let cse_var_112: int32 = (cse_var_1 + 54)
-            let cse_var_111: int32 = (cse_var_1 + 53)
-            let cse_var_110: int32 = (cse_var_1 + 52)
-            let cse_var_109: int32 = (cse_var_1 + 51)
-            let cse_var_108: int32 = (cse_var_1 + 50)
-            let cse_var_107: int32 = (cse_var_1 + 5)
-            let cse_var_106: int32 = (cse_var_1 + 49)
-            let cse_var_105: int32 = (cse_var_1 + 48)
-            let cse_var_104: int32 = (cse_var_1 + 47)
-            let cse_var_103: int32 = (cse_var_1 + 46)
-            let cse_var_102: int32 = (cse_var_1 + 45)
-            let cse_var_101: int32 = (cse_var_1 + 44)
-            let cse_var_100: int32 = (cse_var_1 + 43)
-            let cse_var_99: int32 = (cse_var_1 + 57)
-            let cse_var_98: int32 = (elem_idx*16)
-            let cse_var_97: int32 = (cse_var_1 + 99)
-            let cse_var_96: int32 = (cse_var_1 + 98)
-            let cse_var_95: int32 = (cse_var_1 + 97)
-            let cse_var_94: int32 = (cse_var_1 + 96)
-            let cse_var_93: int32 = (cse_var_1 + 95)
-            let cse_var_92: int32 = (cse_var_1 + 94)
-            let cse_var_91: int32 = (cse_var_1 + 93)
-            let cse_var_90: int32 = (cse_var_1 + 92)
-            let cse_var_89: int32 = (cse_var_1 + 91)
-            let cse_var_88: int32 = (cse_var_1 + 90)
-            let cse_var_87: int32 = (cse_var_1 + 9)
-            let cse_var_86: int32 = (cse_var_1 + 89)
-            let cse_var_85: int32 = (cse_var_1 + 88)
-            let cse_var_84: int32 = (cse_var_1 + 87)
-            let cse_var_83: int32 = (cse_var_1 + 71)
-            let cse_var_82: int32 = (cse_var_1 + 85)
-            let cse_var_81: int32 = (cse_var_1 + 84)
-            let cse_var_80: int32 = (cse_var_1 + 83)
-            let cse_var_79: int32 = (cse_var_1 + 82)
-            let cse_var_78: int32 = (cse_var_1 + 81)
-            let cse_var_77: int32 = (cse_var_1 + 80)
-            let cse_var_76: int32 = (cse_var_1 + 8)
-            let cse_var_75: int32 = (cse_var_1 + 79)
-            let cse_var_74: int32 = (cse_var_1 + 78)
-            let cse_var_73: int32 = (cse_var_1 + 77)
-            let cse_var_72: int32 = (cse_var_1 + 76)
-            let cse_var_71: int32 = (cse_var_1 + 75)
-            let cse_var_70: int32 = (cse_var_1 + 74)
-            let cse_var_69: int32 = (cse_var_1 + 73)
-            let cse_var_68: int32 = (cse_var_1 + 72)
-            let cse_var_67: int32 = (cse_var_1 + 86)
-            let cse_var_66: int32 = (cse_var_1 + 242)
-            let cse_var_65: int32 = (cse_var_1 + 241)
-            let cse_var_64: int32 = (cse_var_1 + 240)
-            let cse_var_63: int32 = (cse_var_1 + 24)
-            let cse_var_62: int32 = (cse_var_1 + 239)
-            let cse_var_61: int32 = (cse_var_1 + 238)
-            let cse_var_60: int32 = (cse_var_1 + 237)
-            let cse_var_59: int32 = (cse_var_1 + 236)
-            let cse_var_58: int32 = (cse_var_1 + 235)
-            let cse_var_57: int32 = (cse_var_1 + 234)
-            let cse_var_56: int32 = (cse_var_1 + 233)
-            let cse_var_55: int32 = (cse_var_1 + 232)
-            let cse_var_54: int32 = (cse_var_1 + 231)
-            let cse_var_53: int32 = (cse_var_1 + 230)
-            let cse_var_52: int32 = (cse_var_1 + 23)
-            let cse_var_51: int32 = (cse_var_1 + 243)
-            let cse_var_50: int32 = (cse_var_1 + 228)
-            let cse_var_49: int32 = (cse_var_1 + 227)
-            let cse_var_48: int32 = (cse_var_1 + 226)
-            let cse_var_47: int32 = (cse_var_1 + 225)
-            let cse_var_46: int32 = (cse_var_1 + 224)
-            let cse_var_45: int32 = (cse_var_1 + 223)
-            let cse_var_44: int32 = (cse_var_1 + 222)
-            let cse_var_43: int32 = (cse_var_1 + 221)
-            let cse_var_42: int32 = (cse_var_1 + 220)
-            let cse_var_41: int32 = (cse_var_1 + 22)
-            let cse_var_40: int32 = (cse_var_1 + 219)
-            let cse_var_39: int32 = (cse_var_1 + 218)
-            let cse_var_38: int32 = (cse_var_1 + 217)
-            let cse_var_37: int32 = (cse_var_1 + 216)
-            let cse_var_36: int32 = (cse_var_1 + 215)
-            let cse_var_35: int32 = (cse_var_1 + 229)
-            let cse_var_34: int32 = (cse_var_1 + 42)
-            let cse_var_33: int32 = (cse_var_1 + 40)
-            let cse_var_32: int32 = (cse_var_1 + 4)
-            let cse_var_31: int32 = (cse_var_1 + 39)
-            let cse_var_30: int32 = (cse_var_1 + 38)
-            let cse_var_29: int32 = (cse_var_1 + 37)
-            let cse_var_28: int32 = (cse_var_1 + 36)
-            let cse_var_27: int32 = (cse_var_1 + 35)
-            let cse_var_26: int32 = (cse_var_1 + 34)
-            let cse_var_25: int32 = (cse_var_1 + 33)
-            let cse_var_24: int32 = (cse_var_1 + 32)
-            let cse_var_23: int32 = (cse_var_1 + 31)
-            let cse_var_22: int32 = (cse_var_1 + 30)
-            let cse_var_21: int32 = (cse_var_1 + 3)
-            let cse_var_20: int32 = (cse_var_1 + 29)
-            let cse_var_19: int32 = (cse_var_1 + 28)
-            let cse_var_18: int32 = (cse_var_1 + 41)
-            let cse_var_17: int32 = (cse_var_1 + 245)
-            let cse_var_16: int32 = (cse_var_1 + 246)
-            let cse_var_15: int32 = (cse_var_1 + 247)
-            let cse_var_14: int32 = (cse_var_1 + 248)
-            let cse_var_13: int32 = (cse_var_1 + 249)
-            let cse_var_12: int32 = (cse_var_1 + 25)
-            let cse_var_11: int32 = (cse_var_1 + 250)
-            let cse_var_10: int32 = (cse_var_1 + 251)
-            let cse_var_9: int32 = (cse_var_1 + 252)
-            let cse_var_8: int32 = (cse_var_1 + 253)
-            let cse_var_7: int32 = (cse_var_1 + 254)
-            let cse_var_6: int32 = (cse_var_1 + 255)
-            let cse_var_5: int32 = (cse_var_1 + 26)
-            let cse_var_4: int32 = (cse_var_1 + 244)
-            let cse_var_3: int32 = (cse_var_1 + 27)
-            let cse_var_2: int32 = ((i0.outer*16384) + (i.outer.inner*4096))
+          compute_4: Buffer(compute_3, float32, [256], [])[cse_var_2] = 0f32
+          compute_4[(cse_var_2 + 1)] = 0f32
+          compute_4[(cse_var_2 + 2)] = 0f32
+          compute_4[(cse_var_2 + 3)] = 0f32
+          compute_4[(cse_var_2 + 4)] = 0f32
+          compute_4[(cse_var_2 + 5)] = 0f32
+          compute_4[(cse_var_2 + 6)] = 0f32
+          compute_4[(cse_var_2 + 7)] = 0f32
+          compute_4[(cse_var_2 + 8)] = 0f32
+          compute_4[(cse_var_2 + 9)] = 0f32
+          compute_4[(cse_var_2 + 10)] = 0f32
+          compute_4[(cse_var_2 + 11)] = 0f32
+          compute_4[(cse_var_2 + 12)] = 0f32
+          compute_4[(cse_var_2 + 13)] = 0f32
+          compute_4[(cse_var_2 + 14)] = 0f32
+          compute_4[(cse_var_2 + 15)] = 0f32
+          compute_4[(cse_var_2 + 32)] = 0f32
+          compute_4[(cse_var_2 + 33)] = 0f32
+          compute_4[(cse_var_2 + 34)] = 0f32
+          compute_4[(cse_var_2 + 35)] = 0f32
+          compute_4[(cse_var_2 + 36)] = 0f32
+          compute_4[(cse_var_2 + 37)] = 0f32
+          compute_4[(cse_var_2 + 38)] = 0f32
+          compute_4[(cse_var_2 + 39)] = 0f32
+          compute_4[(cse_var_2 + 40)] = 0f32
+          compute_4[(cse_var_2 + 41)] = 0f32
+          compute_4[(cse_var_2 + 42)] = 0f32
+          compute_4[(cse_var_2 + 43)] = 0f32
+          compute_4[(cse_var_2 + 44)] = 0f32
+          compute_4[(cse_var_2 + 45)] = 0f32
+          compute_4[(cse_var_2 + 46)] = 0f32
+          compute_4[(cse_var_2 + 47)] = 0f32
+          compute_4[(cse_var_2 + 64)] = 0f32
+          compute_4[(cse_var_2 + 65)] = 0f32
+          compute_4[(cse_var_2 + 66)] = 0f32
+          compute_4[(cse_var_2 + 67)] = 0f32
+          compute_4[(cse_var_2 + 68)] = 0f32
+          compute_4[(cse_var_2 + 69)] = 0f32
+          compute_4[(cse_var_2 + 70)] = 0f32
+          compute_4[(cse_var_2 + 71)] = 0f32
+          compute_4[(cse_var_2 + 72)] = 0f32
+          compute_4[(cse_var_2 + 73)] = 0f32
+          compute_4[(cse_var_2 + 74)] = 0f32
+          compute_4[(cse_var_2 + 75)] = 0f32
+          compute_4[(cse_var_2 + 76)] = 0f32
+          compute_4[(cse_var_2 + 77)] = 0f32
+          compute_4[(cse_var_2 + 78)] = 0f32
+          compute_4[(cse_var_2 + 79)] = 0f32
+          compute_4[(cse_var_2 + 96)] = 0f32
+          compute_4[(cse_var_2 + 97)] = 0f32
+          compute_4[(cse_var_2 + 98)] = 0f32
+          compute_4[(cse_var_2 + 99)] = 0f32
+          compute_4[(cse_var_2 + 100)] = 0f32
+          compute_4[(cse_var_2 + 101)] = 0f32
+          compute_4[(cse_var_2 + 102)] = 0f32
+          compute_4[(cse_var_2 + 103)] = 0f32
+          compute_4[(cse_var_2 + 104)] = 0f32
+          compute_4[(cse_var_2 + 105)] = 0f32
+          compute_4[(cse_var_2 + 106)] = 0f32
+          compute_4[(cse_var_2 + 107)] = 0f32
+          compute_4[(cse_var_2 + 108)] = 0f32
+          compute_4[(cse_var_2 + 109)] = 0f32
+          compute_4[(cse_var_2 + 110)] = 0f32
+          compute_4[(cse_var_2 + 111)] = 0f32
+          compute_4[(cse_var_2 + 128)] = 0f32
+          compute_4[(cse_var_2 + 129)] = 0f32
+          compute_4[(cse_var_2 + 130)] = 0f32
+          compute_4[(cse_var_2 + 131)] = 0f32
+          compute_4[(cse_var_2 + 132)] = 0f32
+          compute_4[(cse_var_2 + 133)] = 0f32
+          compute_4[(cse_var_2 + 134)] = 0f32
+          compute_4[(cse_var_2 + 135)] = 0f32
+          compute_4[(cse_var_2 + 136)] = 0f32
+          compute_4[(cse_var_2 + 137)] = 0f32
+          compute_4[(cse_var_2 + 138)] = 0f32
+          compute_4[(cse_var_2 + 139)] = 0f32
+          compute_4[(cse_var_2 + 140)] = 0f32
+          compute_4[(cse_var_2 + 141)] = 0f32
+          compute_4[(cse_var_2 + 142)] = 0f32
+          compute_4[(cse_var_2 + 143)] = 0f32
+          compute_4[(cse_var_2 + 160)] = 0f32
+          compute_4[(cse_var_2 + 161)] = 0f32
+          compute_4[(cse_var_2 + 162)] = 0f32
+          compute_4[(cse_var_2 + 163)] = 0f32
+          compute_4[(cse_var_2 + 164)] = 0f32
+          compute_4[(cse_var_2 + 165)] = 0f32
+          compute_4[(cse_var_2 + 166)] = 0f32
+          compute_4[(cse_var_2 + 167)] = 0f32
+          compute_4[(cse_var_2 + 168)] = 0f32
+          compute_4[(cse_var_2 + 169)] = 0f32
+          compute_4[(cse_var_2 + 170)] = 0f32
+          compute_4[(cse_var_2 + 171)] = 0f32
+          compute_4[(cse_var_2 + 172)] = 0f32
+          compute_4[(cse_var_2 + 173)] = 0f32
+          compute_4[(cse_var_2 + 174)] = 0f32
+          compute_4[(cse_var_2 + 175)] = 0f32
+          compute_4[(cse_var_2 + 192)] = 0f32
+          compute_4[(cse_var_2 + 193)] = 0f32
+          compute_4[(cse_var_2 + 194)] = 0f32
+          compute_4[(cse_var_2 + 195)] = 0f32
+          compute_4[(cse_var_2 + 196)] = 0f32
+          compute_4[(cse_var_2 + 197)] = 0f32
+          compute_4[(cse_var_2 + 198)] = 0f32
+          compute_4[(cse_var_2 + 199)] = 0f32
+          compute_4[(cse_var_2 + 200)] = 0f32
+          compute_4[(cse_var_2 + 201)] = 0f32
+          compute_4[(cse_var_2 + 202)] = 0f32
+          compute_4[(cse_var_2 + 203)] = 0f32
+          compute_4[(cse_var_2 + 204)] = 0f32
+          compute_4[(cse_var_2 + 205)] = 0f32
+          compute_4[(cse_var_2 + 206)] = 0f32
+          compute_4[(cse_var_2 + 207)] = 0f32
+          compute_4[(cse_var_2 + 224)] = 0f32
+          compute_4[(cse_var_2 + 225)] = 0f32
+          compute_4[(cse_var_2 + 226)] = 0f32
+          compute_4[(cse_var_2 + 227)] = 0f32
+          compute_4[(cse_var_2 + 228)] = 0f32
+          compute_4[(cse_var_2 + 229)] = 0f32
+          compute_4[(cse_var_2 + 230)] = 0f32
+          compute_4[(cse_var_2 + 231)] = 0f32
+          compute_4[(cse_var_2 + 232)] = 0f32
+          compute_4[(cse_var_2 + 233)] = 0f32
+          compute_4[(cse_var_2 + 234)] = 0f32
+          compute_4[(cse_var_2 + 235)] = 0f32
+          compute_4[(cse_var_2 + 236)] = 0f32
+          compute_4[(cse_var_2 + 237)] = 0f32
+          compute_4[(cse_var_2 + 238)] = 0f32
+          compute_4[(cse_var_2 + 239)] = 0f32
+          for (elem_idx: int32, 0, (placeholder_3[(cse_var_1 + 1)] - placeholder_3[cse_var_1])) {
+            let cse_var_131: int32 = (cse_var_2 + 143)
+            let cse_var_130: int32 = (cse_var_2 + 15)
+            let cse_var_129: int32 = (cse_var_2 + 160)
+            let cse_var_128: int32 = (cse_var_2 + 161)
+            let cse_var_127: int32 = (cse_var_2 + 162)
+            let cse_var_126: int32 = (cse_var_2 + 163)
+            let cse_var_125: int32 = (cse_var_2 + 164)
+            let cse_var_124: int32 = (cse_var_2 + 165)
+            let cse_var_123: int32 = (cse_var_2 + 166)
+            let cse_var_122: int32 = (cse_var_2 + 167)
+            let cse_var_121: int32 = (cse_var_2 + 168)
+            let cse_var_120: int32 = (cse_var_2 + 169)
+            let cse_var_119: int32 = (cse_var_2 + 170)
+            let cse_var_118: int32 = (cse_var_2 + 171)
+            let cse_var_117: int32 = (cse_var_2 + 172)
+            let cse_var_116: int32 = (cse_var_2 + 1)
+            let cse_var_115: int32 = (cse_var_2 + 174)
+            let cse_var_114: int32 = (cse_var_2 + 175)
+            let cse_var_113: int32 = (cse_var_2 + 192)
+            let cse_var_112: int32 = (cse_var_2 + 193)
+            let cse_var_111: int32 = (cse_var_2 + 194)
+            let cse_var_110: int32 = (cse_var_2 + 195)
+            let cse_var_109: int32 = (cse_var_2 + 196)
+            let cse_var_108: int32 = (cse_var_2 + 197)
+            let cse_var_107: int32 = (cse_var_2 + 198)
+            let cse_var_106: int32 = (cse_var_2 + 199)
+            let cse_var_105: int32 = (cse_var_2 + 2)
+            let cse_var_104: int32 = (cse_var_2 + 200)
+            let cse_var_103: int32 = (cse_var_2 + 201)
+            let cse_var_102: int32 = (cse_var_2 + 202)
+            let cse_var_101: int32 = (cse_var_2 + 203)
+            let cse_var_100: int32 = (cse_var_2 + 173)
+            let cse_var_99: int32 = (cse_var_2 + 10)
+            let cse_var_98: int32 = (cse_var_2 + 100)
+            let cse_var_97: int32 = (cse_var_2 + 101)
+            let cse_var_96: int32 = (cse_var_2 + 102)
+            let cse_var_95: int32 = (cse_var_2 + 103)
+            let cse_var_94: int32 = (cse_var_2 + 104)
+            let cse_var_93: int32 = (cse_var_2 + 105)
+            let cse_var_92: int32 = (cse_var_2 + 106)
+            let cse_var_91: int32 = (cse_var_2 + 107)
+            let cse_var_90: int32 = (cse_var_2 + 108)
+            let cse_var_89: int32 = (cse_var_2 + 109)
+            let cse_var_88: int32 = (cse_var_2 + 11)
+            let cse_var_87: int32 = (cse_var_2 + 110)
+            let cse_var_86: int32 = (cse_var_2 + 111)
+            let cse_var_85: int32 = (cse_var_2 + 12)
+            let cse_var_84: int32 = (cse_var_2 + 142)
+            let cse_var_83: int32 = (cse_var_2 + 129)
+            let cse_var_82: int32 = (cse_var_2 + 13)
+            let cse_var_81: int32 = (cse_var_2 + 130)
+            let cse_var_80: int32 = (cse_var_2 + 131)
+            let cse_var_79: int32 = (cse_var_2 + 132)
+            let cse_var_78: int32 = (cse_var_2 + 133)
+            let cse_var_77: int32 = (cse_var_2 + 134)
+            let cse_var_76: int32 = (cse_var_2 + 135)
+            let cse_var_75: int32 = (cse_var_2 + 136)
+            let cse_var_74: int32 = (cse_var_2 + 137)
+            let cse_var_73: int32 = (cse_var_2 + 138)
+            let cse_var_72: int32 = (cse_var_2 + 139)
+            let cse_var_71: int32 = (cse_var_2 + 14)
+            let cse_var_70: int32 = (cse_var_2 + 140)
+            let cse_var_69: int32 = (cse_var_2 + 141)
+            let cse_var_68: int32 = (cse_var_2 + 128)
+            let cse_var_67: int32 = (cse_var_2 + 44)
+            let cse_var_66: int32 = (cse_var_2 + 45)
+            let cse_var_65: int32 = (cse_var_2 + 46)
+            let cse_var_64: int32 = (cse_var_2 + 47)
+            let cse_var_63: int32 = (cse_var_2 + 5)
+            let cse_var_62: int32 = (cse_var_2 + 6)
+            let cse_var_61: int32 = (cse_var_2 + 64)
+            let cse_var_60: int32 = (cse_var_2 + 65)
+            let cse_var_59: int32 = (cse_var_2 + 66)
+            let cse_var_58: int32 = (cse_var_2 + 67)
+            let cse_var_57: int32 = (cse_var_2 + 68)
+            let cse_var_56: int32 = (cse_var_2 + 69)
+            let cse_var_55: int32 = (cse_var_2 + 7)
+            let cse_var_54: int32 = (cse_var_2 + 70)
+            let cse_var_53: int32 = (cse_var_2 + 71)
+            let cse_var_52: int32 = (cse_var_2 + 204)
+            let cse_var_51: int32 = (cse_var_2 + 73)
+            let cse_var_50: int32 = (cse_var_2 + 74)
+            let cse_var_49: int32 = (cse_var_2 + 75)
+            let cse_var_48: int32 = (cse_var_2 + 76)
+            let cse_var_47: int32 = (cse_var_2 + 77)
+            let cse_var_46: int32 = (cse_var_2 + 78)
+            let cse_var_45: int32 = (cse_var_2 + 79)
+            let cse_var_44: int32 = (cse_var_2 + 8)
+            let cse_var_43: int32 = (cse_var_2 + 9)
+            let cse_var_42: int32 = (cse_var_2 + 96)
+            let cse_var_41: int32 = (cse_var_2 + 97)
+            let cse_var_40: int32 = (cse_var_2 + 98)
+            let cse_var_39: int32 = (cse_var_2 + 99)
+            let cse_var_38: int32 = (elem_idx*16)
+            let cse_var_37: int32 = (i0.outer*2048)
+            let cse_var_36: int32 = (cse_var_2 + 72)
+            let cse_var_35: int32 = (cse_var_2 + 205)
+            let cse_var_34: int32 = (cse_var_2 + 206)
+            let cse_var_33: int32 = (cse_var_2 + 207)
+            let cse_var_32: int32 = (cse_var_2 + 224)
+            let cse_var_31: int32 = (cse_var_2 + 225)
+            let cse_var_30: int32 = (cse_var_2 + 226)
+            let cse_var_29: int32 = (cse_var_2 + 227)
+            let cse_var_28: int32 = (cse_var_2 + 228)
+            let cse_var_27: int32 = (cse_var_2 + 229)
+            let cse_var_26: int32 = (cse_var_2 + 230)
+            let cse_var_25: int32 = (cse_var_2 + 231)
+            let cse_var_24: int32 = (cse_var_2 + 232)
+            let cse_var_23: int32 = (cse_var_2 + 233)
+            let cse_var_22: int32 = (cse_var_2 + 234)
+            let cse_var_21: int32 = (cse_var_2 + 235)
+            let cse_var_20: int32 = (cse_var_2 + 43)
+            let cse_var_19: int32 = (cse_var_2 + 42)
+            let cse_var_18: int32 = (cse_var_2 + 41)
+            let cse_var_17: int32 = (cse_var_2 + 40)
+            let cse_var_16: int32 = (cse_var_2 + 4)
+            let cse_var_15: int32 = (cse_var_2 + 39)
+            let cse_var_14: int32 = (cse_var_2 + 38)
+            let cse_var_13: int32 = (cse_var_2 + 37)
+            let cse_var_12: int32 = (cse_var_2 + 236)
+            let cse_var_11: int32 = (cse_var_2 + 35)
+            let cse_var_10: int32 = (cse_var_2 + 34)
+            let cse_var_9: int32 = (cse_var_2 + 33)
+            let cse_var_8: int32 = (cse_var_2 + 32)
+            let cse_var_7: int32 = (cse_var_2 + 3)
+            let cse_var_6: int32 = (cse_var_2 + 239)
+            let cse_var_5: int32 = (cse_var_2 + 238)
+            let cse_var_4: int32 = (cse_var_2 + 237)
+            let cse_var_3: int32 = (cse_var_2 + 36)
              {
-              compute_4[cse_var_1] = (compute_4[cse_var_1] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_164] = (compute_4[cse_var_164] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_195] = (compute_4[cse_var_195] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_21] = (compute_4[cse_var_21] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_32] = (compute_4[cse_var_32] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_107] = (compute_4[cse_var_107] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_118] = (compute_4[cse_var_118] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_129] = (compute_4[cse_var_129] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_76] = (compute_4[cse_var_76] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_87] = (compute_4[cse_var_87] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_165] = (compute_4[cse_var_165] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_176] = (compute_4[cse_var_176] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_187] = (compute_4[cse_var_187] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_134] = (compute_4[cse_var_134] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_145] = (compute_4[cse_var_145] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_156] = (compute_4[cse_var_156] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[(cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)])], 0f32)))
-              compute_4[cse_var_231] = (compute_4[cse_var_231] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_242] = (compute_4[cse_var_242] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_253] = (compute_4[cse_var_253] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_200] = (compute_4[cse_var_200] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_212] = (compute_4[cse_var_212] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_223] = (compute_4[cse_var_223] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_41] = (compute_4[cse_var_41] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_52] = (compute_4[cse_var_52] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_63] = (compute_4[cse_var_63] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_22] = (compute_4[cse_var_22] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_23] = (compute_4[cse_var_23] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 256)], 0f32)))
-              compute_4[cse_var_24] = (compute_4[cse_var_24] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_25] = (compute_4[cse_var_25] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_26] = (compute_4[cse_var_26] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_27] = (compute_4[cse_var_27] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_28] = (compute_4[cse_var_28] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_29] = (compute_4[cse_var_29] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_30] = (compute_4[cse_var_30] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_31] = (compute_4[cse_var_31] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_33] = (compute_4[cse_var_33] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_18] = (compute_4[cse_var_18] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_34] = (compute_4[cse_var_34] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_100] = (compute_4[cse_var_100] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_101] = (compute_4[cse_var_101] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_102] = (compute_4[cse_var_102] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_103] = (compute_4[cse_var_103] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_104] = (compute_4[cse_var_104] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 512)], 0f32)))
-              compute_4[cse_var_105] = (compute_4[cse_var_105] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_106] = (compute_4[cse_var_106] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_108] = (compute_4[cse_var_108] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_109] = (compute_4[cse_var_109] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_110] = (compute_4[cse_var_110] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_111] = (compute_4[cse_var_111] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_112] = (compute_4[cse_var_112] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_113] = (compute_4[cse_var_113] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_114] = (compute_4[cse_var_114] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_99] = (compute_4[cse_var_99] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_116] = (compute_4[cse_var_116] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_117] = (compute_4[cse_var_117] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_119] = (compute_4[cse_var_119] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_120] = (compute_4[cse_var_120] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_121] = (compute_4[cse_var_121] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_122] = (compute_4[cse_var_122] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 768)], 0f32)))
-              compute_4[cse_var_123] = (compute_4[cse_var_123] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_124] = (compute_4[cse_var_124] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_125] = (compute_4[cse_var_125] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_126] = (compute_4[cse_var_126] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_127] = (compute_4[cse_var_127] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_128] = (compute_4[cse_var_128] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_130] = (compute_4[cse_var_130] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_83] = (compute_4[cse_var_83] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_68] = (compute_4[cse_var_68] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_69] = (compute_4[cse_var_69] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_70] = (compute_4[cse_var_70] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_71] = (compute_4[cse_var_71] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_72] = (compute_4[cse_var_72] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_73] = (compute_4[cse_var_73] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_74] = (compute_4[cse_var_74] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_75] = (compute_4[cse_var_75] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1024)], 0f32)))
-              compute_4[cse_var_77] = (compute_4[cse_var_77] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_78] = (compute_4[cse_var_78] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_79] = (compute_4[cse_var_79] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_80] = (compute_4[cse_var_80] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_81] = (compute_4[cse_var_81] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_82] = (compute_4[cse_var_82] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_67] = (compute_4[cse_var_67] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_84] = (compute_4[cse_var_84] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_85] = (compute_4[cse_var_85] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_86] = (compute_4[cse_var_86] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_88] = (compute_4[cse_var_88] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_89] = (compute_4[cse_var_89] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_90] = (compute_4[cse_var_90] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_91] = (compute_4[cse_var_91] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_92] = (compute_4[cse_var_92] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_93] = (compute_4[cse_var_93] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1280)], 0f32)))
-              compute_4[cse_var_94] = (compute_4[cse_var_94] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_95] = (compute_4[cse_var_95] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_96] = (compute_4[cse_var_96] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_97] = (compute_4[cse_var_97] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_166] = (compute_4[cse_var_166] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_167] = (compute_4[cse_var_167] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_168] = (compute_4[cse_var_168] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_169] = (compute_4[cse_var_169] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_170] = (compute_4[cse_var_170] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_171] = (compute_4[cse_var_171] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_172] = (compute_4[cse_var_172] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_173] = (compute_4[cse_var_173] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_174] = (compute_4[cse_var_174] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_175] = (compute_4[cse_var_175] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_177] = (compute_4[cse_var_177] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_178] = (compute_4[cse_var_178] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1536)], 0f32)))
-              compute_4[cse_var_163] = (compute_4[cse_var_163] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_180] = (compute_4[cse_var_180] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_181] = (compute_4[cse_var_181] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_182] = (compute_4[cse_var_182] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_183] = (compute_4[cse_var_183] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_184] = (compute_4[cse_var_184] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_185] = (compute_4[cse_var_185] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_186] = (compute_4[cse_var_186] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_188] = (compute_4[cse_var_188] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_189] = (compute_4[cse_var_189] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_190] = (compute_4[cse_var_190] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_191] = (compute_4[cse_var_191] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_192] = (compute_4[cse_var_192] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_193] = (compute_4[cse_var_193] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_194] = (compute_4[cse_var_194] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_147] = (compute_4[cse_var_147] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 1792)], 0f32)))
-              compute_4[cse_var_132] = (compute_4[cse_var_132] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_133] = (compute_4[cse_var_133] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_135] = (compute_4[cse_var_135] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_136] = (compute_4[cse_var_136] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_137] = (compute_4[cse_var_137] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_138] = (compute_4[cse_var_138] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_139] = (compute_4[cse_var_139] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_140] = (compute_4[cse_var_140] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_141] = (compute_4[cse_var_141] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_142] = (compute_4[cse_var_142] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_143] = (compute_4[cse_var_143] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_144] = (compute_4[cse_var_144] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_146] = (compute_4[cse_var_146] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_131] = (compute_4[cse_var_131] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_148] = (compute_4[cse_var_148] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_149] = (compute_4[cse_var_149] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2048)], 0f32)))
-              compute_4[cse_var_150] = (compute_4[cse_var_150] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_151] = (compute_4[cse_var_151] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_152] = (compute_4[cse_var_152] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_153] = (compute_4[cse_var_153] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_154] = (compute_4[cse_var_154] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_155] = (compute_4[cse_var_155] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_157] = (compute_4[cse_var_157] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_158] = (compute_4[cse_var_158] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_159] = (compute_4[cse_var_159] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_160] = (compute_4[cse_var_160] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_161] = (compute_4[cse_var_161] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_162] = (compute_4[cse_var_162] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_179] = (compute_4[cse_var_179] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_228] = (compute_4[cse_var_228] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_229] = (compute_4[cse_var_229] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_230] = (compute_4[cse_var_230] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2304)], 0f32)))
-              compute_4[cse_var_232] = (compute_4[cse_var_232] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_233] = (compute_4[cse_var_233] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_234] = (compute_4[cse_var_234] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_235] = (compute_4[cse_var_235] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_236] = (compute_4[cse_var_236] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_237] = (compute_4[cse_var_237] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_238] = (compute_4[cse_var_238] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_239] = (compute_4[cse_var_239] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_240] = (compute_4[cse_var_240] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_241] = (compute_4[cse_var_241] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_227] = (compute_4[cse_var_227] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_244] = (compute_4[cse_var_244] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_245] = (compute_4[cse_var_245] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_246] = (compute_4[cse_var_246] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_247] = (compute_4[cse_var_247] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_248] = (compute_4[cse_var_248] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2560)], 0f32)))
-              compute_4[cse_var_249] = (compute_4[cse_var_249] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_250] = (compute_4[cse_var_250] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_251] = (compute_4[cse_var_251] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_252] = (compute_4[cse_var_252] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_254] = (compute_4[cse_var_254] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_255] = (compute_4[cse_var_255] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_256] = (compute_4[cse_var_256] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_257] = (compute_4[cse_var_257] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_258] = (compute_4[cse_var_258] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_211] = (compute_4[cse_var_211] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_196] = (compute_4[cse_var_196] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_197] = (compute_4[cse_var_197] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_198] = (compute_4[cse_var_198] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_199] = (compute_4[cse_var_199] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_201] = (compute_4[cse_var_201] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_202] = (compute_4[cse_var_202] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 2816)], 0f32)))
-              compute_4[cse_var_203] = (compute_4[cse_var_203] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_204] = (compute_4[cse_var_204] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_205] = (compute_4[cse_var_205] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_206] = (compute_4[cse_var_206] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_207] = (compute_4[cse_var_207] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_208] = (compute_4[cse_var_208] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_209] = (compute_4[cse_var_209] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_210] = (compute_4[cse_var_210] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_213] = (compute_4[cse_var_213] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_214] = (compute_4[cse_var_214] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_215] = (compute_4[cse_var_215] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_216] = (compute_4[cse_var_216] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_217] = (compute_4[cse_var_217] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_218] = (compute_4[cse_var_218] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_219] = (compute_4[cse_var_219] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_220] = (compute_4[cse_var_220] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3072)], 0f32)))
-              compute_4[cse_var_221] = (compute_4[cse_var_221] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_222] = (compute_4[cse_var_222] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_224] = (compute_4[cse_var_224] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_225] = (compute_4[cse_var_225] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_226] = (compute_4[cse_var_226] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_115] = (compute_4[cse_var_115] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_243] = (compute_4[cse_var_243] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_36] = (compute_4[cse_var_36] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_37] = (compute_4[cse_var_37] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_38] = (compute_4[cse_var_38] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_39] = (compute_4[cse_var_39] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_40] = (compute_4[cse_var_40] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_42] = (compute_4[cse_var_42] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_43] = (compute_4[cse_var_43] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_44] = (compute_4[cse_var_44] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_45] = (compute_4[cse_var_45] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3328)], 0f32)))
-              compute_4[cse_var_46] = (compute_4[cse_var_46] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_47] = (compute_4[cse_var_47] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_48] = (compute_4[cse_var_48] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_49] = (compute_4[cse_var_49] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_50] = (compute_4[cse_var_50] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_35] = (compute_4[cse_var_35] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_53] = (compute_4[cse_var_53] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_54] = (compute_4[cse_var_54] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_55] = (compute_4[cse_var_55] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_56] = (compute_4[cse_var_56] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_57] = (compute_4[cse_var_57] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_58] = (compute_4[cse_var_58] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_59] = (compute_4[cse_var_59] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_60] = (compute_4[cse_var_60] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_61] = (compute_4[cse_var_61] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_62] = (compute_4[cse_var_62] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3584)], 0f32)))
-              compute_4[cse_var_64] = (compute_4[cse_var_64] + (placeholder_1[((placeholder_3[i1.outer]*16) + cse_var_98)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_65] = (compute_4[cse_var_65] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 1)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_66] = (compute_4[cse_var_66] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 2)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_51] = (compute_4[cse_var_51] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 3)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 4)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 5)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 6)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 7)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 8)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 9)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 10)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 11)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 12)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 13)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 14)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
-              compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_1[(((placeholder_3[i1.outer]*16) + cse_var_98) + 15)]*max(placeholder[((cse_var_2 + placeholder_2[(placeholder_3[i1.outer] + elem_idx)]) + 3840)], 0f32)))
+              compute_4[cse_var_2] = (compute_4[cse_var_2] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_116] = (compute_4[cse_var_116] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_105] = (compute_4[cse_var_105] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_63] = (compute_4[cse_var_63] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_62] = (compute_4[cse_var_62] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_55] = (compute_4[cse_var_55] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_44] = (compute_4[cse_var_44] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_43] = (compute_4[cse_var_43] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_99] = (compute_4[cse_var_99] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_88] = (compute_4[cse_var_88] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_85] = (compute_4[cse_var_85] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_82] = (compute_4[cse_var_82] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_71] = (compute_4[cse_var_71] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_130] = (compute_4[cse_var_130] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[(cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)])], 0f32)))
+              compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_18] = (compute_4[cse_var_18] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_67] = (compute_4[cse_var_67] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_66] = (compute_4[cse_var_66] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_65] = (compute_4[cse_var_65] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_64] = (compute_4[cse_var_64] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 256)], 0f32)))
+              compute_4[cse_var_61] = (compute_4[cse_var_61] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_60] = (compute_4[cse_var_60] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_59] = (compute_4[cse_var_59] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_58] = (compute_4[cse_var_58] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_57] = (compute_4[cse_var_57] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_56] = (compute_4[cse_var_56] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_54] = (compute_4[cse_var_54] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_53] = (compute_4[cse_var_53] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_36] = (compute_4[cse_var_36] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_51] = (compute_4[cse_var_51] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_50] = (compute_4[cse_var_50] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_49] = (compute_4[cse_var_49] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_48] = (compute_4[cse_var_48] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_47] = (compute_4[cse_var_47] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_46] = (compute_4[cse_var_46] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_45] = (compute_4[cse_var_45] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 512)], 0f32)))
+              compute_4[cse_var_42] = (compute_4[cse_var_42] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_41] = (compute_4[cse_var_41] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_40] = (compute_4[cse_var_40] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_39] = (compute_4[cse_var_39] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_98] = (compute_4[cse_var_98] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_97] = (compute_4[cse_var_97] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_96] = (compute_4[cse_var_96] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_95] = (compute_4[cse_var_95] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_94] = (compute_4[cse_var_94] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_93] = (compute_4[cse_var_93] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_92] = (compute_4[cse_var_92] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_91] = (compute_4[cse_var_91] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_90] = (compute_4[cse_var_90] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_89] = (compute_4[cse_var_89] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_87] = (compute_4[cse_var_87] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_86] = (compute_4[cse_var_86] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 768)], 0f32)))
+              compute_4[cse_var_68] = (compute_4[cse_var_68] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_83] = (compute_4[cse_var_83] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_81] = (compute_4[cse_var_81] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_80] = (compute_4[cse_var_80] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_79] = (compute_4[cse_var_79] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_78] = (compute_4[cse_var_78] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_77] = (compute_4[cse_var_77] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_76] = (compute_4[cse_var_76] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_75] = (compute_4[cse_var_75] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_74] = (compute_4[cse_var_74] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_73] = (compute_4[cse_var_73] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_72] = (compute_4[cse_var_72] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_70] = (compute_4[cse_var_70] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_69] = (compute_4[cse_var_69] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_84] = (compute_4[cse_var_84] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_131] = (compute_4[cse_var_131] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1024)], 0f32)))
+              compute_4[cse_var_129] = (compute_4[cse_var_129] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_128] = (compute_4[cse_var_128] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_127] = (compute_4[cse_var_127] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_126] = (compute_4[cse_var_126] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_125] = (compute_4[cse_var_125] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_124] = (compute_4[cse_var_124] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_123] = (compute_4[cse_var_123] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_122] = (compute_4[cse_var_122] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_121] = (compute_4[cse_var_121] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_120] = (compute_4[cse_var_120] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_119] = (compute_4[cse_var_119] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_118] = (compute_4[cse_var_118] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_117] = (compute_4[cse_var_117] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_100] = (compute_4[cse_var_100] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_115] = (compute_4[cse_var_115] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_114] = (compute_4[cse_var_114] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1280)], 0f32)))
+              compute_4[cse_var_113] = (compute_4[cse_var_113] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_112] = (compute_4[cse_var_112] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_111] = (compute_4[cse_var_111] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_110] = (compute_4[cse_var_110] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_109] = (compute_4[cse_var_109] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_108] = (compute_4[cse_var_108] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_107] = (compute_4[cse_var_107] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_106] = (compute_4[cse_var_106] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_104] = (compute_4[cse_var_104] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_103] = (compute_4[cse_var_103] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_102] = (compute_4[cse_var_102] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_101] = (compute_4[cse_var_101] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_52] = (compute_4[cse_var_52] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_35] = (compute_4[cse_var_35] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_34] = (compute_4[cse_var_34] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_33] = (compute_4[cse_var_33] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1536)], 0f32)))
+              compute_4[cse_var_32] = (compute_4[cse_var_32] + (placeholder_1[((placeholder_3[cse_var_1]*16) + cse_var_38)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_31] = (compute_4[cse_var_31] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 1)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_30] = (compute_4[cse_var_30] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 2)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_29] = (compute_4[cse_var_29] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 3)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_28] = (compute_4[cse_var_28] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 4)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_27] = (compute_4[cse_var_27] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 5)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_26] = (compute_4[cse_var_26] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 6)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_25] = (compute_4[cse_var_25] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 7)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_24] = (compute_4[cse_var_24] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 8)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_23] = (compute_4[cse_var_23] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 9)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_22] = (compute_4[cse_var_22] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 10)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_21] = (compute_4[cse_var_21] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 11)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 12)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 13)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 14)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
+              compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_1]*16) + cse_var_38) + 15)]*max(placeholder[((cse_var_37 + placeholder_2[(placeholder_3[cse_var_1] + elem_idx)]) + 1792)], 0f32)))
             }
           }
         }
       }
-      for (i0.inner: int32, 0, 64) {
-        let cse_var_259: int32 = (((i0.outer*32768) + (i0.inner*512)) + (i1.outer*16))
-        compute[ramp(cse_var_259, 1, 16)] = max((compute_4[ramp((i0.inner*16), 1, 16)] + placeholder_4[ramp(cse_var_259, 1, 16)]), broadcast(0f32, 16))
+      for (i0.inner: int32, 0, 8) {
+        let cse_var_132: int32 = (((i0.outer*4096) + (i0.inner*512)) + (i1.outer*32))
+        compute[ramp(cse_var_132, 1, 32)] = max((compute_4[ramp((i0.inner*32), 1, 32)] + placeholder_4[ramp(cse_var_132, 1, 32)]), broadcast(0f32, 32))
       }
     }
   }
@@ -1422,7 +1039,7 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
 </pre></div>
 </div>
 <p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 3.337 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 2.706 ms
 </pre></div>
 </div>
 <div class="admonition note">
diff --git a/docs/how_to/tune_with_autotvm/sg_execution_times.html b/docs/how_to/tune_with_autotvm/sg_execution_times.html
index c658425fd..135de22e1 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -300,13 +300,13 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-tune-with-autotvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:43.418</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:43.955</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:42.622</strong>: <a class="reference internal" href="tune_conv2d_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-conv2d-cuda-py"><span class="std std-ref">Tuning High Performance Convolution on NVIDIA GPUs</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_cuda.py</span></code>)</p></li>
-<li><p><strong>00:00.209</strong>: <a class="reference internal" href="tune_relay_x86.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-x86-py"><span class="std std-ref">Auto-tuning a Convolutional Network for x86 CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_x86.py</span></code>)</p></li>
-<li><p><strong>00:00.202</strong>: <a class="reference internal" href="tune_relay_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-cuda-py"><span class="std std-ref">Auto-tuning a Convolutional Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_cuda.py</span></code>)</p></li>
-<li><p><strong>00:00.192</strong>: <a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></li>
-<li><p><strong>00:00.192</strong>: <a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></li>
+<li><p><strong>00:43.085</strong>: <a class="reference internal" href="tune_conv2d_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-conv2d-cuda-py"><span class="std std-ref">Tuning High Performance Convolution on NVIDIA GPUs</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_cuda.py</span></code>)</p></li>
+<li><p><strong>00:00.228</strong>: <a class="reference internal" href="tune_relay_x86.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-x86-py"><span class="std std-ref">Auto-tuning a Convolutional Network for x86 CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_x86.py</span></code>)</p></li>
+<li><p><strong>00:00.216</strong>: <a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></li>
+<li><p><strong>00:00.213</strong>: <a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></li>
+<li><p><strong>00:00.213</strong>: <a class="reference internal" href="tune_relay_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-cuda-py"><span class="std std-ref">Auto-tuning a Convolutional Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_cuda.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
index fa79b4712..560080f20 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -1142,8 +1142,8 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 4, 4, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 1, 128]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2885496
-No: 6   GFLOPS: 103.50/103.50   result: MeasureResult(costs=(0.002236654708333333,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.592435359954834, timestamp=1650063815.6699162)        [(&#39;tile_f&#39;, [-1, 1, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 4, 4]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,3754080
-No: 7   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 6   GFLOPS: 103.76/103.76   result: MeasureResult(costs=(0.0022312065416666667,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5933806896209717, timestamp=1650065132.1649663)      [(&#39;tile_f&#39;, [-1, 1, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 4, 4]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,3754080
+No: 7   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1266,7 +1266,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 16, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 256, 1]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,6225319
-No: 8   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 8   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1389,7 +1389,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 1, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 8, 64]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,943546
-No: 9   GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 9   GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1512,7 +1512,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 4, 16, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 16, 32]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2868708
-No: 10  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 10  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 142, in build
     res = future.result()
   File &quot;/usr/lib/python3.7/concurrent/futures/_base.py&quot;, line 435, in result
@@ -1530,7 +1530,7 @@ No: 10  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
 TimeoutError
 
         [(&#39;tile_f&#39;, [-1, 32, 2, 4]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 4, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4691833
-No: 11  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 11  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1653,7 +1653,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 2, 64]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 4]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,1042124
-No: 12  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 12  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1776,7 +1776,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 32, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 32, 16]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,10013405
-No: 13  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 13  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -1899,7 +1899,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 8, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 4, 32]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,6732082
-No: 14  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 14  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -2022,7 +2022,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 4, 32]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 128]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7536735
-No: 15  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 15  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -2145,7 +2145,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 128, 4]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,482121
-No: 16  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 16  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -2268,7 +2268,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 1, 16]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 32, 8]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2824525
-No: 17  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 17  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -2391,7 +2391,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 64, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 8, 8]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4559286
-No: 18  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 18  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 571, in __call__
     func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 523, in _build_func_common
@@ -2514,7 +2514,7 @@ Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 854, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
 tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 32, 16]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 512]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9677544
-No: 19  GFLOPS: 0.00/103.50     result: Traceback (most recent call last):
+No: 19  GFLOPS: 0.00/103.76     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 721, in __call__
     yield remote, remote.load_module(os.path.split(build_result.filename)[1])
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 685, in run_through_rpc
@@ -2602,7 +2602,7 @@ tvm._ffi.base.TVMError: Traceback (most recent call last):
   15: _PyEval_EvalFrameDefault
   14: 0x0000000000537c30
   13: _PyObject_FastCallKeywords
-  12: 0x00007fc4d562cfa2
+  12: 0x00007f0bfab37fa2
   11: _ctypes_callproc
   10: ffi_call
   9: ffi_call_unix64
@@ -2667,7 +2667,7 @@ Traceback (most recent call last):
   21: _PyFunction_FastCallKeywords
   20: _PyEval_EvalFrameDefault
   19: _PyFunction_FastCall      [(&#39;tile_f&#39;, [-1, 8, 2, 16]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,6390073
-No: 20  GFLOPS: 144.98/144.98   result: MeasureResult(costs=(0.0015968331,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4312255382537842, timestamp=1650063841.3437583)       [(&#39;tile_f&#39;, [-1, 1, 4, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9881539
+No: 20  GFLOPS: 144.55/144.55   result: MeasureResult(costs=(0.0016015775600000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4185152053833008, timestamp=1650065158.492134)       [(&#39;tile_f&#39;, [-1, 1, 4, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9881539
 </pre></div>
 </div>
 <p>Finally we can inspect the best config from log file, check correctness,
@@ -2706,7 +2706,7 @@ and measure running time.</p>
 <p class="sphx-glr-script-out">Out:</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Best config:
 [(&#39;tile_f&#39;, [-1, 1, 4, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9881539
-Time cost of this operator: 0.001952
+Time cost of this operator: 0.002011
 </pre></div>
 </div>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autotvm-tune-conv2d-cuda-py">
diff --git a/docs/how_to/work_with_microtvm/micro_autotune.html b/docs/how_to/work_with_microtvm/micro_autotune.html
index 8ad31a996..8b853d9da 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -553,10 +553,10 @@ the tuned operator.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build without Autotuning ##########
 Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs
 ---------                                     ---                                           --------  -------  -----              ------  -------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  309.7     98.733   (1, 2, 10, 10, 3)  2       1
-tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.067     0.978    (1, 6, 10, 10)     1       1
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.906     0.289    (1, 1, 10, 10, 3)  1       1
-Total_time                                    -                                             313.673   -        -                  -       -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  313.0     98.744   (1, 2, 10, 10, 3)  2       1
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.073     0.969    (1, 6, 10, 10)     1       1
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.907     0.286    (1, 1, 10, 10, 3)  1       1
+Total_time                                    -                                             316.98    -        -                  -       -
 </pre></div>
 </div>
 </div>
@@ -608,10 +608,10 @@ Total_time                                    -
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build with Autotuning ##########
 Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs
 ---------                                     ---                                           --------  -------  -----              ------  -------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  89.1      97.148   (1, 6, 10, 10, 1)  2       1
-tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.715     1.87     (1, 6, 10, 10)     1       1
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.901     0.982    (1, 1, 10, 10, 3)  1       1
-Total_time                                    -                                             91.716    -        -                  -       -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  133.7     98.082   (1, 6, 10, 10, 1)  2       1
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.702     1.248    (1, 6, 10, 10)     1       1
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.913     0.67     (1, 1, 10, 10, 3)  1       1
+Total_time                                    -                                             136.314   -        -                  -       -
 </pre></div>
 </div>
 <div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-autotune-py">
diff --git a/docs/how_to/work_with_microtvm/sg_execution_times.html b/docs/how_to/work_with_microtvm/sg_execution_times.html
index 12d6ceddc..271d54648 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -300,12 +300,12 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-work-with-microtvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:43.326</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>00:43.330</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:39.410</strong>: <a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">Autotuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></li>
-<li><p><strong>00:03.367</strong>: <a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">microTVM with TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></li>
-<li><p><strong>00:00.188</strong>: <a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></li>
-<li><p><strong>00:00.186</strong>: <a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">Executing a Tiny Model with TVMC Micro</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></li>
+<li><p><strong>00:39.365</strong>: <a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">Autotuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></li>
+<li><p><strong>00:03.415</strong>: <a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">microTVM with TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></li>
+<li><p><strong>00:00.190</strong>: <a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">Executing a Tiny Model with TVMC Micro</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></li>
+<li><p><strong>00:00.184</strong>: <a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></li>
 <li><p><strong>00:00.176</strong>: <a class="reference internal" href="micro_reference_vm.html#sphx-glr-how-to-work-with-microtvm-micro-reference-vm-py"><span class="std std-ref">microTVM Reference Virtual Machines</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_reference_vm.py</span></code>)</p></li>
 </ul>
 </div>
diff --git a/docs/how_to/work_with_relay/sg_execution_times.html b/docs/how_to/work_with_relay/sg_execution_times.html
index f39a927b7..a98edb802 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -300,11 +300,11 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-work-with-relay-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:05.817</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:09.148</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:04.046</strong>: <a class="reference internal" href="using_external_lib.html#sphx-glr-how-to-work-with-relay-using-external-lib-py"><span class="std std-ref">Using External Libraries in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_external_lib.py</span></code>)</p></li>
-<li><p><strong>00:01.566</strong>: <a class="reference internal" href="build_gcn.html#sphx-glr-how-to-work-with-relay-build-gcn-py"><span class="std std-ref">Building a Graph Convolutional Network</span></a> (<code class="docutils literal notranslate"><span class="pre">build_gcn.py</span></code>)</p></li>
-<li><p><strong>00:00.205</strong>: <a class="reference internal" href="using_relay_viz.html#sphx-glr-how-to-work-with-relay-using-relay-viz-py"><span class="std std-ref">Use Relay Visualizer to Visualize Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_relay_viz.py</span></code>)</p></li>
+<li><p><strong>00:07.078</strong>: <a class="reference internal" href="using_external_lib.html#sphx-glr-how-to-work-with-relay-using-external-lib-py"><span class="std std-ref">Using External Libraries in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_external_lib.py</span></code>)</p></li>
+<li><p><strong>00:01.882</strong>: <a class="reference internal" href="build_gcn.html#sphx-glr-how-to-work-with-relay-build-gcn-py"><span class="std std-ref">Building a Graph Convolutional Network</span></a> (<code class="docutils literal notranslate"><span class="pre">build_gcn.py</span></code>)</p></li>
+<li><p><strong>00:00.188</strong>: <a class="reference internal" href="using_relay_viz.html#sphx-glr-how-to-work-with-relay-using-relay-viz-py"><span class="std std-ref">Use Relay Visualizer to Visualize Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_relay_viz.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/work_with_schedules/sg_execution_times.html b/docs/how_to/work_with_schedules/sg_execution_times.html
index 2ea39a854..7293a8100 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -300,16 +300,16 @@
             
   <div class="section" id="computation-times">
 <span id="sphx-glr-how-to-work-with-schedules-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:05.195</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:05.384</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
 <ul class="simple">
-<li><p><strong>00:01.939</strong>: <a class="reference internal" href="intrin_math.html#sphx-glr-how-to-work-with-schedules-intrin-math-py"><span class="std std-ref">Intrinsics and Math Functions</span></a> (<code class="docutils literal notranslate"><span class="pre">intrin_math.py</span></code>)</p></li>
-<li><p><strong>00:00.986</strong>: <a class="reference internal" href="tensorize.html#sphx-glr-how-to-work-with-schedules-tensorize-py"><span class="std std-ref">Use Tensorize to Leverage Hardware Intrinsics</span></a> (<code class="docutils literal notranslate"><span class="pre">tensorize.py</span></code>)</p></li>
-<li><p><strong>00:00.687</strong>: <a class="reference internal" href="reduction.html#sphx-glr-how-to-work-with-schedules-reduction-py"><span class="std std-ref">Reduction</span></a> (<code class="docutils literal notranslate"><span class="pre">reduction.py</span></code>)</p></li>
-<li><p><strong>00:00.656</strong>: <a class="reference internal" href="scan.html#sphx-glr-how-to-work-with-schedules-scan-py"><span class="std std-ref">Scan and Recurrent Kernel</span></a> (<code class="docutils literal notranslate"><span class="pre">scan.py</span></code>)</p></li>
-<li><p><strong>00:00.289</strong>: <a class="reference internal" href="extern_op.html#sphx-glr-how-to-work-with-schedules-extern-op-py"><span class="std std-ref">External Tensor Functions</span></a> (<code class="docutils literal notranslate"><span class="pre">extern_op.py</span></code>)</p></li>
-<li><p><strong>00:00.227</strong>: <a class="reference internal" href="schedule_primitives.html#sphx-glr-how-to-work-with-schedules-schedule-primitives-py"><span class="std std-ref">Schedule Primitives in TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">schedule_primitives.py</span></code>)</p></li>
-<li><p><strong>00:00.213</strong>: <a class="reference internal" href="tedd.html#sphx-glr-how-to-work-with-schedules-tedd-py"><span class="std std-ref">Use Tensor Expression Debug Display (TEDD) for Visualization</span></a> (<code class="docutils literal notranslate"><span class="pre">tedd.py</span></code>)</p></li>
-<li><p><strong>00:00.198</strong>: <a class="reference internal" href="tuple_inputs.html#sphx-glr-how-to-work-with-schedules-tuple-inputs-py"><span class="std std-ref">Compute and Reduce with Tuple Inputs</span></a> (<code class="docutils literal notranslate"><span class="pre">tuple_inputs.py</span></code>)</p></li>
+<li><p><strong>00:02.025</strong>: <a class="reference internal" href="intrin_math.html#sphx-glr-how-to-work-with-schedules-intrin-math-py"><span class="std std-ref">Intrinsics and Math Functions</span></a> (<code class="docutils literal notranslate"><span class="pre">intrin_math.py</span></code>)</p></li>
+<li><p><strong>00:01.059</strong>: <a class="reference internal" href="tensorize.html#sphx-glr-how-to-work-with-schedules-tensorize-py"><span class="std std-ref">Use Tensorize to Leverage Hardware Intrinsics</span></a> (<code class="docutils literal notranslate"><span class="pre">tensorize.py</span></code>)</p></li>
+<li><p><strong>00:00.701</strong>: <a class="reference internal" href="reduction.html#sphx-glr-how-to-work-with-schedules-reduction-py"><span class="std std-ref">Reduction</span></a> (<code class="docutils literal notranslate"><span class="pre">reduction.py</span></code>)</p></li>
+<li><p><strong>00:00.690</strong>: <a class="reference internal" href="scan.html#sphx-glr-how-to-work-with-schedules-scan-py"><span class="std std-ref">Scan and Recurrent Kernel</span></a> (<code class="docutils literal notranslate"><span class="pre">scan.py</span></code>)</p></li>
+<li><p><strong>00:00.277</strong>: <a class="reference internal" href="extern_op.html#sphx-glr-how-to-work-with-schedules-extern-op-py"><span class="std std-ref">External Tensor Functions</span></a> (<code class="docutils literal notranslate"><span class="pre">extern_op.py</span></code>)</p></li>
+<li><p><strong>00:00.231</strong>: <a class="reference internal" href="schedule_primitives.html#sphx-glr-how-to-work-with-schedules-schedule-primitives-py"><span class="std std-ref">Schedule Primitives in TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">schedule_primitives.py</span></code>)</p></li>
+<li><p><strong>00:00.207</strong>: <a class="reference internal" href="tedd.html#sphx-glr-how-to-work-with-schedules-tedd-py"><span class="std std-ref">Use Tensor Expression Debug Display (TEDD) for Visualization</span></a> (<code class="docutils literal notranslate"><span class="pre">tedd.py</span></code>)</p></li>
+<li><p><strong>00:00.194</strong>: <a class="reference internal" href="tuple_inputs.html#sphx-glr-how-to-work-with-schedules-tuple-inputs-py"><span class="std std-ref">Compute and Reduce with Tuple Inputs</span></a> (<code class="docutils literal notranslate"><span class="pre">tuple_inputs.py</span></code>)</p></li>
 </ul>
 </div>
 
diff --git a/docs/how_to/work_with_schedules/tensorize.html b/docs/how_to/work_with_schedules/tensorize.html
index c2783957d..0fb69b4be 100644
--- a/docs/how_to/work_with_schedules/tensorize.html
+++ b/docs/how_to/work_with_schedules/tensorize.html
@@ -548,7 +548,7 @@ The importing needs to happen before the tensorized GEMV being executed.</p>
              B: Buffer(B_2: Pointer(float32), float32, [32768], []),
              C: Buffer(C_2: Pointer(float32), float32, [524288], [])}
   buffer_map = {A_1: A, B_1: B, C_1: C} {
-  attr [IterVar(i: int32, (nullptr), &quot;DataPar&quot;, &quot;&quot;)] &quot;pragma_import_llvm&quot; = &quot;; ModuleID = &#39;/tmp/tmphcc2ycl2/input0.cc&#39;\nsource_filename = \&quot;/tmp/tmphcc2ycl2/input0.cc\&quot;\ntarget datalayout = \&quot;e-m:e-i64:64-f80:128-n8:16:32:64-S128\&quot;\ntarget triple = \&quot;x86_64-pc-linux-gnu\&quot;\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n  %7 = allo [...]
+  attr [IterVar(i: int32, (nullptr), &quot;DataPar&quot;, &quot;&quot;)] &quot;pragma_import_llvm&quot; = &quot;; ModuleID = &#39;/tmp/tmp5asqara1/input0.cc&#39;\nsource_filename = \&quot;/tmp/tmp5asqara1/input0.cc\&quot;\ntarget datalayout = \&quot;e-m:e-i64:64-f80:128-n8:16:32:64-S128\&quot;\ntarget triple = \&quot;x86_64-pc-linux-gnu\&quot;\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n  %7 = allo [...]
   for (i, 0, 1024) {
     for (j.outer: int32, 0, 32) {
       @tir.call_extern(&quot;gemv_update&quot;, @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), C_2, ((i*512) + (j.outer*16)), 16, 2, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), A_2, (i*64), 64, 1, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), B_2, (j.outer*1024), 1024, 1, dtype=handle), 16, 64, 64, dtype=int32)
diff --git a/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode.html b/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode.html
index 1a8809f0f..d2127f5a6 100644
--- a/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode.html
+++ b/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode.html
@@ -85,7 +85,7 @@ Inheritance diagram for tvm::LinkedParamNode:</div>
 <div class="dynheader">
 Collaboration diagram for tvm::LinkedParamNode:</div>
 <div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1LinkedParamNode__coll__graph.svg" width="554" height="1346"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1LinkedParamNode__coll__graph.svg" width="563" height="1346"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
 </div>
 </div>
 <table class="memberdecls">
diff --git a/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode__coll__graph.svg
index a73bd888d..9881f3cca 100644
--- a/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1LinkedParamNode__coll__graph.svg
@@ -4,22 +4,22 @@
 <!-- Generated by graphviz version 2.40.1 (20161225.0304)
  -->
 <!-- Title: tvm::LinkedParamNode Pages: 1 -->
-<svg width="415pt" height="1009pt"
- viewBox="0.00 0.00 414.50 1009.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<svg width="422pt" height="1009pt"
+ viewBox="0.00 0.00 422.00 1009.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
 <g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 1005)">
 <title>tvm::LinkedParamNode</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1005 410.5,-1005 410.5,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1005 418,-1005 418,4 -4,4"/>
 <!-- Node2 -->
 <g id="node1" class="node">
 <title>Node2</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="109,-.5 109,-79.5 318,-79.5 318,-.5 109,-.5"/>
-<text text-anchor="middle" x="213.5" y="-67.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::LinkedParamNode</text>
-<polyline fill="none" stroke="#000000" points="109,-60.5 318,-60.5 "/>
-<text text-anchor="start" x="117" y="-48.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ id</text>
-<text text-anchor="start" x="117" y="-37.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
-<polyline fill="none" stroke="#000000" points="109,-30.5 318,-30.5 "/>
-<text text-anchor="start" x="117" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ VisitAttrs()</text>
-<text text-anchor="start" x="117" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_FINAL_OBJECT_INFO()</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="112,-.5 112,-79.5 321,-79.5 321,-.5 112,-.5"/>
+<text text-anchor="middle" x="216.5" y="-67.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::LinkedParamNode</text>
+<polyline fill="none" stroke="#000000" points="112,-60.5 321,-60.5 "/>
+<text text-anchor="start" x="120" y="-48.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ id</text>
+<text text-anchor="start" x="120" y="-37.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
+<polyline fill="none" stroke="#000000" points="112,-30.5 321,-30.5 "/>
+<text text-anchor="start" x="120" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ VisitAttrs()</text>
+<text text-anchor="start" x="120" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_FINAL_OBJECT_INFO()</text>
 </g>
 <!-- Node3 -->
 <g id="node2" class="node">
@@ -67,8 +67,8 @@
 <!-- Node3&#45;&gt;Node2 -->
 <g id="edge1" class="edge">
 <title>Node3&#45;&gt;Node2</title>
-<path fill="none" stroke="#191970" d="M179.6467,-117.9736C185.7598,-103.8934 191.416,-90.8657 196.2852,-79.6505"/>
-<polygon fill="none" stroke="#191970" points="176.3393,-116.8032 175.5672,-127.3698 182.7602,-119.5909 176.3393,-116.8032"/>
+<path fill="none" stroke="#191970" d="M181.8142,-117.9736C188.0777,-103.8934 193.8729,-90.8657 198.8619,-79.6505"/>
+<polygon fill="none" stroke="#191970" points="178.501,-116.8105 177.6344,-127.3698 184.8968,-119.6556 178.501,-116.8105"/>
 </g>
 <!-- Node3&#45;&gt;Node3 -->
 <g id="edge2" class="edge">
@@ -81,103 +81,105 @@
 <g id="node3" class="node">
 <title>Node4</title>
 <g id="a_node3"><a xlink:href="classtvm_1_1runtime_1_1NDArray.html" target="_top" xlink:title="Managed NDArray. The array is backed by reference counted blocks. ">
-<polygon fill="#ffffff" stroke="#000000" points="271,-199 271,-443 402,-443 402,-199 271,-199"/>
-<text text-anchor="middle" x="336.5" y="-431" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
-<polyline fill="none" stroke="#000000" points="271,-424 402,-424 "/>
-<text text-anchor="middle" x="336.5" y="-412" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="271,-405 402,-405 "/>
+<polygon fill="#ffffff" stroke="#000000" points="271,-188 271,-454 414,-454 414,-188 271,-188"/>
+<text text-anchor="middle" x="342.5" y="-442" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
+<polyline fill="none" stroke="#000000" points="271,-435 414,-435 "/>
+<text text-anchor="middle" x="342.5" y="-423" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="271,-416 414,-416 "/>
+<text text-anchor="start" x="279" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
 <text text-anchor="start" x="279" y="-393" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="279" y="-382" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="279" y="-371" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
-<text text-anchor="start" x="279" y="-360" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="279" y="-349" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="279" y="-338" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
+<text text-anchor="start" x="279" y="-382" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
+<text text-anchor="start" x="279" y="-371" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="279" y="-360" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="279" y="-349" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
+<text text-anchor="start" x="279" y="-338" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
 <text text-anchor="start" x="279" y="-327" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="279" y="-316" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="279" y="-305" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
-<text text-anchor="start" x="279" y="-294" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
-<text text-anchor="start" x="279" y="-283" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
-<text text-anchor="start" x="279" y="-272" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
-<text text-anchor="start" x="279" y="-261" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
-<text text-anchor="start" x="279" y="-250" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
-<text text-anchor="start" x="279" y="-239" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="279" y="-228" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
-<text text-anchor="start" x="279" y="-217" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
-<text text-anchor="start" x="279" y="-206" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
+<text text-anchor="start" x="279" y="-316" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
+<text text-anchor="start" x="279" y="-305" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
+<text text-anchor="start" x="279" y="-294" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
+<text text-anchor="start" x="279" y="-283" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
+<text text-anchor="start" x="279" y="-272" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromExternalDLTensor()</text>
+<text text-anchor="start" x="279" y="-261" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NewFromDLTensor()</text>
+<text text-anchor="start" x="279" y="-250" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
+<text text-anchor="start" x="279" y="-239" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
+<text text-anchor="start" x="279" y="-228" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="279" y="-217" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
+<text text-anchor="start" x="279" y="-206" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
+<text text-anchor="start" x="279" y="-195" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
 </a>
 </g>
 </g>
 <!-- Node4&#45;&gt;Node2 -->
 <g id="edge3" class="edge">
 <title>Node4&#45;&gt;Node2</title>
-<path fill="none" stroke="#404040" d="M293.0459,-198.7604C283.6067,-174.7004 273.2175,-149.8442 262.5,-127 256.8065,-114.8643 250.0817,-102.1221 243.4681,-90.247"/>
-<polygon fill="none" stroke="#404040" points="243.3686,-90.0711 236.9323,-86.8193 237.4585,-79.6274 243.8948,-82.8792 243.3686,-90.0711"/>
+<path fill="none" stroke="#404040" d="M289.6216,-187.9935C280.8804,-167.4292 271.6698,-146.5414 262.5,-127 256.8875,-115.0395 250.4685,-102.3666 244.2424,-90.5029"/>
+<polygon fill="none" stroke="#404040" points="244.2303,-90.4801 237.8851,-87.0538 238.6074,-79.8789 244.9525,-83.3052 244.2303,-90.4801"/>
 <text text-anchor="middle" x="274" y="-101" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +param</text>
 </g>
 <!-- Node5 -->
 <g id="node4" class="node">
 <title>Node5</title>
 <g id="a_node4"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="269.5,-552.5 269.5,-774.5 403.5,-774.5 403.5,-552.5 269.5,-552.5"/>
-<text text-anchor="middle" x="336.5" y="-762.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="269.5,-755.5 403.5,-755.5 "/>
-<text text-anchor="start" x="277.5" y="-743.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<polyline fill="none" stroke="#000000" points="269.5,-736.5 403.5,-736.5 "/>
-<text text-anchor="start" x="277.5" y="-724.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="277.5" y="-713.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="277.5" y="-702.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="277.5" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="277.5" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="277.5" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
-<text text-anchor="start" x="277.5" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="277.5" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="277.5" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="277.5" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="277.5" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="277.5" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="277.5" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="277.5" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="277.5" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="277.5" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<polygon fill="#ffffff" stroke="#000000" points="275.5,-552.5 275.5,-774.5 409.5,-774.5 409.5,-552.5 275.5,-552.5"/>
+<text text-anchor="middle" x="342.5" y="-762.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="275.5,-755.5 409.5,-755.5 "/>
+<text text-anchor="start" x="283.5" y="-743.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<polyline fill="none" stroke="#000000" points="275.5,-736.5 409.5,-736.5 "/>
+<text text-anchor="start" x="283.5" y="-724.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="283.5" y="-713.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="283.5" y="-702.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="283.5" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="283.5" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="283.5" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
+<text text-anchor="start" x="283.5" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="283.5" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="283.5" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="283.5" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="283.5" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="283.5" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="283.5" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="283.5" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="283.5" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="283.5" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
 </a>
 </g>
 </g>
 <!-- Node5&#45;&gt;Node4 -->
 <g id="edge4" class="edge">
 <title>Node5&#45;&gt;Node4</title>
-<path fill="none" stroke="#191970" d="M336.5,-541.983C336.5,-509.9111 336.5,-475.2908 336.5,-443.1838"/>
-<polygon fill="none" stroke="#191970" points="333.0001,-542.3 336.5,-552.3001 340.0001,-542.3001 333.0001,-542.3"/>
+<path fill="none" stroke="#191970" d="M342.5,-542.284C342.5,-513.7101 342.5,-483.0989 342.5,-454.0351"/>
+<polygon fill="none" stroke="#191970" points="339.0001,-542.3 342.5,-552.3001 346.0001,-542.3001 339.0001,-542.3"/>
 </g>
 <!-- Node6 -->
 <g id="node5" class="node">
 <title>Node6</title>
 <g id="a_node5"><a xlink:href="classtvm_1_1runtime_1_1ObjectPtr.html" target="_top" xlink:title="{tvm::runtime::ObjectPtr\l\&lt; tvm::runtime::Object \&gt;\n||+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ~ObjectPtr()\l+ swap()\l+ get()\l+ operator&#45;\&gt;()\land 11 more...\l}">
-<polygon fill="#ffffff" stroke="#000000" points="266.5,-822.5 266.5,-1000.5 406.5,-1000.5 406.5,-822.5 266.5,-822.5"/>
-<text text-anchor="start" x="274.5" y="-988.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
-<text text-anchor="middle" x="336.5" y="-977.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
-<polyline fill="none" stroke="#000000" points="266.5,-970.5 406.5,-970.5 "/>
-<text text-anchor="middle" x="336.5" y="-958.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="266.5,-951.5 406.5,-951.5 "/>
-<text text-anchor="start" x="274.5" y="-939.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-928.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-917.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-906.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-895.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-884.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-873.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
-<text text-anchor="start" x="274.5" y="-862.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
-<text text-anchor="start" x="274.5" y="-851.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="274.5" y="-840.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="274.5" y="-829.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
+<polygon fill="#ffffff" stroke="#000000" points="272.5,-822.5 272.5,-1000.5 412.5,-1000.5 412.5,-822.5 272.5,-822.5"/>
+<text text-anchor="start" x="280.5" y="-988.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
+<text text-anchor="middle" x="342.5" y="-977.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
+<polyline fill="none" stroke="#000000" points="272.5,-970.5 412.5,-970.5 "/>
+<text text-anchor="middle" x="342.5" y="-958.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="272.5,-951.5 412.5,-951.5 "/>
+<text text-anchor="start" x="280.5" y="-939.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-928.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-917.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-906.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-895.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-884.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-873.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
+<text text-anchor="start" x="280.5" y="-862.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
+<text text-anchor="start" x="280.5" y="-851.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="280.5" y="-840.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="280.5" y="-829.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
 </a>
 </g>
 </g>
 <!-- Node6&#45;&gt;Node5 -->
 <g id="edge5" class="edge">
 <title>Node6&#45;&gt;Node5</title>
-<path fill="none" stroke="#404040" d="M336.5,-822.3167C336.5,-810.8765 336.5,-799.0062 336.5,-787.1402"/>
-<polygon fill="none" stroke="#404040" points="336.5001,-786.7944 332.5,-780.7944 336.5,-774.7944 340.5,-780.7943 336.5001,-786.7944"/>
-<text text-anchor="middle" x="356" y="-796" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
+<path fill="none" stroke="#404040" d="M342.5,-822.3167C342.5,-810.8765 342.5,-799.0062 342.5,-787.1402"/>
+<polygon fill="none" stroke="#404040" points="342.5001,-786.7944 338.5,-780.7944 342.5,-774.7944 346.5,-780.7943 342.5001,-786.7944"/>
+<text text-anchor="middle" x="362" y="-796" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
 </g>
 </g>
 </svg>
diff --git a/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode.html b/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode.html
index 85d146a3e..996918ea0 100644
--- a/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode.html
+++ b/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode.html
@@ -85,7 +85,7 @@ Inheritance diagram for tvm::relay::ConstantNode:</div>
 <div class="dynheader">
 Collaboration diagram for tvm::relay::ConstantNode:</div>
 <div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1relay_1_1ConstantNode__coll__graph.svg" width="834" height="2178"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1relay_1_1ConstantNode__coll__graph.svg" width="836" height="2207"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
 </div>
 </div>
 <table class="memberdecls">
diff --git a/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode__coll__graph.svg
index 1738ac146..6239f8fad 100644
--- a/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1relay_1_1ConstantNode__coll__graph.svg
@@ -4,293 +4,295 @@
 <!-- Generated by graphviz version 2.40.1 (20161225.0304)
  -->
 <!-- Title: tvm::relay::ConstantNode Pages: 1 -->
-<svg width="625pt" height="1633pt"
- viewBox="0.00 0.00 625.31 1633.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 1629)">
+<svg width="627pt" height="1655pt"
+ viewBox="0.00 0.00 626.58 1655.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 1651)">
 <title>tvm::relay::ConstantNode</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1629 621.3081,-1629 621.3081,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-1651 622.5809,-1651 622.5809,4 -4,4"/>
 <!-- Node4 -->
 <g id="node1" class="node">
 <title>Node4</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="247.1435,-.5 247.1435,-112.5 456.1435,-112.5 456.1435,-.5 247.1435,-.5"/>
-<text text-anchor="middle" x="351.6435" y="-100.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::relay::ConstantNode</text>
-<polyline fill="none" stroke="#000000" points="247.1435,-93.5 456.1435,-93.5 "/>
-<text text-anchor="start" x="255.1435" y="-81.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
-<polyline fill="none" stroke="#000000" points="247.1435,-74.5 456.1435,-74.5 "/>
-<text text-anchor="start" x="255.1435" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ tensor_type()</text>
-<text text-anchor="start" x="255.1435" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ is_scalar()</text>
-<text text-anchor="start" x="255.1435" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ VisitAttrs()</text>
-<text text-anchor="start" x="255.1435" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SEqualReduce()</text>
-<text text-anchor="start" x="255.1435" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SHashReduce()</text>
-<text text-anchor="start" x="255.1435" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_FINAL_OBJECT_INFO()</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="250.2881,-.5 250.2881,-112.5 459.2881,-112.5 459.2881,-.5 250.2881,-.5"/>
+<text text-anchor="middle" x="354.7881" y="-100.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::relay::ConstantNode</text>
+<polyline fill="none" stroke="#000000" points="250.2881,-93.5 459.2881,-93.5 "/>
+<text text-anchor="start" x="258.2881" y="-81.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
+<polyline fill="none" stroke="#000000" points="250.2881,-74.5 459.2881,-74.5 "/>
+<text text-anchor="start" x="258.2881" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ tensor_type()</text>
+<text text-anchor="start" x="258.2881" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ is_scalar()</text>
+<text text-anchor="start" x="258.2881" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ VisitAttrs()</text>
+<text text-anchor="start" x="258.2881" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SEqualReduce()</text>
+<text text-anchor="start" x="258.2881" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SHashReduce()</text>
+<text text-anchor="start" x="258.2881" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_FINAL_OBJECT_INFO()</text>
 </g>
 <!-- Node5 -->
 <g id="node2" class="node">
 <title>Node5</title>
 <g id="a_node2"><a xlink:href="classtvm_1_1RelayExprNode.html" target="_top" xlink:title="Base node of all non&#45;primitive expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="141.1435,-150.5 141.1435,-251.5 348.1435,-251.5 348.1435,-150.5 141.1435,-150.5"/>
-<text text-anchor="middle" x="244.6435" y="-239.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">ExprNode</text>
-<polyline fill="none" stroke="#000000" points="141.1435,-232.5 348.1435,-232.5 "/>
-<text text-anchor="start" x="149.1435" y="-220.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
-<text text-anchor="start" x="149.1435" y="-209.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
-<polyline fill="none" stroke="#000000" points="141.1435,-202.5 348.1435,-202.5 "/>
-<text text-anchor="start" x="149.1435" y="-190.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ checked_type()</text>
-<text text-anchor="start" x="149.1435" y="-179.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ type_as()</text>
-<text text-anchor="start" x="149.1435" y="-168.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ virtual_device()</text>
-<text text-anchor="start" x="149.1435" y="-157.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_BASE_OBJECT_INFO()</text>
+<polygon fill="#ffffff" stroke="#000000" points="143.2881,-150.5 143.2881,-251.5 350.2881,-251.5 350.2881,-150.5 143.2881,-150.5"/>
+<text text-anchor="middle" x="246.7881" y="-239.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">ExprNode</text>
+<polyline fill="none" stroke="#000000" points="143.2881,-232.5 350.2881,-232.5 "/>
+<text text-anchor="start" x="151.2881" y="-220.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
+<text text-anchor="start" x="151.2881" y="-209.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
+<polyline fill="none" stroke="#000000" points="143.2881,-202.5 350.2881,-202.5 "/>
+<text text-anchor="start" x="151.2881" y="-190.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ checked_type()</text>
+<text text-anchor="start" x="151.2881" y="-179.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ type_as()</text>
+<text text-anchor="start" x="151.2881" y="-168.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ virtual_device()</text>
+<text text-anchor="start" x="151.2881" y="-157.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_BASE_OBJECT_INFO()</text>
 </a>
 </g>
 </g>
 <!-- Node5&#45;&gt;Node4 -->
 <g id="edge1" class="edge">
 <title>Node5&#45;&gt;Node4</title>
-<path fill="none" stroke="#191970" d="M288.0839,-142.3352C295.3332,-132.5453 302.838,-122.4102 310.04,-112.6842"/>
-<polygon fill="none" stroke="#191970" points="285.2051,-140.3415 282.0669,-150.4609 290.8307,-144.5072 285.2051,-140.3415"/>
+<path fill="none" stroke="#191970" d="M290.6345,-142.3352C297.9515,-132.5453 305.5265,-122.4102 312.7958,-112.6842"/>
+<polygon fill="none" stroke="#191970" points="287.7445,-140.3556 284.5613,-150.4609 293.3515,-144.5463 287.7445,-140.3556"/>
 </g>
 <!-- Node6 -->
 <g id="node3" class="node">
 <title>Node6</title>
 <g id="a_node3"><a xlink:href="classtvm_1_1BaseExprNode.html" target="_top" xlink:title="Base type of all the expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="274.1435,-533.5 274.1435,-645.5 481.1435,-645.5 481.1435,-533.5 274.1435,-533.5"/>
-<text text-anchor="middle" x="377.6435" y="-633.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExprNode</text>
-<polyline fill="none" stroke="#000000" points="274.1435,-626.5 481.1435,-626.5 "/>
-<text text-anchor="start" x="282.1435" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
-<text text-anchor="start" x="282.1435" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_sequal</text>
-<text text-anchor="start" x="282.1435" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
-<text text-anchor="start" x="282.1435" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_shash</text>
-<text text-anchor="start" x="282.1435" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
-<text text-anchor="start" x="282.1435" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
-<polyline fill="none" stroke="#000000" points="274.1435,-552.5 481.1435,-552.5 "/>
-<text text-anchor="start" x="282.1435" y="-540.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_BASE_OBJECT_INFO()</text>
+<polygon fill="#ffffff" stroke="#000000" points="273.2881,-555.5 273.2881,-667.5 480.2881,-667.5 480.2881,-555.5 273.2881,-555.5"/>
+<text text-anchor="middle" x="376.7881" y="-655.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExprNode</text>
+<polyline fill="none" stroke="#000000" points="273.2881,-648.5 480.2881,-648.5 "/>
+<text text-anchor="start" x="281.2881" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
+<text text-anchor="start" x="281.2881" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_sequal</text>
+<text text-anchor="start" x="281.2881" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
+<text text-anchor="start" x="281.2881" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_shash</text>
+<text text-anchor="start" x="281.2881" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
+<text text-anchor="start" x="281.2881" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
+<polyline fill="none" stroke="#000000" points="273.2881,-574.5 480.2881,-574.5 "/>
+<text text-anchor="start" x="281.2881" y="-562.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DECLARE_BASE_OBJECT_INFO()</text>
 </a>
 </g>
 </g>
 <!-- Node6&#45;&gt;Node5 -->
 <g id="edge2" class="edge">
 <title>Node6&#45;&gt;Node5</title>
-<path fill="none" stroke="#191970" d="M355.167,-523.8448C328.875,-447.0445 285.9131,-321.5506 262.0345,-251.8001"/>
-<polygon fill="none" stroke="#191970" points="351.8959,-525.0963 358.4462,-533.4237 358.5186,-522.8291 351.8959,-525.0963"/>
+<path fill="none" stroke="#191970" d="M355.9132,-545.5835C329.9842,-463.7078 286.3099,-325.7977 262.846,-251.706"/>
+<polygon fill="none" stroke="#191970" points="352.6289,-546.8058 358.9847,-555.2825 359.3023,-544.6924 352.6289,-546.8058"/>
 </g>
 <!-- Node7 -->
 <g id="node4" class="node">
 <title>Node7</title>
 <g id="a_node4"><a xlink:href="classtvm_1_1runtime_1_1Object.html" target="_top" xlink:title="base class of all object containers. ">
-<polygon fill="#ffffff" stroke="#000000" points="286.1435,-751.5 286.1435,-1138.5 469.1435,-1138.5 469.1435,-751.5 286.1435,-751.5"/>
-<text text-anchor="middle" x="377.6435" y="-1126.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::Object</text>
-<polyline fill="none" stroke="#000000" points="286.1435,-1119.5 469.1435,-1119.5 "/>
-<text text-anchor="start" x="294.1435" y="-1107.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
-<text text-anchor="start" x="294.1435" y="-1096.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_final</text>
-<text text-anchor="start" x="294.1435" y="-1085.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
-<text text-anchor="start" x="294.1435" y="-1074.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots_can</text>
-<text text-anchor="start" x="294.1435" y="-1063.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_overflow</text>
-<text text-anchor="start" x="294.1435" y="-1052.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_visit</text>
-<text text-anchor="start" x="294.1435" y="-1041.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_attrs</text>
-<text text-anchor="start" x="294.1435" y="-1030.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_sequal</text>
-<text text-anchor="start" x="294.1435" y="-1019.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
-<text text-anchor="start" x="294.1435" y="-1008.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_shash</text>
-<text text-anchor="start" x="294.1435" y="-997.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
-<text text-anchor="start" x="294.1435" y="-986.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_index</text>
-<text text-anchor="start" x="294.1435" y="-975.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># type_index_</text>
-<text text-anchor="start" x="294.1435" y="-964.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># ref_counter_</text>
-<polyline fill="none" stroke="#000000" points="286.1435,-957.5 469.1435,-957.5 "/>
-<text text-anchor="start" x="294.1435" y="-945.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ type_index()</text>
-<text text-anchor="start" x="294.1435" y="-934.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ GetTypeKey()</text>
-<text text-anchor="start" x="294.1435" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ GetTypeKeyHash()</text>
-<text text-anchor="start" x="294.1435" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsInstance()</text>
-<text text-anchor="start" x="294.1435" y="-901.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="294.1435" y="-890.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
-<text text-anchor="start" x="294.1435" y="-879.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
-<text text-anchor="start" x="294.1435" y="-868.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
-<text text-anchor="start" x="294.1435" y="-857.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator=()</text>
-<text text-anchor="start" x="294.1435" y="-846.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator=()</text>
-<text text-anchor="start" x="294.1435" y="-835.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeIndex2Key()</text>
-<text text-anchor="start" x="294.1435" y="-824.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeIndex2KeyHash()</text>
-<text text-anchor="start" x="294.1435" y="-813.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeKey2Index()</text>
-<text text-anchor="start" x="294.1435" y="-802.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _GetOrAllocRuntimeTypeIndex()</text>
-<text text-anchor="start" x="294.1435" y="-791.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ RuntimeTypeIndex()</text>
-<text text-anchor="start" x="294.1435" y="-780.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># IncRef()</text>
-<text text-anchor="start" x="294.1435" y="-769.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DecRef()</text>
-<text text-anchor="start" x="294.1435" y="-758.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetOrAllocRuntimeTypeIndex()</text>
+<polygon fill="#ffffff" stroke="#000000" points="285.2881,-773.5 285.2881,-1160.5 468.2881,-1160.5 468.2881,-773.5 285.2881,-773.5"/>
+<text text-anchor="middle" x="376.7881" y="-1148.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::Object</text>
+<polyline fill="none" stroke="#000000" points="285.2881,-1141.5 468.2881,-1141.5 "/>
+<text text-anchor="start" x="293.2881" y="-1129.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_key</text>
+<text text-anchor="start" x="293.2881" y="-1118.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_final</text>
+<text text-anchor="start" x="293.2881" y="-1107.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots</text>
+<text text-anchor="start" x="293.2881" y="-1096.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_child_slots_can</text>
+<text text-anchor="start" x="293.2881" y="-1085.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_overflow</text>
+<text text-anchor="start" x="293.2881" y="-1074.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_visit</text>
+<text text-anchor="start" x="293.2881" y="-1063.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_attrs</text>
+<text text-anchor="start" x="293.2881" y="-1052.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_sequal</text>
+<text text-anchor="start" x="293.2881" y="-1041.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
+<text text-anchor="start" x="293.2881" y="-1030.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_has_method_shash</text>
+<text text-anchor="start" x="293.2881" y="-1019.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_reduce</text>
+<text text-anchor="start" x="293.2881" y="-1008.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_index</text>
+<text text-anchor="start" x="293.2881" y="-997.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># type_index_</text>
+<text text-anchor="start" x="293.2881" y="-986.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># ref_counter_</text>
+<polyline fill="none" stroke="#000000" points="285.2881,-979.5 468.2881,-979.5 "/>
+<text text-anchor="start" x="293.2881" y="-967.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ type_index()</text>
+<text text-anchor="start" x="293.2881" y="-956.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ GetTypeKey()</text>
+<text text-anchor="start" x="293.2881" y="-945.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ GetTypeKeyHash()</text>
+<text text-anchor="start" x="293.2881" y="-934.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsInstance()</text>
+<text text-anchor="start" x="293.2881" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="293.2881" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
+<text text-anchor="start" x="293.2881" y="-901.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
+<text text-anchor="start" x="293.2881" y="-890.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Object()</text>
+<text text-anchor="start" x="293.2881" y="-879.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator=()</text>
+<text text-anchor="start" x="293.2881" y="-868.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator=()</text>
+<text text-anchor="start" x="293.2881" y="-857.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeIndex2Key()</text>
+<text text-anchor="start" x="293.2881" y="-846.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeIndex2KeyHash()</text>
+<text text-anchor="start" x="293.2881" y="-835.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TypeKey2Index()</text>
+<text text-anchor="start" x="293.2881" y="-824.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _GetOrAllocRuntimeTypeIndex()</text>
+<text text-anchor="start" x="293.2881" y="-813.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ RuntimeTypeIndex()</text>
+<text text-anchor="start" x="293.2881" y="-802.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># IncRef()</text>
+<text text-anchor="start" x="293.2881" y="-791.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DecRef()</text>
+<text text-anchor="start" x="293.2881" y="-780.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetOrAllocRuntimeTypeIndex()</text>
 </a>
 </g>
 </g>
 <!-- Node7&#45;&gt;Node6 -->
 <g id="edge3" class="edge">
 <title>Node7&#45;&gt;Node6</title>
-<path fill="none" stroke="#191970" d="M377.6435,-740.9811C377.6435,-705.9652 377.6435,-672.3401 377.6435,-645.7036"/>
-<polygon fill="none" stroke="#191970" points="374.1436,-741.3858 377.6435,-751.3859 381.1436,-741.3859 374.1436,-741.3858"/>
+<path fill="none" stroke="#191970" d="M376.7881,-762.9811C376.7881,-727.9652 376.7881,-694.3401 376.7881,-667.7036"/>
+<polygon fill="none" stroke="#191970" points="373.2882,-763.3858 376.7881,-773.3859 380.2882,-763.3859 373.2882,-763.3858"/>
 </g>
 <!-- Node7&#45;&gt;Node7 -->
 <g id="edge4" class="edge">
 <title>Node7&#45;&gt;Node7</title>
-<path fill="none" stroke="#404040" d="M469.506,-978.9248C480.1917,-972.6637 487.1435,-961.3555 487.1435,-945 487.1435,-934.0112 484.0053,-925.3007 478.705,-918.8687"/>
-<polygon fill="none" stroke="#404040" points="478.6619,-918.8322 471.4983,-918.0056 469.506,-911.0752 476.6696,-911.9017 478.6619,-918.8322"/>
-<text text-anchor="middle" x="513.1435" y="-942.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #deleter_</text>
+<path fill="none" stroke="#404040" d="M468.6506,-1000.9248C479.3363,-994.6637 486.2881,-983.3555 486.2881,-967 486.2881,-956.0112 483.15,-947.3007 477.8496,-940.8687"/>
+<polygon fill="none" stroke="#404040" points="477.8065,-940.8322 470.6429,-940.0056 468.6506,-933.0752 475.8142,-933.9017 477.8065,-940.8322"/>
+<text text-anchor="middle" x="512.2881" y="-964.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #deleter_</text>
 </g>
 <!-- Node8 -->
 <g id="node5" class="node">
 <title>Node8</title>
 <g id="a_node5"><a xlink:href="classtvm_1_1Span.html" target="_top" xlink:title="{tvm::Span\n||+ Span()\l+ Merge()\l+ TVM_DEFINE_OBJECT_REF\l_METHODS()\l}">
-<polygon fill="#ffffff" stroke="#000000" points="113.6435,-900 113.6435,-990 267.6435,-990 267.6435,-900 113.6435,-900"/>
-<text text-anchor="middle" x="190.6435" y="-978" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::Span</text>
-<polyline fill="none" stroke="#000000" points="113.6435,-971 267.6435,-971 "/>
-<text text-anchor="middle" x="190.6435" y="-959" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="113.6435,-952 267.6435,-952 "/>
-<text text-anchor="start" x="121.6435" y="-940" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Span()</text>
-<text text-anchor="start" x="121.6435" y="-929" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Merge()</text>
-<text text-anchor="start" x="121.6435" y="-918" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="121.6435" y="-907" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<polygon fill="#ffffff" stroke="#000000" points="112.7881,-922 112.7881,-1012 266.7881,-1012 266.7881,-922 112.7881,-922"/>
+<text text-anchor="middle" x="189.7881" y="-1000" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::Span</text>
+<polyline fill="none" stroke="#000000" points="112.7881,-993 266.7881,-993 "/>
+<text text-anchor="middle" x="189.7881" y="-981" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="112.7881,-974 266.7881,-974 "/>
+<text text-anchor="start" x="120.7881" y="-962" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Span()</text>
+<text text-anchor="start" x="120.7881" y="-951" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Merge()</text>
+<text text-anchor="start" x="120.7881" y="-940" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="120.7881" y="-929" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
 </a>
 </g>
 </g>
 <!-- Node8&#45;&gt;Node6 -->
 <g id="edge5" class="edge">
 <title>Node8&#45;&gt;Node6</title>
-<path fill="none" stroke="#404040" d="M208.8649,-899.891C225.4299,-860.0471 251.2513,-800.7207 277.6435,-751 294.6503,-718.9607 315.5034,-684.6395 333.7581,-655.9119"/>
-<polygon fill="none" stroke="#404040" points="333.8898,-655.7058 333.7519,-648.496 340.354,-645.5957 340.492,-652.8055 333.8898,-655.7058"/>
-<text text-anchor="middle" x="344.1435" y="-696" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +span</text>
+<path fill="none" stroke="#404040" d="M208.0095,-921.891C224.5745,-882.0471 250.3959,-822.7207 276.7881,-773 293.7949,-740.9607 314.648,-706.6395 332.9027,-677.9119"/>
+<polygon fill="none" stroke="#404040" points="333.0344,-677.7058 332.8965,-670.496 339.4986,-667.5957 339.6366,-674.8055 333.0344,-677.7058"/>
+<text text-anchor="middle" x="343.2881" y="-718" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +span</text>
 </g>
 <!-- Node9 -->
 <g id="node6" class="node">
 <title>Node9</title>
 <g id="a_node6"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="50.6435,-1176.5 50.6435,-1398.5 184.6435,-1398.5 184.6435,-1176.5 50.6435,-1176.5"/>
-<text text-anchor="middle" x="117.6435" y="-1386.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="50.6435,-1379.5 184.6435,-1379.5 "/>
-<text text-anchor="start" x="58.6435" y="-1367.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<polyline fill="none" stroke="#000000" points="50.6435,-1360.5 184.6435,-1360.5 "/>
-<text text-anchor="start" x="58.6435" y="-1348.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="58.6435" y="-1337.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="58.6435" y="-1326.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="58.6435" y="-1315.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="58.6435" y="-1304.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="58.6435" y="-1293.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
-<text text-anchor="start" x="58.6435" y="-1282.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="58.6435" y="-1271.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="58.6435" y="-1260.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="58.6435" y="-1249.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="58.6435" y="-1238.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="58.6435" y="-1227.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="58.6435" y="-1216.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="58.6435" y="-1205.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="58.6435" y="-1194.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="58.6435" y="-1183.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<polygon fill="#ffffff" stroke="#000000" points="49.7881,-1198.5 49.7881,-1420.5 183.7881,-1420.5 183.7881,-1198.5 49.7881,-1198.5"/>
+<text text-anchor="middle" x="116.7881" y="-1408.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="49.7881,-1401.5 183.7881,-1401.5 "/>
+<text text-anchor="start" x="57.7881" y="-1389.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<polyline fill="none" stroke="#000000" points="49.7881,-1382.5 183.7881,-1382.5 "/>
+<text text-anchor="start" x="57.7881" y="-1370.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="57.7881" y="-1359.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="57.7881" y="-1348.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="57.7881" y="-1337.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="57.7881" y="-1326.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="57.7881" y="-1315.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
+<text text-anchor="start" x="57.7881" y="-1304.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="57.7881" y="-1293.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="57.7881" y="-1282.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="57.7881" y="-1271.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="57.7881" y="-1260.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="57.7881" y="-1249.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="57.7881" y="-1238.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="57.7881" y="-1227.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="57.7881" y="-1216.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="57.7881" y="-1205.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
 </a>
 </g>
 </g>
 <!-- Node9&#45;&gt;Node5 -->
 <g id="edge8" class="edge">
 <title>Node9&#45;&gt;Node5</title>
-<path fill="none" stroke="#404040" d="M63.3454,-1176.2476C58.7792,-1163.8867 54.7351,-1151.3136 51.6435,-1139 .1318,-933.8361 -18.2223,-871.7412 21.6435,-664 57.169,-478.8763 76.7338,-428.5789 178.6435,-270 180.5339,-267.0584 182.5609,-264.133 184.6891,-261.2413"/>
-<polygon fill="none" stroke="#404040" points="184.8673,-261.0136 185.4156,-253.8233 192.2636,-251.564 191.7153,-258.7542 184.8673,-261.0136"/>
-<text text-anchor="middle" x="64.6435" y="-696" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +virtual_device_</text>
+<path fill="none" stroke="#404040" d="M62.49,-1198.2476C57.9238,-1185.8867 53.8798,-1173.3136 50.7881,-1161 -.7236,-955.8361 -17.1735,-894.0976 20.7881,-686 56.2528,-491.5904 70.6257,-436.0393 177.7881,-270 179.6829,-267.0642 181.7202,-264.1537 183.8648,-261.2832"/>
+<polygon fill="none" stroke="#404040" points="184.0296,-261.0768 184.6487,-253.8923 191.5186,-251.7005 190.8996,-258.885 184.0296,-261.0768"/>
+<text text-anchor="middle" x="63.7881" y="-718" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +virtual_device_</text>
 </g>
 <!-- Node9&#45;&gt;Node8 -->
 <g id="edge6" class="edge">
 <title>Node9&#45;&gt;Node8</title>
-<path fill="none" stroke="#191970" d="M143.4995,-1166.1895C156.5205,-1105.0976 171.522,-1034.7137 181.0402,-990.0565"/>
-<polygon fill="none" stroke="#191970" points="140.006,-1165.7901 141.3445,-1176.3001 146.8523,-1167.2494 140.006,-1165.7901"/>
+<path fill="none" stroke="#191970" d="M142.6441,-1188.1895C155.6651,-1127.0976 170.6666,-1056.7137 180.1848,-1012.0565"/>
+<polygon fill="none" stroke="#191970" points="139.1506,-1187.7901 140.4891,-1198.3001 145.9969,-1189.2494 139.1506,-1187.7901"/>
 </g>
 <!-- Node11 -->
 <g id="node8" class="node">
 <title>Node11</title>
 <g id="a_node8"><a xlink:href="classtvm_1_1Type.html" target="_top" xlink:title="Managed reference to TypeNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="116.6435,-664.5 116.6435,-732.5 270.6435,-732.5 270.6435,-664.5 116.6435,-664.5"/>
-<text text-anchor="middle" x="193.6435" y="-720.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::Type</text>
-<polyline fill="none" stroke="#000000" points="116.6435,-713.5 270.6435,-713.5 "/>
-<text text-anchor="middle" x="193.6435" y="-701.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="116.6435,-694.5 270.6435,-694.5 "/>
-<text text-anchor="start" x="124.6435" y="-682.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="124.6435" y="-671.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<polygon fill="#ffffff" stroke="#000000" points="115.7881,-686.5 115.7881,-754.5 269.7881,-754.5 269.7881,-686.5 115.7881,-686.5"/>
+<text text-anchor="middle" x="192.7881" y="-742.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::Type</text>
+<polyline fill="none" stroke="#000000" points="115.7881,-735.5 269.7881,-735.5 "/>
+<text text-anchor="middle" x="192.7881" y="-723.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="115.7881,-716.5 269.7881,-716.5 "/>
+<text text-anchor="start" x="123.7881" y="-704.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="123.7881" y="-693.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
 </a>
 </g>
 </g>
 <!-- Node9&#45;&gt;Node11 -->
 <g id="edge10" class="edge">
 <title>Node9&#45;&gt;Node11</title>
-<path fill="none" stroke="#191970" d="M80.919,-1166.4783C53.5032,-1051.6676 31.9641,-878.2549 104.6435,-751 108.5582,-744.1458 113.7563,-738.0968 119.6811,-732.7788"/>
-<polygon fill="none" stroke="#191970" points="77.5602,-1167.4784 83.3289,-1176.365 84.3611,-1165.8206 77.5602,-1167.4784"/>
+<path fill="none" stroke="#191970" d="M80.0636,-1188.4783C52.6478,-1073.6676 31.1087,-900.2549 103.7881,-773 107.7028,-766.1458 112.9009,-760.0968 118.8257,-754.7788"/>
+<polygon fill="none" stroke="#191970" points="76.7048,-1189.4784 82.4735,-1198.365 83.5057,-1187.8206 76.7048,-1189.4784"/>
 </g>
 <!-- Node12 -->
 <g id="node9" class="node">
 <title>Node12</title>
 <g id="a_node9"><a xlink:href="classtvm_1_1runtime_1_1NDArray.html" target="_top" xlink:title="Managed NDArray. The array is backed by reference counted blocks. ">
-<polygon fill="#ffffff" stroke="#000000" points="398.1435,-270.5 398.1435,-514.5 529.1435,-514.5 529.1435,-270.5 398.1435,-270.5"/>
-<text text-anchor="middle" x="463.6435" y="-502.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
-<polyline fill="none" stroke="#000000" points="398.1435,-495.5 529.1435,-495.5 "/>
-<text text-anchor="middle" x="463.6435" y="-483.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="398.1435,-476.5 529.1435,-476.5 "/>
-<text text-anchor="start" x="406.1435" y="-464.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="406.1435" y="-453.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="406.1435" y="-442.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
-<text text-anchor="start" x="406.1435" y="-431.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="406.1435" y="-420.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="406.1435" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
-<text text-anchor="start" x="406.1435" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="406.1435" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="406.1435" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
-<text text-anchor="start" x="406.1435" y="-365.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
-<text text-anchor="start" x="406.1435" y="-354.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
-<text text-anchor="start" x="406.1435" y="-343.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
-<text text-anchor="start" x="406.1435" y="-332.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
-<text text-anchor="start" x="406.1435" y="-321.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
-<text text-anchor="start" x="406.1435" y="-310.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="406.1435" y="-299.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
-<text text-anchor="start" x="406.1435" y="-288.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
-<text text-anchor="start" x="406.1435" y="-277.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
+<polygon fill="#ffffff" stroke="#000000" points="395.2881,-270.5 395.2881,-536.5 538.2881,-536.5 538.2881,-270.5 395.2881,-270.5"/>
+<text text-anchor="middle" x="466.7881" y="-524.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
+<polyline fill="none" stroke="#000000" points="395.2881,-517.5 538.2881,-517.5 "/>
+<text text-anchor="middle" x="466.7881" y="-505.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="395.2881,-498.5 538.2881,-498.5 "/>
+<text text-anchor="start" x="403.2881" y="-486.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="403.2881" y="-475.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="403.2881" y="-464.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
+<text text-anchor="start" x="403.2881" y="-453.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="403.2881" y="-442.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="403.2881" y="-431.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
+<text text-anchor="start" x="403.2881" y="-420.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="403.2881" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="403.2881" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
+<text text-anchor="start" x="403.2881" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
+<text text-anchor="start" x="403.2881" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
+<text text-anchor="start" x="403.2881" y="-365.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
+<text text-anchor="start" x="403.2881" y="-354.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromExternalDLTensor()</text>
+<text text-anchor="start" x="403.2881" y="-343.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NewFromDLTensor()</text>
+<text text-anchor="start" x="403.2881" y="-332.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
+<text text-anchor="start" x="403.2881" y="-321.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
+<text text-anchor="start" x="403.2881" y="-310.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="403.2881" y="-299.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
+<text text-anchor="start" x="403.2881" y="-288.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
+<text text-anchor="start" x="403.2881" y="-277.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
 </a>
 </g>
 </g>
 <!-- Node9&#45;&gt;Node12 -->
 <g id="edge12" class="edge">
 <title>Node9&#45;&gt;Node12</title>
-<path fill="none" stroke="#191970" d="M194.8173,-1272.0533C307.1456,-1247.8258 506.881,-1197.5611 548.6435,-1139 680.8964,-953.5497 593.1718,-673.7551 524.0384,-514.6161"/>
-<polygon fill="none" stroke="#191970" points="193.9396,-1268.6617 184.8924,-1274.1753 195.4032,-1275.507 193.9396,-1268.6617"/>
+<path fill="none" stroke="#191970" d="M193.9619,-1294.0533C306.2902,-1269.8258 506.0256,-1219.5611 547.7881,-1161 679.3926,-976.459 598.4967,-699.5268 530.6328,-536.6034"/>
+<polygon fill="none" stroke="#191970" points="193.0842,-1290.6617 184.037,-1296.1753 194.5478,-1297.507 193.0842,-1290.6617"/>
 </g>
 <!-- Node10 -->
 <g id="node7" class="node">
 <title>Node10</title>
 <g id="a_node7"><a xlink:href="classtvm_1_1runtime_1_1ObjectPtr.html" target="_top" xlink:title="{tvm::runtime::ObjectPtr\l\&lt; tvm::runtime::Object \&gt;\n||+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ~ObjectPtr()\l+ swap()\l+ get()\l+ operator&#45;\&gt;()\land 11 more...\l}">
-<polygon fill="#ffffff" stroke="#000000" points="47.6435,-1446.5 47.6435,-1624.5 187.6435,-1624.5 187.6435,-1446.5 47.6435,-1446.5"/>
-<text text-anchor="start" x="55.6435" y="-1612.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
-<text text-anchor="middle" x="117.6435" y="-1601.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
-<polyline fill="none" stroke="#000000" points="47.6435,-1594.5 187.6435,-1594.5 "/>
-<text text-anchor="middle" x="117.6435" y="-1582.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="47.6435,-1575.5 187.6435,-1575.5 "/>
-<text text-anchor="start" x="55.6435" y="-1563.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1552.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1541.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1530.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1519.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1508.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1497.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
-<text text-anchor="start" x="55.6435" y="-1486.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
-<text text-anchor="start" x="55.6435" y="-1475.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="55.6435" y="-1464.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="55.6435" y="-1453.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
+<polygon fill="#ffffff" stroke="#000000" points="46.7881,-1468.5 46.7881,-1646.5 186.7881,-1646.5 186.7881,-1468.5 46.7881,-1468.5"/>
+<text text-anchor="start" x="54.7881" y="-1634.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
+<text text-anchor="middle" x="116.7881" y="-1623.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
+<polyline fill="none" stroke="#000000" points="46.7881,-1616.5 186.7881,-1616.5 "/>
+<text text-anchor="middle" x="116.7881" y="-1604.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="46.7881,-1597.5 186.7881,-1597.5 "/>
+<text text-anchor="start" x="54.7881" y="-1585.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1574.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1563.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1552.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1541.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1530.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1519.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
+<text text-anchor="start" x="54.7881" y="-1508.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
+<text text-anchor="start" x="54.7881" y="-1497.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="54.7881" y="-1486.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="54.7881" y="-1475.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
 </a>
 </g>
 </g>
 <!-- Node10&#45;&gt;Node9 -->
 <g id="edge7" class="edge">
 <title>Node10&#45;&gt;Node9</title>
-<path fill="none" stroke="#404040" d="M117.6435,-1446.3167C117.6435,-1434.8765 117.6435,-1423.0062 117.6435,-1411.1402"/>
-<polygon fill="none" stroke="#404040" points="117.6436,-1410.7944 113.6435,-1404.7944 117.6435,-1398.7944 121.6435,-1404.7943 117.6436,-1410.7944"/>
-<text text-anchor="middle" x="137.1435" y="-1420" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
+<path fill="none" stroke="#404040" d="M116.7881,-1468.3167C116.7881,-1456.8765 116.7881,-1445.0062 116.7881,-1433.1402"/>
+<polygon fill="none" stroke="#404040" points="116.7882,-1432.7944 112.7882,-1426.7944 116.7881,-1420.7944 120.7882,-1426.7943 116.7882,-1432.7944"/>
+<text text-anchor="middle" x="136.2881" y="-1442" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
 </g>
 <!-- Node11&#45;&gt;Node5 -->
 <g id="edge9" class="edge">
 <title>Node11&#45;&gt;Node5</title>
-<path fill="none" stroke="#404040" d="M183.9025,-664.2696C182.5474,-658.2292 181.3711,-651.9653 180.6435,-646 174.5631,-596.1472 175.8687,-582.9947 180.6435,-533 189.7064,-438.1072 213.0245,-330.0091 228.8655,-263.6435"/>
-<polygon fill="none" stroke="#404040" points="228.9203,-263.4159 226.4371,-256.6459 231.7313,-251.7498 234.2145,-258.5198 228.9203,-263.4159"/>
-<text text-anchor="middle" x="223.1435" y="-587" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +checked_type_</text>
+<path fill="none" stroke="#404040" d="M183.0471,-686.2696C181.692,-680.2292 180.5157,-673.9653 179.7881,-668 173.7077,-618.1472 175.0978,-605.0027 179.7881,-555 189.4793,-451.6826 214.7481,-333.6044 231.334,-263.2879"/>
+<polygon fill="none" stroke="#404040" points="231.343,-263.25 228.8405,-256.4871 234.1207,-251.5759 236.6232,-258.3389 231.343,-263.25"/>
+<text text-anchor="middle" x="222.2881" y="-609" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +checked_type_</text>
 </g>
 <!-- Node12&#45;&gt;Node4 -->
 <g id="edge11" class="edge">
 <title>Node12&#45;&gt;Node4</title>
-<path fill="none" stroke="#404040" d="M422.8921,-270.2458C406.484,-221.0216 388.1854,-166.1256 374.2411,-124.2928"/>
-<polygon fill="none" stroke="#404040" points="374.2235,-124.2399 368.5314,-119.8127 370.4287,-112.8557 376.1209,-117.2829 374.2235,-124.2399"/>
-<text text-anchor="middle" x="432.6435" y="-198.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +data</text>
+<path fill="none" stroke="#404040" d="M423.8063,-270.3331C407.714,-220.4757 390.1033,-165.9142 376.6989,-124.3844"/>
+<polygon fill="none" stroke="#404040" points="376.6323,-124.1778 370.9827,-119.6965 372.9463,-112.7579 378.5959,-117.2392 376.6323,-124.1778"/>
+<text text-anchor="middle" x="431.7881" y="-198.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> +data</text>
 </g>
 </g>
 </svg>
diff --git a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray-members.html b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray-members.html
index 78706474b..cd16f3a54 100644
--- a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray-members.html
+++ b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray-members.html
@@ -91,13 +91,15 @@ $(function() {
   <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#ade0e2757904f4f5ba5c667ae01793a47">FFIDecRef</a>(TVMArrayHandle handle)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span><span class="mlabel">static</span></td></tr>
   <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a141e032d848c60f8261046304bdc8c4c">FFIGetHandle</a>(const ObjectRef &amp;nd)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span><span class="mlabel">static</span></td></tr>
   <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#abec485628a0ca451b668c42fd8fa691a">FromDLPack</a>(DLManagedTensor *tensor)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
-  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aadbc0886ffa80162ff31eefd0431ba09">get</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
-  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#ad5d21a1d7a704bfc9504d28910748d39">get_mutable</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span></td></tr>
-  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aed593996e4076632450de8fde776707c">GetDataPtr</a>(const ObjectRef &amp;ref)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span><span class="mlabel">static</span></td></tr>
-  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a1c9d84d35af95d1c29fc5c9c2ced84c8">IsContiguous</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
-  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#ad78792a1e1feb160b0be4474a4c13a4c">Load</a>(dmlc::Stream *stream)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
-  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a4bbb80e8e36317829dd63e7f44ffbb0f">NDArray</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
-  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#af5801a105ceb450616a83d19c5c92326">NDArray</a>(ObjectPtr&lt; Object &gt; data)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e">FromExternalDLTensor</a>(const DLTensor &amp;dl_tensor)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
+  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aadbc0886ffa80162ff31eefd0431ba09">get</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#ad5d21a1d7a704bfc9504d28910748d39">get_mutable</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span></td></tr>
+  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aed593996e4076632450de8fde776707c">GetDataPtr</a>(const ObjectRef &amp;ref)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span><span class="mlabel">static</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a1c9d84d35af95d1c29fc5c9c2ced84c8">IsContiguous</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#ad78792a1e1feb160b0be4474a4c13a4c">Load</a>(dmlc::Stream *stream)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a4bbb80e8e36317829dd63e7f44ffbb0f">NDArray</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+  <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#af5801a105ceb450616a83d19c5c92326">NDArray</a>(ObjectPtr&lt; Object &gt; data)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2">NewFromDLTensor</a>(DLTensor *dl_tensor, Device dev)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
   <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aa07c1f6d66a438ea950637d13ed09471">ObjectRef</a>()=default</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"></td></tr>
   <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#a6a7dd7404edf1c26f8dbd9bd92d03a02">ObjectRef</a>(ObjectPtr&lt; Object &gt; data)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
   <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#aa1bd13a7185cb4b2b6bdde49416e8aa4">operator!=</a>(const ObjectRef &amp;other) const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
diff --git a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray.html b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray.html
index 97f5fe145..bf868c27d 100644
--- a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray.html
+++ b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray.html
@@ -83,13 +83,13 @@ $(function() {
 <div class="dynheader">
 Inheritance diagram for tvm::runtime::NDArray:</div>
 <div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1runtime_1_1NDArray__inherit__graph.svg" width="190" height="698"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1runtime_1_1NDArray__inherit__graph.svg" width="202" height="727"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
 </div>
 </div>
 <div class="dynheader">
 Collaboration diagram for tvm::runtime::NDArray:</div>
 <div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1runtime_1_1NDArray__coll__graph.svg" width="198" height="986"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1runtime_1_1NDArray__coll__graph.svg" width="202" height="1015"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
 </div>
 </div>
 <table class="memberdecls">
@@ -203,6 +203,12 @@ Static Public Member Functions</h2></td></tr>
 <tr class="memitem:a59f41733876e0a161de701de9fd60749"><td class="memItemLeft" align="right" valign="top">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a59f41733876e0a161de701de9fd60749">Empty</a> (<a class="el" href="classtvm_1_1runtime_1_1ShapeTuple.html">ShapeTuple</a> shape, DLDataType dtype, <a class="el" href="namespacetvm.html#a7c2095aed90b2129ba6 [...]
 <tr class="memdesc:a59f41733876e0a161de701de9fd60749"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create an empty <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a>.  <a href="#a59f41733876e0a161de701de9fd60749">More...</a><br /></td></tr>
 <tr class="separator:a59f41733876e0a161de701de9fd60749"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a356d1886b24da68c35a0d0b826c9359e"><td class="memItemLeft" align="right" valign="top">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e">FromExternalDLTensor</a> (const DLTensor &amp;dl_tensor)</td></tr>
+<tr class="memdesc:a356d1886b24da68c35a0d0b826c9359e"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> backed by an external DLTensor.  <a href="#a356d1886b24da68c35a0d0b826c9359e">More...</a><br /></td></tr>
+<tr class="separator:a356d1886b24da68c35a0d0b826c9359e"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:afb6060bb96dad082c1deca26e6b58ae2"><td class="memItemLeft" align="right" valign="top">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2">NewFromDLTensor</a> (DLTensor *dl_tensor, <a class="el" href="namespacetvm.html#a7c2095aed90b2129ba631b90103313a2">Device</a> dev)</td></tr>
+<tr class="memdesc:afb6060bb96dad082c1deca26e6b58ae2"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create new <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a>, data is copied from DLTensor.  <a href="#afb6060bb96dad082c1deca26e6b58ae2">More...</a><br /></td></tr>
+<tr class="separator:afb6060bb96dad082c1deca26e6b58ae2"><td class="memSeparator" colspan="2">&#160;</td></tr>
 <tr class="memitem:abec485628a0ca451b668c42fd8fa691a"><td class="memItemLeft" align="right" valign="top">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1runtime_1_1NDArray.html#abec485628a0ca451b668c42fd8fa691a">FromDLPack</a> (DLManagedTensor *tensor)</td></tr>
 <tr class="memdesc:abec485628a0ca451b668c42fd8fa691a"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> backed by a dlpack tensor.  <a href="#abec485628a0ca451b668c42fd8fa691a">More...</a><br /></td></tr>
 <tr class="separator:abec485628a0ca451b668c42fd8fa691a"><td class="memSeparator" colspan="2">&#160;</td></tr>
@@ -883,6 +889,41 @@ Additional Inherited Members</h2></td></tr>
 </dl>
 <dl class="section return"><dt>Returns</dt><dd>The created <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> view. </dd></dl>
 
+</div>
+</div>
+<a id="a356d1886b24da68c35a0d0b826c9359e"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a356d1886b24da68c35a0d0b826c9359e">&#9670;&nbsp;</a></span>FromExternalDLTensor()</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a> tvm::runtime::NDArray::FromExternalDLTensor </td>
+          <td>(</td>
+          <td class="paramtype">const DLTensor &amp;&#160;</td>
+          <td class="paramname"><em>dl_tensor</em></td><td>)</td>
+          <td></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Create a <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> backed by an external DLTensor. </p>
+<p>This allows us to create a <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> using the memory allocated by an external source. Responsibility for memory retaining lies with the external source. </p><dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">dl_tensor</td><td>The DLTensor to copy from. </td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>The created <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> view. </dd></dl>
+
 </div>
 </div>
 <a id="ad5d21a1d7a704bfc9504d28910748d39"></a>
@@ -972,6 +1013,52 @@ Additional Inherited Members</h2></td></tr>
 </dl>
 <dl class="section return"><dt>Returns</dt><dd>Whether load is successful </dd></dl>
 
+</div>
+</div>
+<a id="afb6060bb96dad082c1deca26e6b58ae2"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#afb6060bb96dad082c1deca26e6b58ae2">&#9670;&nbsp;</a></span>NewFromDLTensor()</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">static <a class="el" href="classtvm_1_1runtime_1_1NDArray.html">NDArray</a> tvm::runtime::NDArray::NewFromDLTensor </td>
+          <td>(</td>
+          <td class="paramtype">DLTensor *&#160;</td>
+          <td class="paramname"><em>dl_tensor</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype"><a class="el" href="namespacetvm.html#a7c2095aed90b2129ba631b90103313a2">Device</a>&#160;</td>
+          <td class="paramname"><em>dev</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">static</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Create new <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a>, data is copied from DLTensor. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">dl_tensor</td><td>The DLTensor to copy from. </td></tr>
+    <tr><td class="paramname">dev</td><td>device location of the created <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a>. </td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>The created <a class="el" href="classtvm_1_1runtime_1_1NDArray.html" title="Managed NDArray. The array is backed by reference counted blocks. ">NDArray</a> view. </dd></dl>
+
 </div>
 </div>
 <a id="ad57933f49a9fd51d7f996e1b16ffd2e0"></a>
diff --git a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__coll__graph.svg
index 865d6c4dd..235c895ad 100644
--- a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__coll__graph.svg
@@ -4,102 +4,104 @@
 <!-- Generated by graphviz version 2.40.1 (20161225.0304)
  -->
 <!-- Title: tvm::runtime::NDArray Pages: 1 -->
-<svg width="148pt" height="739pt"
- viewBox="0.00 0.00 148.00 739.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 735)">
+<svg width="151pt" height="761pt"
+ viewBox="0.00 0.00 151.00 761.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 757)">
 <title>tvm::runtime::NDArray</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-735 144,-735 144,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-757 147,-757 147,4 -4,4"/>
 <!-- Node2 -->
 <g id="node1" class="node">
 <title>Node2</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="4.5,-.5 4.5,-244.5 135.5,-244.5 135.5,-.5 4.5,-.5"/>
-<text text-anchor="middle" x="70" y="-232.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
-<polyline fill="none" stroke="#000000" points="4.5,-225.5 135.5,-225.5 "/>
-<text text-anchor="middle" x="70" y="-213.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="4.5,-206.5 135.5,-206.5 "/>
-<text text-anchor="start" x="12.5" y="-194.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="12.5" y="-183.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="12.5" y="-172.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
-<text text-anchor="start" x="12.5" y="-161.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="12.5" y="-150.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="12.5" y="-139.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
-<text text-anchor="start" x="12.5" y="-128.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="12.5" y="-117.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="12.5" y="-106.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
-<text text-anchor="start" x="12.5" y="-95.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
-<text text-anchor="start" x="12.5" y="-84.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
-<text text-anchor="start" x="12.5" y="-73.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
-<text text-anchor="start" x="12.5" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
-<text text-anchor="start" x="12.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
-<text text-anchor="start" x="12.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="12.5" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
-<text text-anchor="start" x="12.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
-<text text-anchor="start" x="12.5" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="0,-.5 0,-266.5 143,-266.5 143,-.5 0,-.5"/>
+<text text-anchor="middle" x="71.5" y="-254.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
+<polyline fill="none" stroke="#000000" points="0,-247.5 143,-247.5 "/>
+<text text-anchor="middle" x="71.5" y="-235.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-228.5 143,-228.5 "/>
+<text text-anchor="start" x="8" y="-216.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="8" y="-205.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="8" y="-194.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
+<text text-anchor="start" x="8" y="-183.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="8" y="-172.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="8" y="-161.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
+<text text-anchor="start" x="8" y="-150.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="8" y="-139.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="8" y="-128.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
+<text text-anchor="start" x="8" y="-117.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
+<text text-anchor="start" x="8" y="-106.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
+<text text-anchor="start" x="8" y="-95.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
+<text text-anchor="start" x="8" y="-84.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromExternalDLTensor()</text>
+<text text-anchor="start" x="8" y="-73.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NewFromDLTensor()</text>
+<text text-anchor="start" x="8" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
+<text text-anchor="start" x="8" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
+<text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
+<text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
 </g>
 <!-- Node3 -->
 <g id="node2" class="node">
 <title>Node3</title>
 <g id="a_node2"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="3,-282.5 3,-504.5 137,-504.5 137,-282.5 3,-282.5"/>
-<text text-anchor="middle" x="70" y="-492.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="3,-485.5 137,-485.5 "/>
-<text text-anchor="start" x="11" y="-473.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<polyline fill="none" stroke="#000000" points="3,-466.5 137,-466.5 "/>
-<text text-anchor="start" x="11" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="11" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="11" y="-432.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="11" y="-421.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="11" y="-410.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="11" y="-399.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
-<text text-anchor="start" x="11" y="-388.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="11" y="-377.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="11" y="-366.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="11" y="-355.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="11" y="-344.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="11" y="-333.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="11" y="-322.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="11" y="-311.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="11" y="-300.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="11" y="-289.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<polygon fill="#ffffff" stroke="#000000" points="4.5,-304.5 4.5,-526.5 138.5,-526.5 138.5,-304.5 4.5,-304.5"/>
+<text text-anchor="middle" x="71.5" y="-514.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="4.5,-507.5 138.5,-507.5 "/>
+<text text-anchor="start" x="12.5" y="-495.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<polyline fill="none" stroke="#000000" points="4.5,-488.5 138.5,-488.5 "/>
+<text text-anchor="start" x="12.5" y="-476.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="12.5" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="12.5" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="12.5" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="12.5" y="-432.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="12.5" y="-421.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
+<text text-anchor="start" x="12.5" y="-410.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="12.5" y="-399.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="12.5" y="-388.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="12.5" y="-377.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="12.5" y="-366.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="12.5" y="-355.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="12.5" y="-344.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="12.5" y="-333.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="12.5" y="-322.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="12.5" y="-311.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
 </a>
 </g>
 </g>
 <!-- Node3&#45;&gt;Node2 -->
 <g id="edge1" class="edge">
 <title>Node3&#45;&gt;Node2</title>
-<path fill="none" stroke="#191970" d="M70,-272.2451C70,-263.1286 70,-253.9251 70,-244.8022"/>
-<polygon fill="none" stroke="#191970" points="66.5001,-272.492 70,-282.492 73.5001,-272.492 66.5001,-272.492"/>
+<path fill="none" stroke="#191970" d="M71.5,-294.2725C71.5,-285.2083 71.5,-276.0328 71.5,-266.8999"/>
+<polygon fill="none" stroke="#191970" points="68.0001,-294.4509 71.5,-304.4509 75.0001,-294.451 68.0001,-294.4509"/>
 </g>
 <!-- Node4 -->
 <g id="node3" class="node">
 <title>Node4</title>
 <g id="a_node3"><a xlink:href="classtvm_1_1runtime_1_1ObjectPtr.html" target="_top" xlink:title="{tvm::runtime::ObjectPtr\l\&lt; tvm::runtime::Object \&gt;\n||+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ~ObjectPtr()\l+ swap()\l+ get()\l+ operator&#45;\&gt;()\land 11 more...\l}">
-<polygon fill="#ffffff" stroke="#000000" points="0,-552.5 0,-730.5 140,-730.5 140,-552.5 0,-552.5"/>
-<text text-anchor="start" x="8" y="-718.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
-<text text-anchor="middle" x="70" y="-707.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
-<polyline fill="none" stroke="#000000" points="0,-700.5 140,-700.5 "/>
-<text text-anchor="middle" x="70" y="-688.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-681.5 140,-681.5 "/>
-<text text-anchor="start" x="8" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
-<text text-anchor="start" x="8" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
-<text text-anchor="start" x="8" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="8" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="8" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
+<polygon fill="#ffffff" stroke="#000000" points="1.5,-574.5 1.5,-752.5 141.5,-752.5 141.5,-574.5 1.5,-574.5"/>
+<text text-anchor="start" x="9.5" y="-740.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
+<text text-anchor="middle" x="71.5" y="-729.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">&lt; tvm::runtime::Object &gt;</text>
+<polyline fill="none" stroke="#000000" points="1.5,-722.5 141.5,-722.5 "/>
+<text text-anchor="middle" x="71.5" y="-710.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="1.5,-703.5 141.5,-703.5 "/>
+<text text-anchor="start" x="9.5" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
+<text text-anchor="start" x="9.5" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
+<text text-anchor="start" x="9.5" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="9.5" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="9.5" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
 </a>
 </g>
 </g>
 <!-- Node4&#45;&gt;Node3 -->
 <g id="edge2" class="edge">
 <title>Node4&#45;&gt;Node3</title>
-<path fill="none" stroke="#404040" d="M70,-552.3167C70,-540.8765 70,-529.0062 70,-517.1402"/>
-<polygon fill="none" stroke="#404040" points="70.0001,-516.7944 66,-510.7944 70,-504.7944 74,-510.7943 70.0001,-516.7944"/>
-<text text-anchor="middle" x="89.5" y="-526" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
+<path fill="none" stroke="#404040" d="M71.5,-574.3167C71.5,-562.8765 71.5,-551.0062 71.5,-539.1402"/>
+<polygon fill="none" stroke="#404040" points="71.5001,-538.7944 67.5,-532.7944 71.5,-526.7944 75.5,-532.7943 71.5001,-538.7944"/>
+<text text-anchor="middle" x="91" y="-548" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
 </g>
 </g>
 </svg>
diff --git a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__inherit__graph.svg
index 6a69c7f55..377a10999 100644
--- a/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1runtime_1_1NDArray__inherit__graph.svg
@@ -4,72 +4,74 @@
 <!-- Generated by graphviz version 2.40.1 (20161225.0304)
  -->
 <!-- Title: tvm::runtime::NDArray Pages: 1 -->
-<svg width="142pt" height="523pt"
- viewBox="0.00 0.00 142.00 523.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 519)">
+<svg width="151pt" height="545pt"
+ viewBox="0.00 0.00 151.00 545.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 541)">
 <title>tvm::runtime::NDArray</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-519 138,-519 138,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-541 147,-541 147,4 -4,4"/>
 <!-- Node0 -->
 <g id="node1" class="node">
 <title>Node0</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="1.5,-.5 1.5,-244.5 132.5,-244.5 132.5,-.5 1.5,-.5"/>
-<text text-anchor="middle" x="67" y="-232.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
-<polyline fill="none" stroke="#000000" points="1.5,-225.5 132.5,-225.5 "/>
-<text text-anchor="middle" x="67" y="-213.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="1.5,-206.5 132.5,-206.5 "/>
-<text text-anchor="start" x="9.5" y="-194.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="9.5" y="-183.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
-<text text-anchor="start" x="9.5" y="-172.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
-<text text-anchor="start" x="9.5" y="-161.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="9.5" y="-150.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="9.5" y="-139.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
-<text text-anchor="start" x="9.5" y="-128.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="9.5" y="-117.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
-<text text-anchor="start" x="9.5" y="-106.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
-<text text-anchor="start" x="9.5" y="-95.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
-<text text-anchor="start" x="9.5" y="-84.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
-<text text-anchor="start" x="9.5" y="-73.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
-<text text-anchor="start" x="9.5" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
-<text text-anchor="start" x="9.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
-<text text-anchor="start" x="9.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="9.5" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
-<text text-anchor="start" x="9.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
-<text text-anchor="start" x="9.5" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="0,-.5 0,-266.5 143,-266.5 143,-.5 0,-.5"/>
+<text text-anchor="middle" x="71.5" y="-254.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::NDArray</text>
+<polyline fill="none" stroke="#000000" points="0,-247.5 143,-247.5 "/>
+<text text-anchor="middle" x="71.5" y="-235.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-228.5 143,-228.5 "/>
+<text text-anchor="start" x="8" y="-216.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="8" y="-205.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NDArray()</text>
+<text text-anchor="start" x="8" y="-194.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ reset()</text>
+<text text-anchor="start" x="8" y="-183.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="8" y="-172.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="8" y="-161.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ IsContiguous()</text>
+<text text-anchor="start" x="8" y="-150.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="8" y="-139.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFrom()</text>
+<text text-anchor="start" x="8" y="-128.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromBytes()</text>
+<text text-anchor="start" x="8" y="-117.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyTo()</text>
+<text text-anchor="start" x="8" y="-106.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 9 more...</text>
+<text text-anchor="start" x="8" y="-95.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Empty()</text>
+<text text-anchor="start" x="8" y="-84.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromExternalDLTensor()</text>
+<text text-anchor="start" x="8" y="-73.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ NewFromDLTensor()</text>
+<text text-anchor="start" x="8" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ FromDLPack()</text>
+<text text-anchor="start" x="8" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ CopyFromTo()</text>
+<text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDataFromHandle()</text>
+<text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIDecRef()</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIGetHandle()</text>
 </g>
 <!-- Node1 -->
 <g id="node2" class="node">
 <title>Node1</title>
 <g id="a_node2"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-281.5 0,-514.5 134,-514.5 134,-281.5 0,-281.5"/>
-<text text-anchor="middle" x="67" y="-502.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="0,-495.5 134,-495.5 "/>
-<text text-anchor="start" x="8" y="-483.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<text text-anchor="start" x="8" y="-472.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
-<polyline fill="none" stroke="#000000" points="0,-465.5 134,-465.5 "/>
-<text text-anchor="start" x="8" y="-453.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="8" y="-442.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="8" y="-431.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="8" y="-420.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="8" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="8" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
-<text text-anchor="start" x="8" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="8" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="8" y="-365.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
-<text text-anchor="start" x="8" y="-354.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="8" y="-343.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="8" y="-332.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="8" y="-321.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="8" y="-310.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="8" y="-299.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="8" y="-288.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<polygon fill="#ffffff" stroke="#000000" points="4.5,-303.5 4.5,-536.5 138.5,-536.5 138.5,-303.5 4.5,-303.5"/>
+<text text-anchor="middle" x="71.5" y="-524.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="4.5,-517.5 138.5,-517.5 "/>
+<text text-anchor="start" x="12.5" y="-505.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<text text-anchor="start" x="12.5" y="-494.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
+<polyline fill="none" stroke="#000000" points="4.5,-487.5 138.5,-487.5 "/>
+<text text-anchor="start" x="12.5" y="-475.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="12.5" y="-464.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
+<text text-anchor="start" x="12.5" y="-453.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="12.5" y="-442.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="12.5" y="-431.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="12.5" y="-420.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&lt;()</text>
+<text text-anchor="start" x="12.5" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="12.5" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="12.5" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator&#45;&gt;()</text>
+<text text-anchor="start" x="12.5" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="12.5" y="-365.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="12.5" y="-354.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="12.5" y="-343.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="12.5" y="-332.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="12.5" y="-321.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="12.5" y="-310.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
 </a>
 </g>
 </g>
 <!-- Node1&#45;&gt;Node0 -->
 <g id="edge1" class="edge">
 <title>Node1&#45;&gt;Node0</title>
-<path fill="none" stroke="#191970" d="M67,-270.9732C67,-262.2632 67,-253.4957 67,-244.811"/>
-<polygon fill="none" stroke="#191970" points="63.5001,-271.1632 67,-281.1632 70.5001,-271.1632 63.5001,-271.1632"/>
+<path fill="none" stroke="#191970" d="M71.5,-293.2194C71.5,-284.4484 71.5,-275.5939 71.5,-266.7898"/>
+<polygon fill="none" stroke="#191970" points="68.0001,-293.4694 71.5,-303.4695 75.0001,-293.4695 68.0001,-293.4694"/>
 </g>
 </g>
 </svg>
diff --git a/docs/reference/api/doxygen/functions_f.html b/docs/reference/api/doxygen/functions_f.html
index 847ad96a8..d6022ae9c 100644
--- a/docs/reference/api/doxygen/functions_f.html
+++ b/docs/reference/api/doxygen/functions_f.html
@@ -472,6 +472,9 @@ $(function() {
 <li>FromExprInContext()
 : <a class="el" href="classtvm_1_1IRModule.html#a1cc91fc2b2adaca5a4dcfc14baf28c27">tvm::IRModule</a>
 </li>
+<li>FromExternalDLTensor()
+: <a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e">tvm::runtime::NDArray</a>
+</li>
 <li>FromFunc()
 : <a class="el" href="classtvm_1_1tir_1_1IndexMap.html#afa04f25f10b1dac139df9a1b34598cbb">tvm::tir::IndexMap</a>
 </li>
@@ -594,7 +597,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1te_1_1Fuse.html#a10b77eec10eb7dbc536b0c8d65163f9c">tvm::te::Fuse</a>
 </li>
 <li>fuse()
-: <a class="el" href="classtvm_1_1te_1_1Stage.html#a07b721494aa3c0c79e8a8654c433708f">tvm::te::Stage</a>
+: <a class="el" href="classtvm_1_1te_1_1Stage.html#a5658065d9cbbee620bbd107d30c4ae72">tvm::te::Stage</a>
 </li>
 <li>Fuse()
 : <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a4e60246388be76fe6a1f46c4255da3ff">tvm::tir::ScheduleNode</a>
@@ -606,7 +609,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1auto__scheduler_1_1FuseStepNode.html#a19c1a7b47f59a4f004a2dd9f354835eb">tvm::auto_scheduler::FuseStepNode</a>
 </li>
 <li>FuseStep()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1FuseStep.html#a345e30fc54e9782faa9b8744c9ed5d14">tvm::auto_scheduler::FuseStep</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1FuseStep.html#a77c478295e170275d7d5da7345a03546">tvm::auto_scheduler::FuseStep</a>
 </li>
 <li>FVisitAttrs
 : <a class="el" href="classtvm_1_1ReflectionVTable.html#a486eb682af89ac025c0db1f8f6045b95">tvm::ReflectionVTable</a>
diff --git a/docs/reference/api/doxygen/functions_func_f.html b/docs/reference/api/doxygen/functions_func_f.html
index f3bc18051..c7b76c8b3 100644
--- a/docs/reference/api/doxygen/functions_func_f.html
+++ b/docs/reference/api/doxygen/functions_func_f.html
@@ -199,6 +199,9 @@ $(function() {
 <li>FromExprInContext()
 : <a class="el" href="classtvm_1_1IRModule.html#a1cc91fc2b2adaca5a4dcfc14baf28c27">tvm::IRModule</a>
 </li>
+<li>FromExternalDLTensor()
+: <a class="el" href="classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e">tvm::runtime::NDArray</a>
+</li>
 <li>FromFunc()
 : <a class="el" href="classtvm_1_1tir_1_1IndexMap.html#afa04f25f10b1dac139df9a1b34598cbb">tvm::tir::IndexMap</a>
 </li>
@@ -260,7 +263,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1te_1_1Fuse.html#a10b77eec10eb7dbc536b0c8d65163f9c">tvm::te::Fuse</a>
 </li>
 <li>fuse()
-: <a class="el" href="classtvm_1_1te_1_1Stage.html#a07b721494aa3c0c79e8a8654c433708f">tvm::te::Stage</a>
+: <a class="el" href="classtvm_1_1te_1_1Stage.html#a5658065d9cbbee620bbd107d30c4ae72">tvm::te::Stage</a>
 </li>
 <li>Fuse()
 : <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a4e60246388be76fe6a1f46c4255da3ff">tvm::tir::ScheduleNode</a>
diff --git a/docs/reference/api/doxygen/functions_func_n.html b/docs/reference/api/doxygen/functions_func_n.html
index 28bdabbb9..e2d2c11bd 100644
--- a/docs/reference/api/doxygen/functions_func_n.html
+++ b/docs/reference/api/doxygen/functions_func_n.html
@@ -92,6 +92,9 @@ $(function() {
 : <a class="el" href="classtvm_1_1runtime_1_1SimpleObjAllocator_1_1ArrayHandler.html#a310471cff82c5d0836f65ec7f199e621">tvm::runtime::SimpleObjAllocator::ArrayHandler&lt; ArrayType, ElemType &gt;</a>
 , <a class="el" href="classtvm_1_1runtime_1_1SimpleObjAllocator_1_1Handler.html#afedd0ba3dc8dc82c7566bb9120a7c56d">tvm::runtime::SimpleObjAllocator::Handler&lt; T &gt;</a>
 </li>
+<li>NewFromDLTensor()
+: <a class="el" href="classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2">tvm::runtime::NDArray</a>
+</li>
 <li>NextTaskId()
 : <a class="el" href="classtvm_1_1meta__schedule_1_1PyTaskSchedulerNode.html#a23752f62706ef3f0bfac98fb203e5062">tvm::meta_schedule::PyTaskSchedulerNode</a>
 , <a class="el" href="classtvm_1_1meta__schedule_1_1TaskSchedulerNode.html#a079e2964ca86b5c32564140efa3e5626">tvm::meta_schedule::TaskSchedulerNode</a>
diff --git a/docs/reference/api/doxygen/functions_func_s.html b/docs/reference/api/doxygen/functions_func_s.html
index c20507ffa..9ad9dc079 100644
--- a/docs/reference/api/doxygen/functions_func_s.html
+++ b/docs/reference/api/doxygen/functions_func_s.html
@@ -700,7 +700,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1runtime_1_1DeviceAPI.html#ac29b9295c432a87658392872c644864f">tvm::runtime::DeviceAPI</a>
 </li>
 <li>String()
-: <a class="el" href="classtvm_1_1runtime_1_1String.html#acf549b3c43142639879e0fc31ea5cd77">tvm::runtime::String</a>
+: <a class="el" href="classtvm_1_1runtime_1_1String.html#ac5d930b522e9fef9c07e51819d96d2f3">tvm::runtime::String</a>
 </li>
 <li>StringImm()
 : <a class="el" href="classtvm_1_1tir_1_1StringImm.html#a0f2830290e055f677c5d5dea98aab726">tvm::tir::StringImm</a>
diff --git a/docs/reference/api/doxygen/functions_func_t.html b/docs/reference/api/doxygen/functions_func_t.html
index 067ed566c..07b41b291 100644
--- a/docs/reference/api/doxygen/functions_func_t.html
+++ b/docs/reference/api/doxygen/functions_func_t.html
@@ -981,7 +981,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1runtime_1_1TVMArgsSetter.html#a5882f7eda112e825eb5a87e45aeb85b0">tvm::runtime::TVMArgsSetter</a>
 </li>
 <li>TVMArgValue()
-: <a class="el" href="classtvm_1_1runtime_1_1TVMArgValue.html#a5fbd71750e5bbba6edc9094178af9276">tvm::runtime::TVMArgValue</a>
+: <a class="el" href="classtvm_1_1runtime_1_1TVMArgValue.html#a987b2fb283cea5484d4655e3f711c046">tvm::runtime::TVMArgValue</a>
 </li>
 <li>TVMMovableArgValue_()
 : <a class="el" href="classtvm_1_1runtime_1_1TVMMovableArgValue__.html#a8eca9048535541f374a5806f9648131b">tvm::runtime::TVMMovableArgValue_</a>
@@ -1022,7 +1022,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1TypedEnvFunc_3_01R_07Args_8_8_8_08_4.html#a0d72a6fa7263821c14bcd37837998ed9">tvm::TypedEnvFunc&lt; R(Args...)&gt;</a>
 </li>
 <li>TypedPackedFunc()
-: <a class="el" href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#af45a2ceff92e6f6c394ea766a45027a0">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;</a>
+: <a class="el" href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#a36ca0d1876544463ee848766e70e5e96">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;</a>
 </li>
 <li>TypeIndex2Key()
 : <a class="el" href="classtvm_1_1runtime_1_1Object.html#a817ba6c23b7ee1821c48a75edf255a30">tvm::runtime::Object</a>
@@ -1045,7 +1045,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1TypeRelation.html#ac26b1897eab8197ed26606ab81b7403b">tvm::TypeRelation</a>
 </li>
 <li>TypeReporter()
-: <a class="el" href="classtvm_1_1TypeReporter.html#aa3dc38a3c84d324d0b3a9f358460a091">tvm::TypeReporter</a>
+: <a class="el" href="classtvm_1_1TypeReporter.html#a8e7e05a07f9f7ad9bea91f27afac9051">tvm::TypeReporter</a>
 </li>
 <li>TypeVar()
 : <a class="el" href="classtvm_1_1TypeVar.html#adf5ef8e89d162735519b5d125c89e3e3">tvm::TypeVar</a>
diff --git a/docs/reference/api/doxygen/functions_n.html b/docs/reference/api/doxygen/functions_n.html
index 2147f7259..637e3ac2c 100644
--- a/docs/reference/api/doxygen/functions_n.html
+++ b/docs/reference/api/doxygen/functions_n.html
@@ -143,6 +143,9 @@ $(function() {
 : <a class="el" href="classtvm_1_1runtime_1_1SimpleObjAllocator_1_1ArrayHandler.html#a310471cff82c5d0836f65ec7f199e621">tvm::runtime::SimpleObjAllocator::ArrayHandler&lt; ArrayType, ElemType &gt;</a>
 , <a class="el" href="classtvm_1_1runtime_1_1SimpleObjAllocator_1_1Handler.html#afedd0ba3dc8dc82c7566bb9120a7c56d">tvm::runtime::SimpleObjAllocator::Handler&lt; T &gt;</a>
 </li>
+<li>NewFromDLTensor()
+: <a class="el" href="classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2">tvm::runtime::NDArray</a>
+</li>
 <li>newshape
 : <a class="el" href="structtvm_1_1relay_1_1ReshapeAttrs.html#a9bca32c3acff2ed8fd6bc63a50f82051">tvm::relay::ReshapeAttrs</a>
 , <a class="el" href="structtvm_1_1relay_1_1ReshapeTensorAttrs.html#aaacd1ab5124b54316a9e1f3ef5a5ec3c">tvm::relay::ReshapeTensorAttrs</a>
diff --git a/docs/reference/api/doxygen/functions_s.html b/docs/reference/api/doxygen/functions_s.html
index f8e02414f..87255dbb3 100644
--- a/docs/reference/api/doxygen/functions_s.html
+++ b/docs/reference/api/doxygen/functions_s.html
@@ -782,7 +782,7 @@ $(function() {
 , <a class="el" href="classtvm_1_1SpanNode.html#ad573167f93facbfbee19983b08bbba3d">tvm::SpanNode</a>
 </li>
 <li>SourceMap()
-: <a class="el" href="classtvm_1_1parser_1_1SourceMap.html#a43518e78ad2060e9400d893078c48008">tvm::parser::SourceMap</a>
+: <a class="el" href="classtvm_1_1parser_1_1SourceMap.html#afc48463cc0967ab79876178613a5aff2">tvm::parser::SourceMap</a>
 </li>
 <li>space_generator
 : <a class="el" href="classtvm_1_1meta__schedule_1_1TuneContextNode.html#a7bdfdd48530bfe380c5f6c143158a07f">tvm::meta_schedule::TuneContextNode</a>
@@ -877,8 +877,8 @@ $(function() {
 : <a class="el" href="structtvm_1_1relay_1_1DeviceCopyAttrs.html#aac5b2c76325a587bbefaa5af87b4138f">tvm::relay::DeviceCopyAttrs</a>
 </li>
 <li>Stage()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#af0643fe8c1298451c9a322f915c48843">tvm::auto_scheduler::Stage</a>
-, <a class="el" href="classtvm_1_1te_1_1Stage.html#a1ecdc9a000be62c9cc26a96d4c33e36e">tvm::te::Stage</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#a39ffbb1b4e189180bc4067e74965f42b">tvm::auto_scheduler::Stage</a>
+, <a class="el" href="classtvm_1_1te_1_1Stage.html#aa6ace38b6312e42aaf9389c8749ae0a4">tvm::te::Stage</a>
 </li>
 <li>stage_id
 : <a class="el" href="classtvm_1_1auto__scheduler_1_1StepNode.html#afcc7aaf263348f66139307affbfcee09">tvm::auto_scheduler::StepNode</a>
@@ -991,7 +991,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a93d1d23f24d903db844f75f51fe09a36">tvm::tir::ScheduleNode</a>
 </li>
 <li>StorageAlignStep()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1StorageAlignStep.html#a99dbb8c55d9e7d78268b6d43fd348bc7">tvm::auto_scheduler::StorageAlignStep</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1StorageAlignStep.html#af50b7c2f020f8e0a80f5bcc8e559b394">tvm::auto_scheduler::StorageAlignStep</a>
 </li>
 <li>StorageType
 : <a class="el" href="classtvm_1_1runtime_1_1SimpleObjAllocator_1_1ArrayHandler.html#a67e86db3290b1d3bd4aca7e7a2faf187">tvm::runtime::SimpleObjAllocator::ArrayHandler&lt; ArrayType, ElemType &gt;</a>
@@ -1046,7 +1046,7 @@ $(function() {
 , <a class="el" href="classtvm_1_1tir_1_1BufferNode.html#ac18ddd10b79a30ae57d3a8283686259d">tvm::tir::BufferNode</a>
 </li>
 <li>String()
-: <a class="el" href="classtvm_1_1runtime_1_1String.html#ac5d930b522e9fef9c07e51819d96d2f3">tvm::runtime::String</a>
+: <a class="el" href="classtvm_1_1runtime_1_1String.html#acf549b3c43142639879e0fc31ea5cd77">tvm::runtime::String</a>
 , <a class="el" href="classtvm_1_1runtime_1_1StringObj_1_1FromStd.html#a7fb804f7dc96dd9f705c84095f37f1ca">tvm::runtime::StringObj::FromStd</a>
 , <a class="el" href="classtvm_1_1runtime_1_1StringObj.html#a7fb804f7dc96dd9f705c84095f37f1ca">tvm::runtime::StringObj</a>
 </li>
diff --git a/docs/reference/api/doxygen/functions_t.html b/docs/reference/api/doxygen/functions_t.html
index 16067442a..baeab243d 100644
--- a/docs/reference/api/doxygen/functions_t.html
+++ b/docs/reference/api/doxygen/functions_t.html
@@ -78,7 +78,7 @@ $(function() {
 , <a class="el" href="structtvm_1_1runtime_1_1vm_1_1Instruction.html#a46879dbe84105fb621a6167f8d73b223">tvm::runtime::vm::Instruction</a>
 </li>
 <li>Target()
-: <a class="el" href="classtvm_1_1Target.html#ab825b350cf478bf948d807b6fdf636a0">tvm::Target</a>
+: <a class="el" href="classtvm_1_1Target.html#a77f3d7cc97d8cfd7172af58b4e784d89">tvm::Target</a>
 </li>
 <li>target
 : <a class="el" href="classtvm_1_1VirtualDeviceNode.html#a8b2d427d9e21886ccaeaae5e9cc55aaf">tvm::VirtualDeviceNode</a>
@@ -1263,7 +1263,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1TypedEnvFunc_3_01R_07Args_8_8_8_08_4.html#a41a6b9014d0feeb628ca7edfd0d26f0b">tvm::TypedEnvFunc&lt; R(Args...)&gt;</a>
 </li>
 <li>TypedPackedFunc()
-: <a class="el" href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#a6b346a6d0b601eff5a100c7a207e9c86">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;</a>
+: <a class="el" href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#a0161d426f9ca366c860ad48c384f7192">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;</a>
 </li>
 <li>TypeIndex2Key()
 : <a class="el" href="classtvm_1_1runtime_1_1Object.html#a817ba6c23b7ee1821c48a75edf255a30">tvm::runtime::Object</a>
diff --git a/docs/reference/api/doxygen/functions_v.html b/docs/reference/api/doxygen/functions_v.html
index 6b588d83f..dcbd6760d 100644
--- a/docs/reference/api/doxygen/functions_v.html
+++ b/docs/reference/api/doxygen/functions_v.html
@@ -566,8 +566,8 @@ $(function() {
 </li>
 <li>VisitType_()
 : <a class="el" href="classtvm_1_1TypeFunctor_3_01R_07const_01Type_01_6n_00_01Args_8_8_8_08_4.html#a949688cb9b4a8fa16ddc3e5fbbf13580">tvm::TypeFunctor&lt; R(const Type &amp;n, Args...)&gt;</a>
-, <a class="el" href="classtvm_1_1TypeMutator.html#a11e7e2f91d7dd05bee32aee6260eb459">tvm::TypeMutator</a>
-, <a class="el" href="classtvm_1_1TypeVisitor.html#a11378b4db6f704c04a97bec1c8ea8261">tvm::TypeVisitor</a>
+, <a class="el" href="classtvm_1_1TypeMutator.html#a18a04668d3fb464d957f3a26a4274104">tvm::TypeMutator</a>
+, <a class="el" href="classtvm_1_1TypeVisitor.html#a82c83b1524502579f56d194138badd3e">tvm::TypeVisitor</a>
 </li>
 <li>VisitTypeDefault_()
 : <a class="el" href="classtvm_1_1TypeFunctor_3_01R_07const_01Type_01_6n_00_01Args_8_8_8_08_4.html#a91553f9e04c39b3821a70ae4f7b0c597">tvm::TypeFunctor&lt; R(const Type &amp;n, Args...)&gt;</a>
diff --git a/docs/reference/api/doxygen/ndarray_8h_source.html b/docs/reference/api/doxygen/ndarray_8h_source.html
index 72d1d8b9e..856057d40 100644
--- a/docs/reference/api/doxygen/ndarray_8h_source.html
+++ b/docs/reference/api/doxygen/ndarray_8h_source.html
@@ -66,17 +66,17 @@ $(function() {
 <div class="title">ndarray.h</div>  </div>
 </div><!--header-->
 <div class="contents">
-<a href="ndarray_8h.html">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="comment">/*</span></div><div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;<span class="comment"> * Licensed to the Apache Software Foundation (ASF) under one</span></div><div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="comment"> * or more con [...]
+<a href="ndarray_8h.html">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno">    1</span>&#160;<span class="comment">/*</span></div><div class="line"><a name="l00002"></a><span class="lineno">    2</span>&#160;<span class="comment"> * Licensed to the Apache Software Foundation (ASF) under one</span></div><div class="line"><a name="l00003"></a><span class="lineno">    3</span>&#160;<span class="comment"> * or more con [...]
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ab7238434803d6a171318495fe46dc977"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ab7238434803d6a171318495fe46dc977">tvm::runtime::NDArray::ToDLPack</a></div><div class="ttdeci">DLManagedTensor * ToDLPack() const</div><div class="ttdoc">Create a reference view of NDArray that represents as DLManagedTensor. </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html">tvm::runtime::TVMRetValue</a></div><div class="ttdoc">Return Value container, Unlike TVMArgValue, which only holds reference and do not delete the underlyi...</div><div class="ttdef"><b>Definition:</b> packed_func.h:799</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ade0e2757904f4f5ba5c667ae01793a47"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ade0e2757904f4f5ba5c667ae01793a47">tvm::runtime::NDArray::FFIDecRef</a></div><div class="ttdeci">static void FFIDecRef(TVMArrayHandle handle)</div><div class="ttdoc">DecRef resource managed by an FFI array handle. </div><div class="ttdef"><b>Definition:</b> ndarray.h:396</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ade0e2757904f4f5ba5c667ae01793a47"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ade0e2757904f4f5ba5c667ae01793a47">tvm::runtime::NDArray::FFIDecRef</a></div><div class="ttdeci">static void FFIDecRef(TVMArrayHandle handle)</div><div class="ttdoc">DecRef resource managed by an FFI array handle. </div><div class="ttdef"><b>Definition:</b> ndarray.h:414</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a87d9cadd0c232324c101f7ed231aa193"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a87d9cadd0c232324c101f7ed231aa193">tvm::runtime::NDArray::DataType</a></div><div class="ttdeci">runtime::DataType DataType() const</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ObjectPtr_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1ObjectPtr.html">tvm::runtime::ObjectPtr</a></div><div class="ttdoc">A custom smart pointer for Object. </div><div class="ttdef"><b>Definition:</b> object.h:358</div></div>
 <div class="ttc" id="optional_8h_html"><div class="ttname"><a href="optional_8h.html">optional.h</a></div><div class="ttdoc">Runtime Optional container types. </div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad78792a1e1feb160b0be4474a4c13a4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad78792a1e1feb160b0be4474a4c13a4c">tvm::runtime::NDArray::Load</a></div><div class="ttdeci">bool Load(dmlc::Stream *stream)</div><div class="ttdoc">Load NDArray from stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:456</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad78792a1e1feb160b0be4474a4c13a4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad78792a1e1feb160b0be4474a4c13a4c">tvm::runtime::NDArray::Load</a></div><div class="ttdeci">bool Load(dmlc::Stream *stream)</div><div class="ttdoc">Load NDArray from stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:474</div></div>
 <div class="ttc" id="string_8h_html"><div class="ttname"><a href="string_8h.html">string.h</a></div><div class="ttdoc">Runtime String container types. </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMPODValue___html"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMPODValue__.html">tvm::runtime::TVMPODValue_</a></div><div class="ttdoc">Internal base class to handle conversion to POD values. </div><div class="ttdef"><b>Definition:</b> packed_func.h:541</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a1550151d3616e918d45e047840b81e1e"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a1550151d3616e918d45e047840b81e1e">tvm::runtime::NDArray::CopyFrom</a></div><div class="ttdeci">void CopyFrom(const DLTensor *other)</div><div class="ttdoc">Copy data content from another array. </div><div class="ttdef"><b>Definition:</b> ndarray.h:345</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a1550151d3616e918d45e047840b81e1e"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a1550151d3616e918d45e047840b81e1e">tvm::runtime::NDArray::CopyFrom</a></div><div class="ttdeci">void CopyFrom(const DLTensor *other)</div><div class="ttdoc">Copy data content from another array. </div><div class="ttdef"><b>Definition:</b> ndarray.h:363</div></div>
 <div class="ttc" id="namespacetvm_html"><div class="ttname"><a href="namespacetvm.html">tvm</a></div><div class="ttdoc">runtime implementation for LibTorch/TorchScript. </div><div class="ttdef"><b>Definition:</b> analyzer.h:36</div></div>
 <div class="ttc" id="c__runtime__api_8h_html_a73ca58cb32f4a4adf71d274dc1e27be4"><div class="ttname"><a href="c__runtime__api_8h.html#a73ca58cb32f4a4adf71d274dc1e27be4">TVMArrayHandle</a></div><div class="ttdeci">DLTensor * TVMArrayHandle</div><div class="ttdoc">the array handle </div><div class="ttdef"><b>Definition:</b> c_runtime_api.h:138</div></div>
 <div class="ttc" id="serializer_8h_html"><div class="ttname"><a href="serializer_8h.html">serializer.h</a></div><div class="ttdoc">Serializer extension to support TVM data types Include this file to enable serialization of DLDataTyp...</div></div>
@@ -84,11 +84,11 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html_ac9e5eed7719e322117bde996a171e33a"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html#ac9e5eed7719e322117bde996a171e33a">tvm::runtime::Object::IncRef</a></div><div class="ttdeci">void IncRef()</div><div class="ttdoc">developer function, increases reference counter. </div><div class="ttdef"><b>Definition:</b> object.h:799</div></div>
 <div class="ttc" id="namespacestd_html"><div class="ttname"><a href="namespacestd.html">std</a></div><div class="ttdef"><b>Definition:</b> loop_state.h:456</div></div>
 <div class="ttc" id="c__runtime__api_8h_html_ae246eaa00342c042f3f194605ad9bc7a"><div class="ttname"><a href="c__runtime__api_8h.html#ae246eaa00342c042f3f194605ad9bc7a">TVMArrayCopyToBytes</a></div><div class="ttdeci">int TVMArrayCopyToBytes(TVMArrayHandle handle, void *data, size_t nbytes)</div><div class="ttdoc">Copy array data to CPU byte array. </div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a7b077581a4cd4b1d4f783b49ccac112b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a7b077581a4cd4b1d4f783b49ccac112b">tvm::runtime::NDArray::CopyTo</a></div><div class="ttdeci">void CopyTo(DLTensor *other) const</div><div class="ttdoc">Copy data content into another array. </div><div class="ttdef"><b>Definition:</b> ndarray.h:356</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a7b077581a4cd4b1d4f783b49ccac112b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a7b077581a4cd4b1d4f783b49ccac112b">tvm::runtime::NDArray::CopyTo</a></div><div class="ttdeci">void CopyTo(DLTensor *other) const</div><div class="ttdoc">Copy data content into another array. </div><div class="ttdef"><b>Definition:</b> ndarray.h:374</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a59f41733876e0a161de701de9fd60749"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a59f41733876e0a161de701de9fd60749">tvm::runtime::NDArray::Empty</a></div><div class="ttdeci">static NDArray Empty(ShapeTuple shape, DLDataType dtype, Device dev, Optional&lt; String &gt; mem_scope=NullOpt)</div><div class="ttdoc">Create an empty NDArray. </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a1c9d84d35af95d1c29fc5c9c2ced84c8"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a1c9d84d35af95d1c29fc5c9c2ced84c8">tvm::runtime::NDArray::IsContiguous</a></div><div class="ttdeci">bool IsContiguous() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:341</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_afdd7050eda88b079f0a962bd413a34ea"><div class="ttname"><a href="namespacetvm_1_1runtime.html#afdd7050eda88b079f0a962bd413a34ea">tvm::runtime::TVMArrayHandleToObjectHandle</a></div><div class="ttdeci">Object * TVMArrayHandleToObjectHandle(TVMArrayHandle handle)</div><div class="ttdef"><b>Definition:</b> ndarray.h:400</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a1c9d84d35af95d1c29fc5c9c2ced84c8"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a1c9d84d35af95d1c29fc5c9c2ced84c8">tvm::runtime::NDArray::IsContiguous</a></div><div class="ttdeci">bool IsContiguous() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:359</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_afdd7050eda88b079f0a962bd413a34ea"><div class="ttname"><a href="namespacetvm_1_1runtime.html#afdd7050eda88b079f0a962bd413a34ea">tvm::runtime::TVMArrayHandleToObjectHandle</a></div><div class="ttdeci">Object * TVMArrayHandleToObjectHandle(TVMArrayHandle handle)</div><div class="ttdef"><b>Definition:</b> ndarray.h:418</div></div>
 <div class="ttc" id="c__runtime__api_8h_html_a775383bcd8c0237e36bdf0c9654d62c3"><div class="ttname"><a href="c__runtime__api_8h.html#a775383bcd8c0237e36bdf0c9654d62c3">TVMGetLastError</a></div><div class="ttdeci">const char * TVMGetLastError(void)</div><div class="ttdoc">return str message of the last error all function in this file will return 0 when success and nonzero...</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_af2a8ccab95d432d1ecad7a389e11bcd3"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#af2a8ccab95d432d1ecad7a389e11bcd3">tvm::runtime::NDArray::reset</a></div><div class="ttdeci">void reset()</div><div class="ttdoc">reset the content of NDArray to be nullptr </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></div><div class="ttdoc">Managed NDArray. The array is backed by reference counted blocks. </div><div class="ttdef"><b>Definition:</b> ndarray.h:59</div></div>
@@ -96,37 +96,39 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html_a70fb5361147634605d6595bb89381f03"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html#a70fb5361147634605d6595bb89381f03">tvm::runtime::Object::DecRef</a></div><div class="ttdeci">void DecRef()</div><div class="ttdoc">developer function, decrease reference counter. </div><div class="ttdef"><b>Definition:</b> object.h:801</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html_ad94d79729ac85aa7c976e23d39066383"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html#ad94d79729ac85aa7c976e23d39066383">tvm::runtime::Object::RuntimeTypeIndex</a></div><div class="ttdeci">static uint32_t RuntimeTypeIndex()</div><div class="ttdef"><b>Definition:</b> object.h:225</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ae2a878fb8c847666d2318b979714cefa"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ae2a878fb8c847666d2318b979714cefa">tvm::runtime::NDArray::CopyFromTo</a></div><div class="ttdeci">static void CopyFromTo(const DLTensor *from, DLTensor *to, TVMStreamHandle stream=nullptr)</div><div class="ttdoc">Function to copy data from one array to another. </div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_aa1e7d2346052e198b409966eb67be92b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#aa1e7d2346052e198b409966eb67be92b">tvm::runtime::NDArray::use_count</a></div><div class="ttdeci">int use_count() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:375</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ae6f82ad564a648d21e9a2e4d0ff07b39"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ae6f82ad564a648d21e9a2e4d0ff07b39">tvm::runtime::NDArray::Save</a></div><div class="ttdeci">void Save(dmlc::Stream *stream) const</div><div class="ttdoc">Save NDArray to stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:454</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad57933f49a9fd51d7f996e1b16ffd2e0"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad57933f49a9fd51d7f996e1b16ffd2e0">tvm::runtime::NDArray::operator-&gt;</a></div><div class="ttdeci">const DLTensor * operator-&gt;() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:377</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ab76ba9c638e1d6db8d6f0ba8c4d38670"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ab76ba9c638e1d6db8d6f0ba8c4d38670">tvm::runtime::NDArray::FFIDataFromHandle</a></div><div class="ttdeci">static ObjectPtr&lt; Object &gt; FFIDataFromHandle(TVMArrayHandle handle)</div><div class="ttdoc">Construct NDArray&amp;#39;s Data field from array handle in FFI. </div><div class="ttdef"><b>Definition:</b> ndarray.h:383</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_aa1e7d2346052e198b409966eb67be92b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#aa1e7d2346052e198b409966eb67be92b">tvm::runtime::NDArray::use_count</a></div><div class="ttdeci">int use_count() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:393</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ae6f82ad564a648d21e9a2e4d0ff07b39"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ae6f82ad564a648d21e9a2e4d0ff07b39">tvm::runtime::NDArray::Save</a></div><div class="ttdeci">void Save(dmlc::Stream *stream) const</div><div class="ttdoc">Save NDArray to stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:472</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad57933f49a9fd51d7f996e1b16ffd2e0"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad57933f49a9fd51d7f996e1b16ffd2e0">tvm::runtime::NDArray::operator-&gt;</a></div><div class="ttdeci">const DLTensor * operator-&gt;() const</div><div class="ttdef"><b>Definition:</b> ndarray.h:395</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ab76ba9c638e1d6db8d6f0ba8c4d38670"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ab76ba9c638e1d6db8d6f0ba8c4d38670">tvm::runtime::NDArray::FFIDataFromHandle</a></div><div class="ttdeci">static ObjectPtr&lt; Object &gt; FFIDataFromHandle(TVMArrayHandle handle)</div><div class="ttdoc">Construct NDArray&amp;#39;s Data field from array handle in FFI. </div><div class="ttdef"><b>Definition:</b> ndarray.h:401</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_af5801a105ceb450616a83d19c5c92326"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#af5801a105ceb450616a83d19c5c92326">tvm::runtime::NDArray::NDArray</a></div><div class="ttdeci">NDArray(ObjectPtr&lt; Object &gt; data)</div><div class="ttdoc">constructor. </div><div class="ttdef"><b>Definition:</b> ndarray.h:73</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1DataType_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1DataType.html">tvm::runtime::DataType</a></div><div class="ttdoc">Runtime primitive data type. </div><div class="ttdef"><b>Definition:</b> data_type.h:41</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad273c7bc59b73fb026fd64fc764cbebc"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad273c7bc59b73fb026fd64fc764cbebc">tvm::runtime::NDArray::Shape</a></div><div class="ttdeci">ShapeTuple Shape() const</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad5d21a1d7a704bfc9504d28910748d39"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad5d21a1d7a704bfc9504d28910748d39">tvm::runtime::NDArray::get_mutable</a></div><div class="ttdeci">Container * get_mutable() const</div><div class="ttdoc">Get mutable internal container pointer. </div><div class="ttdef"><b>Definition:</b> ndarray.h:379</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:261</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a5f3c42a1a5d71d914d3ca326553e4f79"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a5f3c42a1a5d71d914d3ca326553e4f79">tvm::runtime::NDArray::Container::Container</a></div><div class="ttdeci">Container(void *data, ShapeTuple shape, DLDataType dtype, Device dev)</div><div class="ttdef"><b>Definition:</b> ndarray.h:274</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_adb2ed1227b418f5846d43d3234b52391"><div class="ttname"><a href="namespacetvm_1_1runtime.html#adb2ed1227b418f5846d43d3234b52391">tvm::runtime::IsContiguous</a></div><div class="ttdeci">bool IsContiguous(const DLTensor &amp;arr)</div><div class="ttdoc">check if a DLTensor is contiguous. </div><div class="ttdef"><b>Definition:</b> ndarray.h:330</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ad5d21a1d7a704bfc9504d28910748d39"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ad5d21a1d7a704bfc9504d28910748d39">tvm::runtime::NDArray::get_mutable</a></div><div class="ttdeci">Container * get_mutable() const</div><div class="ttdoc">Get mutable internal container pointer. </div><div class="ttdef"><b>Definition:</b> ndarray.h:397</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:279</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a5f3c42a1a5d71d914d3ca326553e4f79"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a5f3c42a1a5d71d914d3ca326553e4f79">tvm::runtime::NDArray::Container::Container</a></div><div class="ttdeci">Container(void *data, ShapeTuple shape, DLDataType dtype, Device dev)</div><div class="ttdef"><b>Definition:</b> ndarray.h:292</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_adb2ed1227b418f5846d43d3234b52391"><div class="ttname"><a href="namespacetvm_1_1runtime.html#adb2ed1227b418f5846d43d3234b52391">tvm::runtime::IsContiguous</a></div><div class="ttdeci">bool IsContiguous(const DLTensor &amp;arr)</div><div class="ttdoc">check if a DLTensor is contiguous. </div><div class="ttdef"><b>Definition:</b> ndarray.h:348</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ObjectRef_html_ac261cdb80487fb29ac42b28678f8cbef"><div class="ttname"><a href="classtvm_1_1runtime_1_1ObjectRef.html#ac261cdb80487fb29ac42b28678f8cbef">tvm::runtime::ObjectRef::data_</a></div><div class="ttdeci">ObjectPtr&lt; Object &gt; data_</div><div class="ttdoc">Internal pointer that backs the reference. </div><div class="ttdef"><b>Definition:</b> object.h:574</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase_html_a1063a9d01075d5b7b0e8fa31d4d72e0b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase.html#a1063a9d01075d5b7b0e8fa31d4d72e0b">tvm::runtime::NDArray::ContainerBase::dl_tensor</a></div><div class="ttdeci">DLTensor dl_tensor</div><div class="ttdoc">The corresponding dl_tensor field. </div><div class="ttdef"><b>Definition:</b> ndarray.h:239</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a56109cfc826b26172f084c3790144351"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a56109cfc826b26172f084c3790144351">tvm::runtime::NDArray::Container::SetDeleter</a></div><div class="ttdeci">void SetDeleter(FDeleter deleter)</div><div class="ttdoc">Set the deleter field. </div><div class="ttdef"><b>Definition:</b> ndarray.h:290</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase_html_a1063a9d01075d5b7b0e8fa31d4d72e0b"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase.html#a1063a9d01075d5b7b0e8fa31d4d72e0b">tvm::runtime::NDArray::ContainerBase::dl_tensor</a></div><div class="ttdeci">DLTensor dl_tensor</div><div class="ttdoc">The corresponding dl_tensor field. </div><div class="ttdef"><b>Definition:</b> ndarray.h:257</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a56109cfc826b26172f084c3790144351"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a56109cfc826b26172f084c3790144351">tvm::runtime::NDArray::Container::SetDeleter</a></div><div class="ttdeci">void SetDeleter(FDeleter deleter)</div><div class="ttdoc">Set the deleter field. </div><div class="ttdef"><b>Definition:</b> ndarray.h:308</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_afb6060bb96dad082c1deca26e6b58ae2"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2">tvm::runtime::NDArray::NewFromDLTensor</a></div><div class="ttdeci">static NDArray NewFromDLTensor(DLTensor *dl_tensor, Device dev)</div><div class="ttdoc">Create new NDArray, data is copied from DLTensor. </div></div>
 <div class="ttc" id="namespacetvm_1_1topi_html_af30c02f3a3f37c7963b3af60fb9c72a1"><div class="ttname"><a href="namespacetvm_1_1topi.html#af30c02f3a3f37c7963b3af60fb9c72a1">tvm::topi::shape</a></div><div class="ttdeci">Tensor shape(const Tensor &amp;src, DataType dtype, const std::string name=&quot;T_shape&quot;, const std::string tag=kInjective)</div><div class="ttdoc">Get the shape of input tensor. </div><div class="ttdef"><b>Definition:</b> transform.h:1700</div></div>
 <div class="ttc" id="c__runtime__api_8h_html_ace8007daffd9f2c6d954c24d870bfcc4"><div class="ttname"><a href="c__runtime__api_8h.html#ace8007daffd9f2c6d954c24d870bfcc4">tvm_index_t</a></div><div class="ttdeci">int64_t tvm_index_t</div><div class="ttdoc">type of array index. </div><div class="ttdef"><b>Definition:</b> c_runtime_api.h:81</div></div>
 <div class="ttc" id="namespacetvm_html_a7c2095aed90b2129ba631b90103313a2"><div class="ttname"><a href="namespacetvm.html#a7c2095aed90b2129ba631b90103313a2">tvm::Device</a></div><div class="ttdeci">DLDevice Device</div><div class="ttdef"><b>Definition:</b> ndarray.h:43</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ObjectRef_html_aadbc0886ffa80162ff31eefd0431ba09"><div class="ttname"><a href="classtvm_1_1runtime_1_1ObjectRef.html#aadbc0886ffa80162ff31eefd0431ba09">tvm::runtime::ObjectRef::get</a></div><div class="ttdeci">const Object * get() const</div><div class="ttdef"><b>Definition:</b> object.h:546</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ObjectRef_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></div><div class="ttdoc">Base class of all object reference. </div><div class="ttdef"><b>Definition:</b> object.h:511</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a356d1886b24da68c35a0d0b826c9359e"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e">tvm::runtime::NDArray::FromExternalDLTensor</a></div><div class="ttdeci">static NDArray FromExternalDLTensor(const DLTensor &amp;dl_tensor)</div><div class="ttdoc">Create a NDArray backed by an external DLTensor. </div></div>
 <div class="ttc" id="shape__tuple_8h_html"><div class="ttname"><a href="shape__tuple_8h.html">shape_tuple.h</a></div><div class="ttdoc">Runtime ShapeTuple container types. </div></div>
 <div class="ttc" id="object_8h_html"><div class="ttname"><a href="object_8h.html">object.h</a></div><div class="ttdoc">A managed object in the TVM runtime. </div></div>
 <div class="ttc" id="data__type_8h_html"><div class="ttname"><a href="data__type_8h.html">data_type.h</a></div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a141e032d848c60f8261046304bdc8c4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a141e032d848c60f8261046304bdc8c4c">tvm::runtime::NDArray::FFIGetHandle</a></div><div class="ttdeci">static TVMArrayHandle FFIGetHandle(const ObjectRef &amp;nd)</div><div class="ttdoc">Get FFI Array handle from ndarray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:388</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a39b39ce5a2a658b44944381f1835404a"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a39b39ce5a2a658b44944381f1835404a">tvm::runtime::NDArray::Container::Container</a></div><div class="ttdeci">Container()</div><div class="ttdoc">default constructor </div><div class="ttdef"><b>Definition:</b> ndarray.h:264</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_acf4599f17bfe79ae1fe8afc1af053b43"><div class="ttname"><a href="namespacetvm_1_1runtime.html#acf4599f17bfe79ae1fe8afc1af053b43">tvm::runtime::kTVMNDArrayMagic</a></div><div class="ttdeci">constexpr uint64_t kTVMNDArrayMagic</div><div class="ttdoc">Magic number for NDArray file. </div><div class="ttdef"><b>Definition:</b> ndarray.h:405</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a141e032d848c60f8261046304bdc8c4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a141e032d848c60f8261046304bdc8c4c">tvm::runtime::NDArray::FFIGetHandle</a></div><div class="ttdeci">static TVMArrayHandle FFIGetHandle(const ObjectRef &amp;nd)</div><div class="ttdoc">Get FFI Array handle from ndarray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:406</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html_a39b39ce5a2a658b44944381f1835404a"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html#a39b39ce5a2a658b44944381f1835404a">tvm::runtime::NDArray::Container::Container</a></div><div class="ttdeci">Container()</div><div class="ttdoc">default constructor </div><div class="ttdef"><b>Definition:</b> ndarray.h:282</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_acf4599f17bfe79ae1fe8afc1af053b43"><div class="ttname"><a href="namespacetvm_1_1runtime.html#acf4599f17bfe79ae1fe8afc1af053b43">tvm::runtime::kTVMNDArrayMagic</a></div><div class="ttdeci">constexpr uint64_t kTVMNDArrayMagic</div><div class="ttdoc">Magic number for NDArray file. </div><div class="ttdef"><b>Definition:</b> ndarray.h:423</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a4bbb80e8e36317829dd63e7f44ffbb0f"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a4bbb80e8e36317829dd63e7f44ffbb0f">tvm::runtime::NDArray::NDArray</a></div><div class="ttdeci">NDArray()</div><div class="ttdoc">default constructor </div><div class="ttdef"><b>Definition:</b> ndarray.h:68</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ShapeTuple_html_aea4f2fe4a3b8d36cb88193298f346228"><div class="ttname"><a href="classtvm_1_1runtime_1_1ShapeTuple.html#aea4f2fe4a3b8d36cb88193298f346228">tvm::runtime::ShapeTuple::index_type</a></div><div class="ttdeci">ShapeTupleObj::index_type index_type</div><div class="ttdoc">The type of shape index element. </div><div class="ttdef"><b>Definition:</b> shape_tuple.h:84</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_af4d489a1208be9cc4248b592769bccf2"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#af4d489a1208be9cc4248b592769bccf2">tvm::runtime::NDArray::CopyFromBytes</a></div><div class="ttdeci">void CopyFromBytes(const void *data, size_t nbytes)</div><div class="ttdoc">Copy data content from a byte buffer. </div></div>
 <div class="ttc" id="namespacetvm_html_a0da40d3e210aa3b38a17982a7b7866b8"><div class="ttname"><a href="namespacetvm.html#a0da40d3e210aa3b38a17982a7b7866b8">tvm::ret</a></div><div class="ttdeci">PrimExpr ret(PrimExpr value, Span span=Span())</div><div class="ttdoc">Return the value. </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Optional_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Optional.html">tvm::runtime::Optional</a></div><div class="ttdoc">Optional container that to represent to a Nullable variant of T. </div><div class="ttdef"><b>Definition:</b> optional.h:51</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_a8fb37910dcd9bb6899e6a3a47f006514"><div class="ttname"><a href="namespacetvm_1_1runtime.html#a8fb37910dcd9bb6899e6a3a47f006514">tvm::runtime::SaveDLTensor</a></div><div class="ttdeci">bool SaveDLTensor(dmlc::Stream *strm, const DLTensor *tensor)</div><div class="ttdoc">Save a DLTensor to stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:407</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_a8fb37910dcd9bb6899e6a3a47f006514"><div class="ttname"><a href="namespacetvm_1_1runtime.html#a8fb37910dcd9bb6899e6a3a47f006514">tvm::runtime::SaveDLTensor</a></div><div class="ttdeci">bool SaveDLTensor(dmlc::Stream *strm, const DLTensor *tensor)</div><div class="ttdoc">Save a DLTensor to stream. </div><div class="ttdef"><b>Definition:</b> ndarray.h:425</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_abec485628a0ca451b668c42fd8fa691a"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#abec485628a0ca451b668c42fd8fa691a">tvm::runtime::NDArray::FromDLPack</a></div><div class="ttdeci">static NDArray FromDLPack(DLManagedTensor *tensor)</div><div class="ttdoc">Create a NDArray backed by a dlpack tensor. </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMArgsSetter_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMArgsSetter.html">tvm::runtime::TVMArgsSetter</a></div><div class="ttdef"><b>Definition:</b> packed_func.h:1514</div></div>
 <div class="ttc" id="namespacetvm_html_aae7034e3e41c18e7fb78ff32bfc6a318"><div class="ttname"><a href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">tvm::NullOpt</a></div><div class="ttdeci">constexpr runtime::NullOptType NullOpt</div><div class="ttdef"><b>Definition:</b> optional.h:160</div></div>
@@ -135,8 +137,8 @@ $(function() {
 <div class="ttc" id="object_8h_html_ac2b7418e9549512b5db0126cf2a716f1"><div class="ttname"><a href="object_8h.html#ac2b7418e9549512b5db0126cf2a716f1">TVM_DECLARE_BASE_OBJECT_INFO</a></div><div class="ttdeci">#define TVM_DECLARE_BASE_OBJECT_INFO(TypeName, ParentType)</div><div class="ttdoc">helper macro to declare a base object type that can be inherited. </div><div class="ttdef"><b>Definition:</b> object.h:648</div></div>
 <div class="ttc" id="c__runtime__api_8h_html"><div class="ttname"><a href="c__runtime__api_8h.html">c_runtime_api.h</a></div></div>
 <div class="ttc" id="namespacetvm_html_ab3c85920678b8ba5d925d386b66c0261"><div class="ttname"><a href="namespacetvm.html#ab3c85920678b8ba5d925d386b66c0261">tvm::kInvalidDeviceType</a></div><div class="ttdeci">constexpr DLDeviceType kInvalidDeviceType</div><div class="ttdef"><b>Definition:</b> ndarray.h:51</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase_html_aa5597a1760c9f8c9d1fd51584b1283fb"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase.html#aa5597a1760c9f8c9d1fd51584b1283fb">tvm::runtime::NDArray::ContainerBase::shape_</a></div><div class="ttdeci">ShapeTuple shape_</div><div class="ttdoc">The shape container, can be used used for shape data. </div><div class="ttdef"><b>Definition:</b> ndarray.h:254</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_a59940b6d63dd4c5175c0fe875047c1cf"><div class="ttname"><a href="namespacetvm_1_1runtime.html#a59940b6d63dd4c5175c0fe875047c1cf">tvm::runtime::GetDataSize</a></div><div class="ttdeci">size_t GetDataSize(const DLTensor &amp;arr)</div><div class="ttdoc">return the size of data the DLTensor hold, in term of number of bytes </div><div class="ttdef"><b>Definition:</b> ndarray.h:316</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase_html_aa5597a1760c9f8c9d1fd51584b1283fb"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1ContainerBase.html#aa5597a1760c9f8c9d1fd51584b1283fb">tvm::runtime::NDArray::ContainerBase::shape_</a></div><div class="ttdeci">ShapeTuple shape_</div><div class="ttdoc">The shape container, can be used used for shape data. </div><div class="ttdef"><b>Definition:</b> ndarray.h:272</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_a59940b6d63dd4c5175c0fe875047c1cf"><div class="ttname"><a href="namespacetvm_1_1runtime.html#a59940b6d63dd4c5175c0fe875047c1cf">tvm::runtime::GetDataSize</a></div><div class="ttdeci">size_t GetDataSize(const DLTensor &amp;arr)</div><div class="ttdoc">return the size of data the DLTensor hold, in term of number of bytes </div><div class="ttdef"><b>Definition:</b> ndarray.h:334</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1ShapeTuple_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1ShapeTuple.html">tvm::runtime::ShapeTuple</a></div><div class="ttdoc">Reference to shape tuple objects. </div><div class="ttdef"><b>Definition:</b> shape_tuple.h:81</div></div>
 <div class="ttc" id="structtvm_1_1runtime_1_1TypeIndex_html_aed93c7318efc8052201d4c404b21a40da48232c4de1fa5119f58c3ba3fc88334c"><div class="ttname"><a href="structtvm_1_1runtime_1_1TypeIndex.html#aed93c7318efc8052201d4c404b21a40da48232c4de1fa5119f58c3ba3fc88334c">tvm::runtime::TypeIndex::kRuntimeNDArray</a></div><div class="ttdoc">runtime::NDArray. </div><div class="ttdef"><b>Definition:</b> object.h:64</div></div>
 </div><!-- fragment --></div><!-- contents -->
diff --git a/docs/reference/api/doxygen/packed__func_8h_source.html b/docs/reference/api/doxygen/packed__func_8h_source.html
index 45529d6bb..793866bfe 100644
--- a/docs/reference/api/doxygen/packed__func_8h_source.html
+++ b/docs/reference/api/doxygen/packed__func_8h_source.html
@@ -75,7 +75,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1ArrayNode_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1ArrayNode.html">tvm::runtime::ArrayNode</a></div><div class="ttdoc">array node content in array </div><div class="ttdef"><b>Definition:</b> array.h:38</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html">tvm::runtime::TVMRetValue</a></div><div class="ttdoc">Return Value container, Unlike TVMArgValue, which only holds reference and do not delete the underlyi...</div><div class="ttdef"><b>Definition:</b> packed_func.h:799</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Module_html_a1233f7b896bb299ef07f9e41a4ffdc17"><div class="ttname"><a href="classtvm_1_1runtime_1_1Module.html#a1233f7b896bb299ef07f9e41a4ffdc17">tvm::runtime::Module::GetFunction</a></div><div class="ttdeci">PackedFunc GetFunction(const std::string &amp;name, bool query_imports=false)</div><div class="ttdoc">Get packed function from current module by name. </div><div class="ttdef"><b>Definition:</b> packed_func.h:1937</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ade0e2757904f4f5ba5c667ae01793a47"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ade0e2757904f4f5ba5c667ae01793a47">tvm::runtime::NDArray::FFIDecRef</a></div><div class="ttdeci">static void FFIDecRef(TVMArrayHandle handle)</div><div class="ttdoc">DecRef resource managed by an FFI array handle. </div><div class="ttdef"><b>Definition:</b> ndarray.h:396</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ade0e2757904f4f5ba5c667ae01793a47"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ade0e2757904f4f5ba5c667ae01793a47">tvm::runtime::NDArray::FFIDecRef</a></div><div class="ttdeci">static void FFIDecRef(TVMArrayHandle handle)</div><div class="ttdoc">DecRef resource managed by an FFI array handle. </div><div class="ttdef"><b>Definition:</b> ndarray.h:414</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html_ae47baae854e2ff66d0ef87178727d8f4"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html#ae47baae854e2ff66d0ef87178727d8f4">tvm::runtime::TVMRetValue::operator=</a></div><div class="ttdeci">TVMRetValue &amp; operator=(TVMMovableArgValue_ &amp;&amp;other)</div><div class="ttdef"><b>Definition:</b> packed_func.h:940</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMArgsSetter_html_a4c6dffcee3ea29b0b5862f1cc42c5c1c"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMArgsSetter.html#a4c6dffcee3ea29b0b5862f1cc42c5c1c">tvm::runtime::TVMArgsSetter::operator()</a></div><div class="ttdeci">TVM_ALWAYS_INLINE void operator()(size_t i, TObjectRef &amp;&amp;value) const</div><div class="ttdef"><b>Definition:</b> packed_func.h:1596</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html_a74ea2767d491c57cb9c71e26ee934344"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html#a74ea2767d491c57cb9c71e26ee934344">tvm::runtime::TVMRetValue::MoveToCHost</a></div><div class="ttdeci">void MoveToCHost(TVMValue *ret_value, int *ret_type_code)</div><div class="ttdoc">Move the value back to front-end via C API. This marks the current container as null. The managed resources are moved to the front-end. The fr [...]
@@ -113,7 +113,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
 <div class="ttc" id="c__runtime__api_8h_html_a190e81769e805cca153514137a66e793abe76912731a3c65bd7cdd9ab9a462c66"><div class="ttname"><a href="c__runtime__api_8h.html#a190e81769e805cca153514137a66e793abe76912731a3c65bd7cdd9ab9a462c66">kTVMObjectHandle</a></div><div class="ttdef"><b>Definition:</b> c_runtime_api.h:118</div></div>
 <div class="ttc" id="structTVMByteArray_html_ab124e3227a75e0e4d55452f675f4fde1"><div class="ttname"><a href="structTVMByteArray.html#ab124e3227a75e0e4d55452f675f4fde1">TVMByteArray::data</a></div><div class="ttdeci">const char * data</div><div class="ttdef"><b>Definition:</b> c_runtime_api.h:158</div></div>
-<div class="ttc" id="namespacetvm_1_1runtime_html_afdd7050eda88b079f0a962bd413a34ea"><div class="ttname"><a href="namespacetvm_1_1runtime.html#afdd7050eda88b079f0a962bd413a34ea">tvm::runtime::TVMArrayHandleToObjectHandle</a></div><div class="ttdeci">Object * TVMArrayHandleToObjectHandle(TVMArrayHandle handle)</div><div class="ttdef"><b>Definition:</b> ndarray.h:400</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_afdd7050eda88b079f0a962bd413a34ea"><div class="ttname"><a href="namespacetvm_1_1runtime.html#afdd7050eda88b079f0a962bd413a34ea">tvm::runtime::TVMArrayHandleToObjectHandle</a></div><div class="ttdeci">Object * TVMArrayHandleToObjectHandle(TVMArrayHandle handle)</div><div class="ttdef"><b>Definition:</b> ndarray.h:418</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMPODValue___html_a8ffdfcc7099faf19ee07a5c03ce06af8"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMPODValue__.html#a8ffdfcc7099faf19ee07a5c03ce06af8">tvm::runtime::TVMPODValue_::value_</a></div><div class="ttdeci">TVMValue value_</div><div class="ttdoc">The value. </div><div class="ttdef"><b>Definition:</b> packed_func.h:635</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html_a77455a8fe7d27b90a01a64f1cd28e9ec"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html#a77455a8fe7d27b90a01a64f1cd28e9ec">tvm::runtime::TVMRetValue::TVMRetValue</a></div><div class="ttdeci">TVMRetValue()</div><div class="ttdoc">default constructor </div><div class="ttdef"><b>Definition:</b> packed_func.h:802</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html">tvm::runtime::NDArray</a></div><div class="ttdoc">Managed NDArray. The array is backed by reference counted blocks. </div><div class="ttdef"><b>Definition:</b> ndarray.h:59</div></div>
@@ -138,7 +138,7 @@ $(function() {
 <div class="ttc" id="ndarray_8h_html"><div class="ttname"><a href="ndarray_8h.html">ndarray.h</a></div><div class="ttdoc">A device-independent managed NDArray abstraction. </div></div>
 <div class="ttc" id="namespacetvm_1_1runtime_html_a277f104e659f71cd8885744700016341"><div class="ttname"><a href="namespacetvm_1_1runtime.html#a277f104e659f71cd8885744700016341">tvm::runtime::String2DLDataType</a></div><div class="ttdeci">DLDataType String2DLDataType(std::string s)</div><div class="ttdoc">convert a string to TVM type. </div><div class="ttdef"><b>Definition:</b> data_type.h:339</div></div>
 <div class="ttc" id="structtvm_1_1runtime_1_1PackedFuncValueConverter_3_1_1tvm_1_1runtime_1_1String_01_4_html_a9ac48d52f86dc3718590acc119e88741"><div class="ttname"><a href="structtvm_1_1runtime_1_1PackedFuncValueConverter_3_1_1tvm_1_1runtime_1_1String_01_4.html#a9ac48d52f86dc3718590acc119e88741">tvm::runtime::PackedFuncValueConverter&lt;::tvm::runtime::String &gt;::From</a></div><div class="ttdeci">static String From(const TVMRetValue &amp;val)</div><div class="ttdef"><b>Definition:</b> [...]
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ab76ba9c638e1d6db8d6f0ba8c4d38670"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ab76ba9c638e1d6db8d6f0ba8c4d38670">tvm::runtime::NDArray::FFIDataFromHandle</a></div><div class="ttdeci">static ObjectPtr&lt; Object &gt; FFIDataFromHandle(TVMArrayHandle handle)</div><div class="ttdoc">Construct NDArray&amp;#39;s Data field from array handle in FFI. </div><div class="ttdef"><b>Definition:</b> ndarray.h:383</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_ab76ba9c638e1d6db8d6f0ba8c4d38670"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#ab76ba9c638e1d6db8d6f0ba8c4d38670">tvm::runtime::NDArray::FFIDataFromHandle</a></div><div class="ttdeci">static ObjectPtr&lt; Object &gt; FFIDataFromHandle(TVMArrayHandle handle)</div><div class="ttdoc">Construct NDArray&amp;#39;s Data field from array handle in FFI. </div><div class="ttdef"><b>Definition:</b> ndarray.h:401</div></div>
 <div class="ttc" id="structtvm_1_1runtime_1_1ObjectTypeChecker_html"><div class="ttname"><a href="structtvm_1_1runtime_1_1ObjectTypeChecker.html">tvm::runtime::ObjectTypeChecker</a></div><div class="ttdoc">Type traits for runtime type check during FFI conversion. </div><div class="ttdef"><b>Definition:</b> packed_func.h:430</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMPODValue___html_a67fdc41da2d772f61a924b8fa2d820e5"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMPODValue__.html#a67fdc41da2d772f61a924b8fa2d820e5">tvm::runtime::TVMPODValue_::IsObjectRef</a></div><div class="ttdeci">bool IsObjectRef() const</div><div class="ttdef"><b>Definition:</b> packed_func.h:1792</div></div>
 <div class="ttc" id="structtvm_1_1runtime_1_1PackedFuncValueConverter_3_01Optional_3_01T_01_4_01_4_html_a6748e04a16945df4c15edb53d0aaba70"><div class="ttname"><a href="structtvm_1_1runtime_1_1PackedFuncValueConverter_3_01Optional_3_01T_01_4_01_4.html#a6748e04a16945df4c15edb53d0aaba70">tvm::runtime::PackedFuncValueConverter&lt; Optional&lt; T &gt; &gt;::From</a></div><div class="ttdeci">static Optional&lt; T &gt; From(const TVMArgValue &amp;val)</div><div class="ttdef"><b>Definition:</b>  [...]
@@ -158,7 +158,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4_html_aa590b3e712e06867805b41aaf17019ed"><div class="ttname"><a href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#aa590b3e712e06867805b41aaf17019ed">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;::operator=</a></div><div class="ttdeci">TSelf &amp; operator=(PackedFunc packed)</div><div class="ttdoc">copy assignment operator from PackedFunc. </div><div class="ttdef"><b>Definition:< [...]
 <div class="ttc" id="structtvm_1_1runtime_1_1TypeIndex_html_aed93c7318efc8052201d4c404b21a40da23eb6ecfca76a965bf4727a4e931584b"><div class="ttname"><a href="structtvm_1_1runtime_1_1TypeIndex.html#aed93c7318efc8052201d4c404b21a40da23eb6ecfca76a965bf4727a4e931584b">tvm::runtime::TypeIndex::kRuntimePackedFunc</a></div><div class="ttdoc">runtime::PackedFunc. </div><div class="ttdef"><b>Definition:</b> object.h:74</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html_ad5446f5812132852387dca7335989e88"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html#ad5446f5812132852387dca7335989e88">tvm::runtime::TVMRetValue::operator=</a></div><div class="ttdeci">TVMRetValue &amp; operator=(Module m)</div><div class="ttdef"><b>Definition:</b> packed_func.h:920</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:261</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:279</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMRetValue_html_a4ab194932127e4b1c372e5e58e450721"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMRetValue.html#a4ab194932127e4b1c372e5e58e450721">tvm::runtime::TVMRetValue::operator=</a></div><div class="ttdeci">TVMRetValue &amp; operator=(PackedFunc f)</div><div class="ttdef"><b>Definition:</b> packed_func.h:924</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1PackedFuncObj_html_a063452b7982696b09f35b20993ac3138"><div class="ttname"><a href="classtvm_1_1runtime_1_1PackedFuncObj.html#a063452b7982696b09f35b20993ac3138">tvm::runtime::PackedFuncObj::PackedFuncObj</a></div><div class="ttdeci">PackedFuncObj(FCallPacked *f_call_pack)</div><div class="ttdoc">Constructing a packed function object from a function pointer. </div><div class="ttdef"><b>Definition:</b> packed_func.h:103</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMArgsSetter_html_a3a8c3436ed1235e1c6bc4825b1e606a8"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMArgsSetter.html#a3a8c3436ed1235e1c6bc4825b1e606a8">tvm::runtime::TVMArgsSetter::operator()</a></div><div class="ttdeci">TVM_ALWAYS_INLINE void operator()(size_t i, const TObjectRef &amp;value) const</div><div class="ttdef"><b>Definition:</b> packed_func.h:1589</div></div>
@@ -201,7 +201,7 @@ $(function() {
 <div class="ttc" id="object_8h_html"><div class="ttname"><a href="object_8h.html">object.h</a></div><div class="ttdoc">A managed object in the TVM runtime. </div></div>
 <div class="ttc" id="data__type_8h_html"><div class="ttname"><a href="data__type_8h.html">data_type.h</a></div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMArgsSetter_html_af0f7e76657fd06170c6347f47a42a342"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMArgsSetter.html#af0f7e76657fd06170c6347f47a42a342">tvm::runtime::TVMArgsSetter::operator()</a></div><div class="ttdeci">TVM_ALWAYS_INLINE void operator()(size_t i, const TVMArgValue &amp;value) const</div><div class="ttdef"><b>Definition:</b> packed_func.h:1536</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a141e032d848c60f8261046304bdc8c4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a141e032d848c60f8261046304bdc8c4c">tvm::runtime::NDArray::FFIGetHandle</a></div><div class="ttdeci">static TVMArrayHandle FFIGetHandle(const ObjectRef &amp;nd)</div><div class="ttdoc">Get FFI Array handle from ndarray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:388</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_html_a141e032d848c60f8261046304bdc8c4c"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray.html#a141e032d848c60f8261046304bdc8c4c">tvm::runtime::NDArray::FFIGetHandle</a></div><div class="ttdeci">static TVMArrayHandle FFIGetHandle(const ObjectRef &amp;nd)</div><div class="ttdoc">Get FFI Array handle from ndarray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:406</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TVMArgsSetter_html_a46970e7b592c6bc08278ba7774ceb3be"><div class="ttname"><a href="classtvm_1_1runtime_1_1TVMArgsSetter.html#a46970e7b592c6bc08278ba7774ceb3be">tvm::runtime::TVMArgsSetter::operator()</a></div><div class="ttdeci">void operator()(size_t i, const TVMRetValue &amp;value) const</div><div class="ttdef"><b>Definition:</b> packed_func.h:1576</div></div>
 <div class="ttc" id="runtime_2module_8h_html"><div class="ttname"><a href="runtime_2module_8h.html">module.h</a></div><div class="ttdoc">Runtime container of the functions generated by TVM, This is used to support dynamically link...</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4_html_a0e4a4d01d86eca79c5d9e1e90322c5cb"><div class="ttname"><a href="classtvm_1_1runtime_1_1TypedPackedFunc_3_01R_07Args_8_8_8_08_4.html#a0e4a4d01d86eca79c5d9e1e90322c5cb">tvm::runtime::TypedPackedFunc&lt; R(Args...)&gt;::operator==</a></div><div class="ttdeci">bool operator==(std::nullptr_t null) const</div><div class="ttdef"><b>Definition:</b> packed_func.h:360</div></div>
diff --git a/docs/reference/api/doxygen/search/all_11.js b/docs/reference/api/doxygen/search/all_11.js
index 83413fcd5..12e511ba2 100644
--- a/docs/reference/api/doxygen/search/all_11.js
+++ b/docs/reference/api/doxygen/search/all_11.js
@@ -33,7 +33,7 @@ var searchData=
   ['page_5fallocator_2eh',['page_allocator.h',['../page__allocator_8h.html',1,'']]],
   ['pagememorymanagercreate',['PageMemoryManagerCreate',['../page__allocator_8h.html#a720dbc7474ac13b93fafb974cfc20bc7',1,'page_allocator.h']]],
   ['papi_2eh',['papi.h',['../papi_8h.html',1,'']]],
-  ['parallel',['parallel',['../classtvm_1_1auto__scheduler_1_1State.html#a2376f0180bc5b5dd4b456f2a75d4a366',1,'tvm::auto_scheduler::State::parallel()'],['../classtvm_1_1te_1_1Stage.html#a60a6be10a1a96cb594c1399efabafef3',1,'tvm::te::Stage::parallel()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789',1,'tvm::tir::ScheduleNode::Parallel()']]],
+  ['parallel',['Parallel',['../classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789',1,'tvm::tir::ScheduleNode::Parallel()'],['../classtvm_1_1auto__scheduler_1_1State.html#a2376f0180bc5b5dd4b456f2a75d4a366',1,'tvm::auto_scheduler::State::parallel()'],['../classtvm_1_1te_1_1Stage.html#a60a6be10a1a96cb594c1399efabafef3',1,'tvm::te::Stage::parallel()']]],
   ['parallel_5ffor',['parallel_for',['../namespacetvm_1_1support.html#a8bf1225e8bb1db575578ca2d645fb23c',1,'tvm::support']]],
   ['parallel_5ffor_2eh',['parallel_for.h',['../parallel__for_8h.html',1,'']]],
   ['parallel_5ffor_5fdynamic',['parallel_for_dynamic',['../namespacetvm_1_1support.html#afe4271363c794f1644ce7af5c2266530',1,'tvm::support']]],
diff --git a/docs/reference/api/doxygen/search/all_13.js b/docs/reference/api/doxygen/search/all_13.js
index 609d0696f..1ef177e99 100644
--- a/docs/reference/api/doxygen/search/all_13.js
+++ b/docs/reference/api/doxygen/search/all_13.js
@@ -79,7 +79,7 @@ var searchData=
   ['registerconfigoption',['RegisterConfigOption',['../classtvm_1_1transform_1_1PassContext.html#a6f1d1040cc97320414b4690203f87919',1,'tvm::transform::PassContext']]],
   ['registergenericfunc',['RegisterGenericFunc',['../classtvm_1_1GenericFunc.html#a909acecbf2f34f847a34e587a4570dce',1,'tvm::GenericFunc']]],
   ['registerorget',['RegisterOrGet',['../classtvm_1_1OpRegEntry.html#a39a4d3e7f905eb4e29ca464bcedb05bd',1,'tvm::OpRegEntry::RegisterOrGet()'],['../classtvm_1_1relay_1_1ExecutorRegEntry.html#a03347a2b68269b853a7c0399994951ef',1,'tvm::relay::ExecutorRegEntry::RegisterOrGet()'],['../classtvm_1_1relay_1_1RuntimeRegEntry.html#ae8b479159ccd8b35b75950fcda58dd9d',1,'tvm::relay::RuntimeRegEntry::RegisterOrGet()'],['../classtvm_1_1TargetTagRegEntry.html#a07e0631600484dc0985ca62b1620461c',1,'tvm::T [...]
-  ['registry',['Registry',['../classtvm_1_1ReflectionVTable_1_1Registry.html',1,'tvm::ReflectionVTable::Registry'],['../classtvm_1_1runtime_1_1Registry.html',1,'tvm::runtime::Registry'],['../structTVMMutableFuncRegistry.html#acc1fcd6554c627c1bf3b3c00e1120e9b',1,'TVMMutableFuncRegistry::registry()'],['../structTVMModule.html#a6db21005b9e983207b341e65af4c4ab7',1,'TVMModule::registry()'],['../classtvm_1_1ReflectionVTable_1_1Registry.html#ac8f4637640aa9dffed745303a4cfa827',1,'tvm::Reflection [...]
+  ['registry',['Registry',['../classtvm_1_1ReflectionVTable_1_1Registry.html',1,'tvm::ReflectionVTable::Registry'],['../classtvm_1_1runtime_1_1Registry.html',1,'tvm::runtime::Registry'],['../classtvm_1_1ReflectionVTable_1_1Registry.html#ac8f4637640aa9dffed745303a4cfa827',1,'tvm::ReflectionVTable::Registry::Registry()'],['../structTVMMutableFuncRegistry.html#acc1fcd6554c627c1bf3b3c00e1120e9b',1,'TVMMutableFuncRegistry::registry()'],['../structTVMModule.html#a6db21005b9e983207b341e65af4c4a [...]
   ['registry_2eh',['registry.h',['../registry_8h.html',1,'']]],
   ['regname',['RegName',['../namespacetvm_1_1runtime_1_1vm.html#a3bbbf700719e9dc3dda2bc25210c18ae',1,'tvm::runtime::vm']]],
   ['reinterpret',['reinterpret',['../namespacetvm_1_1tir_1_1builtin.html#a7b555bc5cca2f5e7b26c1037bc0001ce',1,'tvm::tir::builtin::reinterpret()'],['../namespacetvm.html#a34084606675cd2c73c6b0f10e1618280',1,'tvm::reinterpret()'],['../namespacetvm_1_1topi.html#a25239505894bdae140e53f4abc146f92',1,'tvm::topi::reinterpret()']]],
@@ -148,7 +148,7 @@ var searchData=
   ['resize2dattrs',['Resize2DAttrs',['../structtvm_1_1relay_1_1Resize2DAttrs.html',1,'tvm::relay']]],
   ['resize3dattrs',['Resize3DAttrs',['../structtvm_1_1relay_1_1Resize3DAttrs.html',1,'tvm::relay']]],
   ['resolvedependency',['ResolveDependency',['../classtvm_1_1transform_1_1SequentialNode.html#a5549edf77e0a64bd6fcb692603967b8e',1,'tvm::transform::SequentialNode']]],
-  ['result',['Result',['../classtvm_1_1meta__schedule_1_1RunnerFutureNode.html#a1b5438c21c436ce7a864487583fd32b2',1,'tvm::meta_schedule::RunnerFutureNode::Result()'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#ae0d33229af059c727db2abd3616660e0',1,'tvm::runtime::vm::Instruction::result()'],['../classtvm_1_1tir_1_1CommReducerNode.html#a7030917568a088215da423fc56882814',1,'tvm::tir::CommReducerNode::result()']]],
+  ['result',['result',['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#ae0d33229af059c727db2abd3616660e0',1,'tvm::runtime::vm::Instruction::result()'],['../classtvm_1_1tir_1_1CommReducerNode.html#a7030917568a088215da423fc56882814',1,'tvm::tir::CommReducerNode::result()'],['../classtvm_1_1meta__schedule_1_1RunnerFutureNode.html#a1b5438c21c436ce7a864487583fd32b2',1,'tvm::meta_schedule::RunnerFutureNode::Result()']]],
   ['result_5f',['result_',['../classtvm_1_1detail_1_1AttrsSEqualVisitor.html#aeda3a91f0b2d1a7a9a075828954ff77f',1,'tvm::detail::AttrsSEqualVisitor']]],
   ['result_5ftype',['result_type',['../classtvm_1_1TypeFunctor_3_01R_07const_01Type_01_6n_00_01Args_8_8_8_08_4.html#a24d4a3522ee6c4cdeed80dcdcc1424ad',1,'tvm::TypeFunctor&lt; R(const Type &amp;n, Args...)&gt;::result_type()'],['../classtvm_1_1NodeFunctor_3_01R_07const_01ObjectRef_01_6n_00_01Args_8_8_8_08_4.html#ac7f687cb7dda02407b578a6683fa708a',1,'tvm::NodeFunctor&lt; R(const ObjectRef &amp;n, Args...)&gt;::result_type()'],['../classtvm_1_1relay_1_1ExprFunctor_3_01R_07const_01Expr_01_6n [...]
   ['resulttype',['ResultType',['../structtvm_1_1runtime_1_1Array_1_1ValueConverter.html#a0db77cfd8032391d76dffc88eae8e09b',1,'tvm::runtime::Array::ValueConverter']]],
diff --git a/docs/reference/api/doxygen/search/all_14.js b/docs/reference/api/doxygen/search/all_14.js
index ec493d53b..911493efc 100644
--- a/docs/reference/api/doxygen/search/all_14.js
+++ b/docs/reference/api/doxygen/search/all_14.js
@@ -91,7 +91,7 @@ var searchData=
   ['selectshashreduce_3c_20t_2c_20traitname_2c_20false_20_3e',['SelectSHashReduce&lt; T, TraitName, false &gt;',['../structtvm_1_1detail_1_1SelectSHashReduce_3_01T_00_01TraitName_00_01false_01_4.html',1,'tvm::detail']]],
   ['selectvisitattrs',['SelectVisitAttrs',['../structtvm_1_1detail_1_1SelectVisitAttrs.html',1,'tvm::detail']]],
   ['selectvisitattrs_3c_20t_2c_20traitname_2c_20false_20_3e',['SelectVisitAttrs&lt; T, TraitName, false &gt;',['../structtvm_1_1detail_1_1SelectVisitAttrs_3_01T_00_01TraitName_00_01false_01_4.html',1,'tvm::detail']]],
-  ['self',['self',['../classtvm_1_1runtime_1_1MapNode_1_1iterator.html#a5bac4439279428fb3c0d44aa6b1cc798',1,'tvm::runtime::MapNode::iterator::self()'],['../classtvm_1_1runtime_1_1InplaceArrayBase.html#ae447f7c7a742fb3f5613a632706509df',1,'tvm::runtime::InplaceArrayBase::Self()']]],
+  ['self',['Self',['../classtvm_1_1runtime_1_1InplaceArrayBase.html#ae447f7c7a742fb3f5613a632706509df',1,'tvm::runtime::InplaceArrayBase::Self()'],['../classtvm_1_1runtime_1_1MapNode_1_1iterator.html#a5bac4439279428fb3c0d44aa6b1cc798',1,'tvm::runtime::MapNode::iterator::self()']]],
   ['sendbodychunk',['SendBodyChunk',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a37b77101825145283cced6cd05eb502c',1,'tvm::runtime::micro_rpc::Session']]],
   ['sendmessage',['SendMessage',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a6e540521a7e9188564da712c0641619c',1,'tvm::runtime::micro_rpc::Session']]],
   ['seq',['seq',['../classtvm_1_1tir_1_1SeqStmtNode.html#a0e548955529d35c56e646fcaac38f865',1,'tvm::tir::SeqStmtNode']]],
@@ -211,7 +211,7 @@ var searchData=
   ['singleton',['Singleton',['../classtvm_1_1te_1_1Singleton.html',1,'tvm::te::Singleton'],['../classtvm_1_1te_1_1Singleton.html#a94450b853dcd5e9865546d8c8fe351a1',1,'tvm::te::Singleton::Singleton()']]],
   ['singletonnode',['SingletonNode',['../classtvm_1_1te_1_1SingletonNode.html',1,'tvm::te']]],
   ['sinh',['sinh',['../namespacetvm.html#ad828bc801c73df761c58d9f8877d52ee',1,'tvm::sinh()'],['../namespacetvm_1_1topi.html#af9694f5470ba2cabc19866be3b00fe8d',1,'tvm::topi::sinh()']]],
-  ['size',['Size',['../classtvm_1_1TensorTypeNode.html#a1f08dac86ae8aea81d058ef64cfd38b4',1,'tvm::TensorTypeNode::Size()'],['../classtvm_1_1meta__schedule_1_1DatabaseNode.html#aae5b9ab9f7e497654b90c23a2159a5cc',1,'tvm::meta_schedule::DatabaseNode::Size()'],['../classtvm_1_1meta__schedule_1_1PyDatabaseNode.html#a36817d04978253571fef7d01427ce9c0',1,'tvm::meta_schedule::PyDatabaseNode::Size()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1FrameBuffer.html#ae395a0f1c6e79e825aa7a244c74a5d7b',1,' [...]
+  ['size',['size',['../structtvm_1_1relay_1_1Resize1DAttrs.html#afb1175c0ff019e485ed65d98305b5f62',1,'tvm::relay::Resize1DAttrs::size()'],['../structtvm_1_1relay_1_1Resize2DAttrs.html#ab3e26dbbc2dc1da40764832a99459c30',1,'tvm::relay::Resize2DAttrs::size()'],['../structtvm_1_1relay_1_1Resize3DAttrs.html#aab61649fe8417a8a7fbc849090bac083',1,'tvm::relay::Resize3DAttrs::size()'],['../structtvm_1_1relay_1_1LRNAttrs.html#a3758ed1f8a8bcf73008ae1dd2bfa148e',1,'tvm::relay::LRNAttrs::size()'],['.. [...]
   ['size_5f',['size_',['../classtvm_1_1runtime_1_1MapNode.html#a2285f106f6afa29f512a7818ad59e9e5',1,'tvm::runtime::MapNode']]],
   ['size_5fbytes',['size_bytes',['../structtvm_1_1tir_1_1usmp_1_1BufferInfoNode.html#a0a5d4bd6072c268df05b90d267b4c0a0',1,'tvm::tir::usmp::BufferInfoNode']]],
   ['size_5fhint_5fbytes',['size_hint_bytes',['../structtvm_1_1PoolInfoNode.html#ac073aeb75bf031ff8687e132bc112f92',1,'tvm::PoolInfoNode']]],
@@ -268,7 +268,7 @@ var searchData=
   ['specialize',['Specialize',['../namespacetvm_1_1tir.html#a69b6f1b0014dc6e7dd390cff746e9782',1,'tvm::tir']]],
   ['specializedcondition',['SpecializedCondition',['../classtvm_1_1te_1_1SpecializedCondition.html',1,'tvm::te::SpecializedCondition'],['../classtvm_1_1te_1_1SpecializedCondition.html#a48d119ee1c6033929a5592cfc2592e60',1,'tvm::te::SpecializedCondition::SpecializedCondition()']]],
   ['specializedconditionnode',['SpecializedConditionNode',['../classtvm_1_1te_1_1SpecializedConditionNode.html',1,'tvm::te']]],
-  ['split',['Split',['../classtvm_1_1te_1_1Split.html',1,'tvm::te::Split'],['../classtvm_1_1te_1_1Split.html#a328e0c093ce5b41ebaf33e0e80592764',1,'tvm::te::Split::Split()'],['../classtvm_1_1tir_1_1Layout.html#ad7657af7789fe040d3224c0149976bb4',1,'tvm::tir::Layout::Split()'],['../classtvm_1_1tir_1_1ScheduleNode.html#af8a330c32b06dc16c8835c76177ffa11',1,'tvm::tir::ScheduleNode::Split()'],['../classtvm_1_1auto__scheduler_1_1State.html#a5815f21fc90ba7cc379c2410c05ab54c',1,'tvm::auto_schedule [...]
+  ['split',['Split',['../classtvm_1_1te_1_1Split.html',1,'tvm::te::Split'],['../classtvm_1_1auto__scheduler_1_1State.html#a5815f21fc90ba7cc379c2410c05ab54c',1,'tvm::auto_scheduler::State::split()'],['../classtvm_1_1te_1_1Stage.html#a5a7cd562be59b68a187ad97085a3425d',1,'tvm::te::Stage::split()'],['../classtvm_1_1te_1_1Split.html#a328e0c093ce5b41ebaf33e0e80592764',1,'tvm::te::Split::Split()'],['../classtvm_1_1tir_1_1Layout.html#ad7657af7789fe040d3224c0149976bb4',1,'tvm::tir::Layout::Split( [...]
   ['split_5fby_5fnparts',['split_by_nparts',['../classtvm_1_1te_1_1Stage.html#a51432f38d9ec4792a2525023179ae604',1,'tvm::te::Stage']]],
   ['split_5fsections',['split_sections',['../namespacetvm_1_1topi.html#acc643e2ed166fa2ed82a95853e145619',1,'tvm::topi']]],
   ['splitargs',['SplitArgs',['../namespacetvm_1_1relay_1_1transform.html#a2425d757b896168a109498e8d34ba960',1,'tvm::relay::transform']]],
@@ -309,13 +309,13 @@ var searchData=
   ['stagenode',['StageNode',['../classtvm_1_1auto__scheduler_1_1StageNode.html',1,'tvm::auto_scheduler::StageNode'],['../classtvm_1_1te_1_1StageNode.html',1,'tvm::te::StageNode']]],
   ['stages',['stages',['../classtvm_1_1auto__scheduler_1_1StateNode.html#a881e14990bf228ee3fddb3721c451b9e',1,'tvm::auto_scheduler::StateNode::stages()'],['../classtvm_1_1te_1_1ScheduleNode.html#ab5649969db603d6b7b4d155c0d09cdd5',1,'tvm::te::ScheduleNode::stages()']]],
   ['stagetoaxesmap',['StageToAxesMap',['../namespacetvm_1_1auto__scheduler.html#a8f12e558fc4b8fbb990e7e204c06beeb',1,'tvm::auto_scheduler']]],
-  ['start',['start',['../structtvm_1_1relay_1_1ArangeAttrs.html#ae8ae5bc1551b406a4f52395af343c2ce',1,'tvm::relay::ArangeAttrs::start()'],['../classtvm_1_1runtime_1_1TimerNode.html#aa11fc338c39ee2137448e54a10efe0ae',1,'tvm::runtime::TimerNode::Start()'],['../classtvm_1_1runtime_1_1Timer.html#a89bcaa433499bc68902cb473d5eba6ca',1,'tvm::runtime::Timer::Start()'],['../classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#a44fadfb7b0f961a7fb2275e3b5dbcd88',1,'tvm::runtime::profiling::Me [...]
+  ['start',['Start',['../classtvm_1_1runtime_1_1TimerNode.html#aa11fc338c39ee2137448e54a10efe0ae',1,'tvm::runtime::TimerNode::Start()'],['../classtvm_1_1runtime_1_1Timer.html#a89bcaa433499bc68902cb473d5eba6ca',1,'tvm::runtime::Timer::Start()'],['../classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#a44fadfb7b0f961a7fb2275e3b5dbcd88',1,'tvm::runtime::profiling::MetricCollectorNode::Start()'],['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#aee5452075c8e022b8aaa6fb365f68e14 [...]
   ['start_5findex',['start_index',['../namespacetvm_1_1topi_1_1nn.html#a752c4130dac73fd2de0390c5f6b24b15',1,'tvm::topi::nn']]],
   ['startcall',['StartCall',['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#a1fe322f7ba92be44d7e7c8cb184f3833',1,'tvm::runtime::profiling::Profiler']]],
   ['startmessage',['StartMessage',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#acd512b977c6dd888f90c4fd6d2b9500f',1,'tvm::runtime::micro_rpc::Session']]],
   ['startpacket',['StartPacket',['../classtvm_1_1runtime_1_1micro__rpc_1_1Framer.html#ade10d3bd3a26e3b7af881ae134e9a998',1,'tvm::runtime::micro_rpc::Framer']]],
   ['startsession',['StartSession',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a15d3f9ecb8b22bf2d330f6f0a16c5239',1,'tvm::runtime::micro_rpc::Session']]],
-  ['state',['State',['../classtvm_1_1auto__scheduler_1_1State.html',1,'tvm::auto_scheduler::State'],['../classtvm_1_1auto__scheduler_1_1State.html#a9e8198b1f51b42cfbbee4b9f42160749',1,'tvm::auto_scheduler::State::State()'],['../classtvm_1_1auto__scheduler_1_1MeasureInputNode.html#afb23aaf6133189687d2541ec6e1352f4',1,'tvm::auto_scheduler::MeasureInputNode::state()'],['../classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26',1,'tvm::tir::ScheduleNode::state()']]],
+  ['state',['State',['../classtvm_1_1auto__scheduler_1_1State.html',1,'tvm::auto_scheduler::State'],['../classtvm_1_1auto__scheduler_1_1MeasureInputNode.html#afb23aaf6133189687d2541ec6e1352f4',1,'tvm::auto_scheduler::MeasureInputNode::state()'],['../classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26',1,'tvm::tir::ScheduleNode::state()'],['../classtvm_1_1auto__scheduler_1_1State.html#a9e8198b1f51b42cfbbee4b9f42160749',1,'tvm::auto_scheduler::State::State()']]],
   ['state_2eh',['state.h',['../state_8h.html',1,'']]],
   ['state_5fplaceholder',['state_placeholder',['../classtvm_1_1te_1_1ScanOpNode.html#a69105f6a84dd4fb912a16bfaa68aebf6',1,'tvm::te::ScanOpNode']]],
   ['statenode',['StateNode',['../classtvm_1_1auto__scheduler_1_1StateNode.html',1,'tvm::auto_scheduler']]],
diff --git a/docs/reference/api/doxygen/search/all_15.js b/docs/reference/api/doxygen/search/all_15.js
index 444289c38..73d5b2a10 100644
--- a/docs/reference/api/doxygen/search/all_15.js
+++ b/docs/reference/api/doxygen/search/all_15.js
@@ -32,7 +32,7 @@ var searchData=
   ['takeattrs',['TakeAttrs',['../structtvm_1_1relay_1_1TakeAttrs.html',1,'tvm::relay']]],
   ['tan',['tan',['../namespacetvm.html#af99838098788d40c80b402f29b3c2e8c',1,'tvm::tan()'],['../namespacetvm_1_1topi.html#a13b757fe52775f43a58d91c0a1330f97',1,'tvm::topi::tan()']]],
   ['tanh',['tanh',['../namespacetvm.html#a12c5457301d8a2c03a2ba1163edd7cee',1,'tvm::tanh()'],['../namespacetvm_1_1topi.html#aec153e599d33c78a7592007cde1c02cb',1,'tvm::topi::tanh()']]],
-  ['target',['Target',['../classtvm_1_1Target.html',1,'tvm::Target'],['../classtvm_1_1auto__scheduler_1_1SearchTaskNode.html#acf4407e0c8dced81b05b34ec0426c933',1,'tvm::auto_scheduler::SearchTaskNode::target()'],['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#afc001f3e427cfc8c05236b615cfd2868',1,'tvm::meta_schedule::BuilderInputNode::target()'],['../classtvm_1_1meta__schedule_1_1TuningRecordNode.html#ab9cbbf8eb7941995e9c7552948eac02b',1,'tvm::meta_schedule::TuningRecordNode::targ [...]
+  ['target',['Target',['../classtvm_1_1Target.html',1,'tvm::Target'],['../classtvm_1_1Target.html#a58a5a1e042e265fe5a6973045226fe1a',1,'tvm::Target::Target(std::nullptr_t)'],['../classtvm_1_1Target.html#a77f3d7cc97d8cfd7172af58b4e784d89',1,'tvm::Target::Target(const String &amp;tag_or_config_or_target_str)'],['../classtvm_1_1Target.html#ab825b350cf478bf948d807b6fdf636a0',1,'tvm::Target::Target(const Map&lt; String, ObjectRef &gt; &amp;config)'],['../classtvm_1_1Target.html#a1abb29217d8e3 [...]
   ['target_2eh',['target.h',['../target_8h.html',1,'']]],
   ['target_5faccess',['target_access',['../structtvm_1_1PoolInfoNode.html#a78514ba53ee1471fa6069800c56c0612',1,'tvm::PoolInfoNode']]],
   ['target_5fburst_5fbytes',['target_burst_bytes',['../structtvm_1_1PoolInfoNode.html#a747c03e3eafc83b053637b735244c6d7',1,'tvm::PoolInfoNode']]],
@@ -147,7 +147,7 @@ var searchData=
   ['touchtask',['TouchTask',['../classtvm_1_1meta__schedule_1_1TaskSchedulerNode.html#af6fa276674945d3432c129bdf9cea599',1,'tvm::meta_schedule::TaskSchedulerNode::TouchTask()'],['../classtvm_1_1meta__schedule_1_1PyTaskSchedulerNode.html#a7de09f81c8aceb580b43107f266e6b40',1,'tvm::meta_schedule::PyTaskSchedulerNode::TouchTask()']]],
   ['tovar',['ToVar',['../classtvm_1_1tir_1_1AnyNode.html#ae01ebbba2378afb6509a22de97f8fb30',1,'tvm::tir::AnyNode']]],
   ['tparent',['TParent',['../classtvm_1_1OpAttrMap.html#a316480ca7450209650fc1a62f7ce4a14',1,'tvm::OpAttrMap::TParent()'],['../classtvm_1_1TargetKindAttrMap.html#a37eb6bfb0d881cf897147b17ff7d3265',1,'tvm::TargetKindAttrMap::TParent()']]],
-  ['trace',['Trace',['../classtvm_1_1tir_1_1Trace.html',1,'tvm::tir::Trace'],['../classtvm_1_1tir_1_1Trace.html#a8e09abffd0b9b1afac7b832cf16c142d',1,'tvm::tir::Trace::Trace()'],['../classtvm_1_1tir_1_1Trace.html#af79bccf1bde25efea387bb1b82dacaa6',1,'tvm::tir::Trace::Trace(Array&lt; Instruction &gt; insts, Map&lt; Instruction, ObjectRef &gt; decisions)'],['../classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a8cc2d64f796593a1a774eef259f17b29',1,'tvm::meta_schedule::TuningRecordNode::tra [...]
+  ['trace',['Trace',['../classtvm_1_1tir_1_1Trace.html',1,'tvm::tir::Trace'],['../classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a8cc2d64f796593a1a774eef259f17b29',1,'tvm::meta_schedule::TuningRecordNode::trace()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a953bca4123b5a758adfdcd65634a5f3b',1,'tvm::tir::ScheduleNode::trace()'],['../classtvm_1_1tir_1_1Trace.html#a8e09abffd0b9b1afac7b832cf16c142d',1,'tvm::tir::Trace::Trace()'],['../classtvm_1_1tir_1_1Trace.html#af79bccf1bde25efea387bb [...]
   ['trace_2eh',['trace.h',['../trace_8h.html',1,'']]],
   ['traced',['Traced',['../classtvm_1_1tir_1_1Schedule.html#a295d432b86621101f67b20fadb367b91',1,'tvm::tir::Schedule']]],
   ['tracenode',['TraceNode',['../classtvm_1_1tir_1_1TraceNode.html',1,'tvm::tir']]],
diff --git a/docs/reference/api/doxygen/search/all_16.js b/docs/reference/api/doxygen/search/all_16.js
index 08dc5ba40..20d4f627d 100644
--- a/docs/reference/api/doxygen/search/all_16.js
+++ b/docs/reference/api/doxygen/search/all_16.js
@@ -15,7 +15,7 @@ var searchData=
   ['unionregion',['UnionRegion',['../namespacetvm_1_1arith.html#ad27c4f216e41eb8e81296fb7ec4b9453',1,'tvm::arith']]],
   ['unionregionlowerbound',['UnionRegionLowerBound',['../namespacetvm_1_1arith.html#a4c3dedfa4cba4ad39c953eb51eb83e4d',1,'tvm::arith']]],
   ['unipolar',['unipolar',['../structtvm_1_1relay_1_1BinaryConv2DAttrs.html#a7e0ad68dce226079b769a678aa01dc49',1,'tvm::relay::BinaryConv2DAttrs::unipolar()'],['../structtvm_1_1relay_1_1BinaryDenseAttrs.html#af21cdb9dac67ab9ecea5a19642658d8a',1,'tvm::relay::BinaryDenseAttrs::unipolar()']]],
-  ['unique',['unique',['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()'],['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()']]],
+  ['unique',['Unique',['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()'],['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()']]],
   ['uniqueattrs',['UniqueAttrs',['../structtvm_1_1relay_1_1UniqueAttrs.html',1,'tvm::relay']]],
   ['unit_5fbits',['unit_bits',['../classtvm_1_1MemoryInfoNode.html#aa935f1ee9d8d2f06633ca4b3c44f7725',1,'tvm::MemoryInfoNode']]],
   ['units',['units',['../structtvm_1_1relay_1_1BinaryDenseAttrs.html#a5373b2f2aac19653ae21aec74c69cdb0',1,'tvm::relay::BinaryDenseAttrs::units()'],['../structtvm_1_1relay_1_1MatmulAttrs.html#a5893df9ad99c6717c4e6cb440d60c6a1',1,'tvm::relay::MatmulAttrs::units()'],['../structtvm_1_1relay_1_1DenseAttrs.html#a497487f7ccced8c7492a5ed03f78fa8f',1,'tvm::relay::DenseAttrs::units()'],['../structtvm_1_1relay_1_1DensePackAttrs.html#aa0096c26c832166de13881a032ba3fbf',1,'tvm::relay::DensePackAttrs:: [...]
diff --git a/docs/reference/api/doxygen/search/all_17.js b/docs/reference/api/doxygen/search/all_17.js
index 54b74dab8..6a8d7ec76 100644
--- a/docs/reference/api/doxygen/search/all_17.js
+++ b/docs/reference/api/doxygen/search/all_17.js
@@ -30,7 +30,7 @@ var searchData=
   ['vector_5funit_5fbytes',['vector_unit_bytes',['../classtvm_1_1auto__scheduler_1_1HardwareParamsNode.html#a6f2dd9161fdb3233417a9912c8854434',1,'tvm::auto_scheduler::HardwareParamsNode']]],
   ['vectorcombine',['vectorcombine',['../namespacetvm_1_1tir_1_1builtin.html#a30dff65bc2c142b57fae7f60e378ff43',1,'tvm::tir::builtin']]],
   ['vectorhigh',['vectorhigh',['../namespacetvm_1_1tir_1_1builtin.html#a45bf65ca7ca01d2016e0b609117d7e25',1,'tvm::tir::builtin']]],
-  ['vectorize',['Vectorize',['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()'],['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()']]],
+  ['vectorize',['vectorize',['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()']]],
   ['vectorizeloop',['VectorizeLoop',['../namespacetvm_1_1tir_1_1transform.html#af3cecb50a8b8fc8021f6a87bc27587da',1,'tvm::tir::transform']]],
   ['vectorizer',['Vectorizer',['../classtvm_1_1tir_1_1BufferLoadNode.html#a842a72b9d02a9f8541b512478932fece',1,'tvm::tir::BufferLoadNode']]],
   ['vectorjacobianproduct',['VectorJacobianProduct',['../namespacetvm_1_1te.html#a547183f5a311af53ab598faba423fd64',1,'tvm::te']]],
diff --git a/docs/reference/api/doxygen/search/all_18.js b/docs/reference/api/doxygen/search/all_18.js
index ff7ab68e1..d9ca258ac 100644
--- a/docs/reference/api/doxygen/search/all_18.js
+++ b/docs/reference/api/doxygen/search/all_18.js
@@ -26,7 +26,7 @@ var searchData=
   ['withfields',['WithFields',['../namespacetvm_1_1relay.html#acd80501d29e4d951be6746c79934a70c',1,'tvm::relay::WithFields(Clause clause, Optional&lt; Pattern &gt; opt_lhs=Optional&lt; Pattern &gt;(), Optional&lt; Expr &gt; opt_rhs=Optional&lt; Expr &gt;())'],['../namespacetvm_1_1relay.html#adb39b46f86b66a5e7252f6d9102deb7b',1,'tvm::relay::WithFields(Match match, Optional&lt; Expr &gt; opt_data=Optional&lt; Expr &gt;(), Optional&lt; Array&lt; Clause &gt;&gt; opt_clauses=Optional&lt; Arra [...]
   ['withhost',['WithHost',['../classtvm_1_1Target.html#a509ce63995f082c80742ea5ca6ac112f',1,'tvm::Target']]],
   ['withoutattr',['WithoutAttr',['../namespacetvm.html#a7e2bc626db8be997b1562c79df3d9e11',1,'tvm']]],
-  ['workload',['Workload',['../classtvm_1_1meta__schedule_1_1Workload.html',1,'tvm::meta_schedule::Workload'],['../classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a42c87f1ec62dae6806c3fe9629c5e7f0',1,'tvm::meta_schedule::TuningRecordNode::workload()'],['../classtvm_1_1meta__schedule_1_1Workload.html#a21ccf9c956b82d50a2579f1c0f592fd0',1,'tvm::meta_schedule::Workload::Workload(IRModule mod)'],['../classtvm_1_1meta__schedule_1_1Workload.html#a8880877517679c82ae63520e28d5e1d8',1,'tvm::me [...]
+  ['workload',['Workload',['../classtvm_1_1meta__schedule_1_1Workload.html',1,'tvm::meta_schedule::Workload'],['../classtvm_1_1meta__schedule_1_1Workload.html#a21ccf9c956b82d50a2579f1c0f592fd0',1,'tvm::meta_schedule::Workload::Workload(IRModule mod)'],['../classtvm_1_1meta__schedule_1_1Workload.html#a8880877517679c82ae63520e28d5e1d8',1,'tvm::meta_schedule::Workload::Workload(IRModule mod, THashCode shash)'],['../classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a42c87f1ec62dae6806c3fe9 [...]
   ['workload_5fkey',['workload_key',['../classtvm_1_1auto__scheduler_1_1SearchTaskNode.html#a20045d677ba2bc5c5ce461e78543b3e2',1,'tvm::auto_scheduler::SearchTaskNode']]],
   ['workloadequal',['WorkloadEqual',['../structtvm_1_1meta__schedule_1_1WorkloadEqual.html',1,'tvm::meta_schedule']]],
   ['workloadhash',['WorkloadHash',['../structtvm_1_1meta__schedule_1_1WorkloadHash.html',1,'tvm::meta_schedule']]],
diff --git a/docs/reference/api/doxygen/search/all_7.js b/docs/reference/api/doxygen/search/all_7.js
index a7b725d7c..f3b6a7923 100644
--- a/docs/reference/api/doxygen/search/all_7.js
+++ b/docs/reference/api/doxygen/search/all_7.js
@@ -193,6 +193,7 @@ var searchData=
   ['fromdlpack',['FromDLPack',['../classtvm_1_1runtime_1_1NDArray.html#abec485628a0ca451b668c42fd8fa691a',1,'tvm::runtime::NDArray']]],
   ['fromexpr',['FromExpr',['../classtvm_1_1IRModule.html#a59099426f65dbeac227e51f8864e322a',1,'tvm::IRModule']]],
   ['fromexprincontext',['FromExprInContext',['../classtvm_1_1IRModule.html#a1cc91fc2b2adaca5a4dcfc14baf28c27',1,'tvm::IRModule']]],
+  ['fromexternaldltensor',['FromExternalDLTensor',['../classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e',1,'tvm::runtime::NDArray']]],
   ['fromfunc',['FromFunc',['../classtvm_1_1tir_1_1IndexMap.html#afa04f25f10b1dac139df9a1b34598cbb',1,'tvm::tir::IndexMap']]],
   ['fromjson',['FromJSON',['../classtvm_1_1meta__schedule_1_1ArgInfo.html#afc2cfa9fdf0bcedc79e90f07a596f74a',1,'tvm::meta_schedule::ArgInfo::FromJSON()'],['../classtvm_1_1meta__schedule_1_1TensorInfo.html#a1d4166d5ac2777b955c7263ebcaa068a',1,'tvm::meta_schedule::TensorInfo::FromJSON()'],['../classtvm_1_1meta__schedule_1_1Workload.html#a1c3076818c9a20d8e7c675a8ce58f8f3',1,'tvm::meta_schedule::Workload::FromJSON()'],['../classtvm_1_1meta__schedule_1_1TuningRecord.html#aabec8835c7178c808063 [...]
   ['fromminextent',['FromMinExtent',['../classtvm_1_1arith_1_1IntSet.html#a5eaf5c75ebfc33cf04373bc2d0071465',1,'tvm::arith::IntSet::FromMinExtent()'],['../classtvm_1_1Range.html#a91e7301ca1d135ca5f8ed199efbb9818',1,'tvm::Range::FromMinExtent()']]],
diff --git a/docs/reference/api/doxygen/search/all_e.js b/docs/reference/api/doxygen/search/all_e.js
index a0afc259e..1ef02082c 100644
--- a/docs/reference/api/doxygen/search/all_e.js
+++ b/docs/reference/api/doxygen/search/all_e.js
@@ -63,7 +63,7 @@ var searchData=
   ['matmulattrs',['MatmulAttrs',['../structtvm_1_1relay_1_1MatmulAttrs.html',1,'tvm::relay']]],
   ['matrix_5fset_5fdiag',['matrix_set_diag',['../namespacetvm_1_1topi.html#aead477c6c9d4f4589d22b8acff82040c',1,'tvm::topi']]],
   ['matrixsetdiagattrs',['MatrixSetDiagAttrs',['../structtvm_1_1relay_1_1MatrixSetDiagAttrs.html',1,'tvm::relay']]],
-  ['max',['Max',['../classtvm_1_1tir_1_1Max.html',1,'tvm::tir::Max'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, [...]
+  ['max',['Max',['../classtvm_1_1tir_1_1Max.html',1,'tvm::tir::Max'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, [...]
   ['max_5fcontinuous_5ferror',['max_continuous_error',['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#abdc38da91bcdf77be765c1e3d5af3648',1,'tvm::auto_scheduler::ProgramMeasurerNode']]],
   ['max_5fdisplacement',['max_displacement',['../structtvm_1_1relay_1_1CorrelationAttrs.html#ad1d16e2ba537736c8baee2553e1e32bf',1,'tvm::relay::CorrelationAttrs']]],
   ['max_5ffunctions',['max_functions',['../structTVMMutableFuncRegistry.html#a41745f8e0f73f8e4fb2074f5b154b49c',1,'TVMMutableFuncRegistry']]],
@@ -165,7 +165,7 @@ var searchData=
   ['mixedmodemutator',['MixedModeMutator',['../classtvm_1_1relay_1_1MixedModeMutator.html',1,'tvm::relay::MixedModeMutator'],['../classtvm_1_1relay_1_1MixedModeMutator.html#add94673de95267df385059ac4cc9c519',1,'tvm::relay::MixedModeMutator::MixedModeMutator()']]],
   ['mixedmodevisitor',['MixedModeVisitor',['../classtvm_1_1relay_1_1MixedModeVisitor.html',1,'tvm::relay::MixedModeVisitor'],['../classtvm_1_1relay_1_1MixedModeVisitor.html#ad98abd46ba60b173c3e33a96aa3ef9b4',1,'tvm::relay::MixedModeVisitor::MixedModeVisitor()']]],
   ['mixedmodulepassmanager',['MixedModulePassManager',['../namespacetvm.html#abc01352eff102d4902632d097adc0e08',1,'tvm']]],
-  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html',1,'tvm::tir::Mod'],['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#ab2fb058ca54af03b5bc47bf4fac23cf7',1,'tvm::meta_schedule::BuilderInputNode::mod()'],['../classtvm_1_1meta__schedule_1_1WorkloadNode.html#a3929f2761c168c25de6be2247b913911',1,'tvm::meta_schedule::WorkloadNode::mod()'],['../classtvm_1_1meta__schedule_1_1ExtractedTaskNode.html#a50c40aa8beb57d0f31c36ef360042be6',1,'tvm::meta_schedule::ExtractedTaskNode::mod()'],['../c [...]
+  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html',1,'tvm::tir::Mod'],['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#ab2fb058ca54af03b5bc47bf4fac23cf7',1,'tvm::meta_schedule::BuilderInputNode::mod()'],['../classtvm_1_1meta__schedule_1_1WorkloadNode.html#a3929f2761c168c25de6be2247b913911',1,'tvm::meta_schedule::WorkloadNode::mod()'],['../classtvm_1_1meta__schedule_1_1ExtractedTaskNode.ht [...]
   ['mod_5fname',['mod_name',['../structTVMMetadata.html#a32e45fcae0f9328e944a35a885d94276',1,'TVMMetadata']]],
   ['mode',['mode',['../structtvm_1_1relay_1_1MirrorPadAttrs.html#af5381d72f1d9c9abcb9d2e522966ad86',1,'tvm::relay::MirrorPadAttrs::mode()'],['../structtvm_1_1relay_1_1SubPixelAttrs.html#a6f0822aa1ad7672a18ab73c64e83fa99',1,'tvm::relay::SubPixelAttrs::mode()'],['../structtvm_1_1relay_1_1ScatterNDAttrs.html#ab13eeaa700fe7e41666ac04179e0fd62',1,'tvm::relay::ScatterNDAttrs::mode()'],['../structtvm_1_1relay_1_1TakeAttrs.html#a0bf9d25ced9bfc91e766494e5f641e70',1,'tvm::relay::TakeAttrs::mode()' [...]
   ['modnode',['ModNode',['../classtvm_1_1tir_1_1ModNode.html',1,'tvm::tir']]],
diff --git a/docs/reference/api/doxygen/search/all_f.js b/docs/reference/api/doxygen/search/all_f.js
index 2c1182999..b1f9ecc8d 100644
--- a/docs/reference/api/doxygen/search/all_f.js
+++ b/docs/reference/api/doxygen/search/all_f.js
@@ -24,6 +24,7 @@ var searchData=
   ['negative',['negative',['../namespacetvm_1_1topi.html#af6b3e60333fce92bcf0930e45683a8f6',1,'tvm::topi']]],
   ['nenode',['NENode',['../classtvm_1_1tir_1_1NENode.html',1,'tvm::tir']]],
   ['new',['New',['../classtvm_1_1runtime_1_1SimpleObjAllocator_1_1Handler.html#afedd0ba3dc8dc82c7566bb9120a7c56d',1,'tvm::runtime::SimpleObjAllocator::Handler::New()'],['../classtvm_1_1runtime_1_1SimpleObjAllocator_1_1ArrayHandler.html#a310471cff82c5d0836f65ec7f199e621',1,'tvm::runtime::SimpleObjAllocator::ArrayHandler::New()']]],
+  ['newfromdltensor',['NewFromDLTensor',['../classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2',1,'tvm::runtime::NDArray']]],
   ['newshape',['newshape',['../structtvm_1_1relay_1_1ReshapeAttrs.html#a9bca32c3acff2ed8fd6bc63a50f82051',1,'tvm::relay::ReshapeAttrs::newshape()'],['../structtvm_1_1relay_1_1ReshapeTensorAttrs.html#aaacd1ab5124b54316a9e1f3ef5a5ec3c',1,'tvm::relay::ReshapeTensorAttrs::newshape()'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a5602ccf14b6bd90e33a38c6ab82b0b57',1,'tvm::runtime::vm::Instruction::newshape()']]],
   ['next_5falloc',['next_alloc',['../structtvm__workspace__t.html#a5da9eaf15149d785a9b537f7c9e3945b',1,'tvm_workspace_t']]],
   ['nextafter',['nextafter',['../namespacetvm.html#a96d86ba91e4855c84879ba886465cacf',1,'tvm']]],
diff --git a/docs/reference/api/doxygen/search/functions_10.js b/docs/reference/api/doxygen/search/functions_10.js
index 134a2c594..0bea9cdd7 100644
--- a/docs/reference/api/doxygen/search/functions_10.js
+++ b/docs/reference/api/doxygen/search/functions_10.js
@@ -9,7 +9,7 @@ var searchData=
   ['packimportstollvm',['PackImportsToLLVM',['../namespacetvm_1_1codegen.html#ab2cd2a65bac4b26427a8ca0abe4e0bd6',1,'tvm::codegen']]],
   ['pad',['Pad',['../namespacetvm_1_1topi.html#a97c798d0a0ec20a95d351618b83d5121',1,'tvm::topi::Pad(const Array&lt; PrimExpr &gt; shape, int odim)'],['../namespacetvm_1_1topi.html#a3305d377f96cd20c23032eeada2756d5',1,'tvm::topi::pad(const tvm::te::Tensor &amp;t, const tvm::Array&lt; tvm::PrimExpr &gt; &amp;pad_before, tvm::Array&lt; tvm::PrimExpr &gt; pad_after=tvm::Array&lt; tvm::PrimExpr &gt;(), PrimExpr pad_value=PrimExpr(), std::string name=&quot;T_pad&quot;, std::string tag=kElement [...]
   ['pagememorymanagercreate',['PageMemoryManagerCreate',['../page__allocator_8h.html#a720dbc7474ac13b93fafb974cfc20bc7',1,'page_allocator.h']]],
-  ['parallel',['parallel',['../classtvm_1_1auto__scheduler_1_1State.html#a2376f0180bc5b5dd4b456f2a75d4a366',1,'tvm::auto_scheduler::State::parallel()'],['../classtvm_1_1te_1_1Stage.html#a60a6be10a1a96cb594c1399efabafef3',1,'tvm::te::Stage::parallel()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789',1,'tvm::tir::ScheduleNode::Parallel()']]],
+  ['parallel',['Parallel',['../classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789',1,'tvm::tir::ScheduleNode::Parallel()'],['../classtvm_1_1auto__scheduler_1_1State.html#a2376f0180bc5b5dd4b456f2a75d4a366',1,'tvm::auto_scheduler::State::parallel()'],['../classtvm_1_1te_1_1Stage.html#a60a6be10a1a96cb594c1399efabafef3',1,'tvm::te::Stage::parallel()']]],
   ['parallel_5ffor',['parallel_for',['../namespacetvm_1_1support.html#a8bf1225e8bb1db575578ca2d645fb23c',1,'tvm::support']]],
   ['parallel_5ffor_5fdynamic',['parallel_for_dynamic',['../namespacetvm_1_1support.html#afe4271363c794f1644ce7af5c2266530',1,'tvm::support']]],
   ['parallelizevectorizeunroll',['ParallelizeVectorizeUnroll',['../classtvm_1_1meta__schedule_1_1ScheduleRule.html#a0ef9b604081db7a8bf960f3fbfd3a804',1,'tvm::meta_schedule::ScheduleRule']]],
diff --git a/docs/reference/api/doxygen/search/functions_13.js b/docs/reference/api/doxygen/search/functions_13.js
index 181b84ced..f99732a1a 100644
--- a/docs/reference/api/doxygen/search/functions_13.js
+++ b/docs/reference/api/doxygen/search/functions_13.js
@@ -125,7 +125,7 @@ var searchData=
   ['singlepoint',['SinglePoint',['../classtvm_1_1arith_1_1IntSet.html#a58aeb0d34656b1b43ac2532e4dfa12ed',1,'tvm::arith::IntSet']]],
   ['singleton',['Singleton',['../classtvm_1_1te_1_1Singleton.html#a94450b853dcd5e9865546d8c8fe351a1',1,'tvm::te::Singleton']]],
   ['sinh',['sinh',['../namespacetvm.html#ad828bc801c73df761c58d9f8877d52ee',1,'tvm::sinh()'],['../namespacetvm_1_1topi.html#af9694f5470ba2cabc19866be3b00fe8d',1,'tvm::topi::sinh()']]],
-  ['size',['Size',['../classtvm_1_1TensorTypeNode.html#a1f08dac86ae8aea81d058ef64cfd38b4',1,'tvm::TensorTypeNode::Size()'],['../classtvm_1_1meta__schedule_1_1DatabaseNode.html#aae5b9ab9f7e497654b90c23a2159a5cc',1,'tvm::meta_schedule::DatabaseNode::Size()'],['../classtvm_1_1meta__schedule_1_1PyDatabaseNode.html#a36817d04978253571fef7d01427ce9c0',1,'tvm::meta_schedule::PyDatabaseNode::Size()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1FrameBuffer.html#ae395a0f1c6e79e825aa7a244c74a5d7b',1,' [...]
+  ['size',['size',['../classtvm_1_1runtime_1_1ADT.html#af51613add20f67643684b1c7fdd5569a',1,'tvm::runtime::ADT::size()'],['../classtvm_1_1runtime_1_1ArrayNode.html#a3e88cee6eb31d0e495f7debd94b7573d',1,'tvm::runtime::ArrayNode::size()'],['../classtvm_1_1runtime_1_1Array.html#aed6387e67d18b9d5ad18f510fd600a25',1,'tvm::runtime::Array::size()'],['../classtvm_1_1runtime_1_1MapNode.html#a5c0c770f7667f911aa8bec879e3ac214',1,'tvm::runtime::MapNode::size()'],['../classtvm_1_1runtime_1_1Map.html#a [...]
   ['sizevar',['SizeVar',['../classtvm_1_1tir_1_1SizeVar.html#ac470249315d9e395ad581d35dd5dcb05',1,'tvm::tir::SizeVar::SizeVar(ObjectPtr&lt; Object &gt; n)'],['../classtvm_1_1tir_1_1SizeVar.html#a0f8cb8a92feb96343939d223db90f7cd',1,'tvm::tir::SizeVar::SizeVar(String name_hint=&quot;s&quot;, DataType t=DataType::Int(32), Span span=Span())']]],
   ['skipassert',['SkipAssert',['../namespacetvm_1_1tir_1_1transform.html#a6fdd5910b00af823071dcdddd21cd2d3',1,'tvm::tir::transform']]],
   ['slice',['Slice',['../classtvm_1_1te_1_1Tensor_1_1Slice.html#ab314819e8bcca6421e9a4f33e48578c3',1,'tvm::te::Tensor::Slice']]],
@@ -145,7 +145,7 @@ var searchData=
   ['sparse_5fto_5fdense',['sparse_to_dense',['../namespacetvm_1_1topi.html#a877e6fdffb6b6c051c29602ec6fe995c',1,'tvm::topi']]],
   ['specialize',['Specialize',['../namespacetvm_1_1tir.html#a69b6f1b0014dc6e7dd390cff746e9782',1,'tvm::tir']]],
   ['specializedcondition',['SpecializedCondition',['../classtvm_1_1te_1_1SpecializedCondition.html#a48d119ee1c6033929a5592cfc2592e60',1,'tvm::te::SpecializedCondition']]],
-  ['split',['Split',['../classtvm_1_1te_1_1Split.html#a328e0c093ce5b41ebaf33e0e80592764',1,'tvm::te::Split::Split()'],['../classtvm_1_1tir_1_1Layout.html#ad7657af7789fe040d3224c0149976bb4',1,'tvm::tir::Layout::Split()'],['../classtvm_1_1tir_1_1ScheduleNode.html#af8a330c32b06dc16c8835c76177ffa11',1,'tvm::tir::ScheduleNode::Split()'],['../classtvm_1_1auto__scheduler_1_1State.html#a5815f21fc90ba7cc379c2410c05ab54c',1,'tvm::auto_scheduler::State::split()'],['../classtvm_1_1te_1_1Stage.html#a [...]
+  ['split',['split',['../classtvm_1_1auto__scheduler_1_1State.html#a5815f21fc90ba7cc379c2410c05ab54c',1,'tvm::auto_scheduler::State::split()'],['../classtvm_1_1te_1_1Stage.html#a5a7cd562be59b68a187ad97085a3425d',1,'tvm::te::Stage::split()'],['../classtvm_1_1te_1_1Split.html#a328e0c093ce5b41ebaf33e0e80592764',1,'tvm::te::Split::Split()'],['../classtvm_1_1tir_1_1Layout.html#ad7657af7789fe040d3224c0149976bb4',1,'tvm::tir::Layout::Split()'],['../classtvm_1_1tir_1_1ScheduleNode.html#af8a330c3 [...]
   ['split_5fby_5fnparts',['split_by_nparts',['../classtvm_1_1te_1_1Stage.html#a51432f38d9ec4792a2525023179ae604',1,'tvm::te::Stage']]],
   ['split_5fsections',['split_sections',['../namespacetvm_1_1topi.html#acc643e2ed166fa2ed82a95853e145619',1,'tvm::topi']]],
   ['splitargs',['SplitArgs',['../namespacetvm_1_1relay_1_1transform.html#a2425d757b896168a109498e8d34ba960',1,'tvm::relay::transform']]],
@@ -166,7 +166,7 @@ var searchData=
   ['startmessage',['StartMessage',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#acd512b977c6dd888f90c4fd6d2b9500f',1,'tvm::runtime::micro_rpc::Session']]],
   ['startpacket',['StartPacket',['../classtvm_1_1runtime_1_1micro__rpc_1_1Framer.html#ade10d3bd3a26e3b7af881ae134e9a998',1,'tvm::runtime::micro_rpc::Framer']]],
   ['startsession',['StartSession',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a15d3f9ecb8b22bf2d330f6f0a16c5239',1,'tvm::runtime::micro_rpc::Session']]],
-  ['state',['State',['../classtvm_1_1auto__scheduler_1_1State.html#a9e8198b1f51b42cfbbee4b9f42160749',1,'tvm::auto_scheduler::State::State()'],['../classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26',1,'tvm::tir::ScheduleNode::state()']]],
+  ['state',['state',['../classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26',1,'tvm::tir::ScheduleNode::state()'],['../classtvm_1_1auto__scheduler_1_1State.html#a9e8198b1f51b42cfbbee4b9f42160749',1,'tvm::auto_scheduler::State::State()']]],
   ['stats',['Stats',['../classtvm_1_1runtime_1_1vm_1_1Executable.html#a5445bd71aa14ec97552fa099dc3bd787',1,'tvm::runtime::vm::Executable']]],
   ['stepapplytoschedule',['StepApplyToSchedule',['../namespacetvm_1_1auto__scheduler.html#ac58f7548a94b92f801b2b9a6f65bd785',1,'tvm::auto_scheduler']]],
   ['stepapplytostate',['StepApplyToState',['../namespacetvm_1_1auto__scheduler.html#a6909bc5a99d1cc8372201e9392717832',1,'tvm::auto_scheduler']]],
diff --git a/docs/reference/api/doxygen/search/functions_14.js b/docs/reference/api/doxygen/search/functions_14.js
index 7bebca7c1..e684a6493 100644
--- a/docs/reference/api/doxygen/search/functions_14.js
+++ b/docs/reference/api/doxygen/search/functions_14.js
@@ -48,7 +48,7 @@ var searchData=
   ['totupletype',['ToTupleType',['../namespacetvm_1_1relay.html#ae6757a008816e31cce4109e8dfc2bc16',1,'tvm::relay']]],
   ['touchtask',['TouchTask',['../classtvm_1_1meta__schedule_1_1TaskSchedulerNode.html#af6fa276674945d3432c129bdf9cea599',1,'tvm::meta_schedule::TaskSchedulerNode::TouchTask()'],['../classtvm_1_1meta__schedule_1_1PyTaskSchedulerNode.html#a7de09f81c8aceb580b43107f266e6b40',1,'tvm::meta_schedule::PyTaskSchedulerNode::TouchTask()']]],
   ['tovar',['ToVar',['../classtvm_1_1tir_1_1AnyNode.html#ae01ebbba2378afb6509a22de97f8fb30',1,'tvm::tir::AnyNode']]],
-  ['trace',['Trace',['../classtvm_1_1tir_1_1Trace.html#a8e09abffd0b9b1afac7b832cf16c142d',1,'tvm::tir::Trace::Trace()'],['../classtvm_1_1tir_1_1Trace.html#af79bccf1bde25efea387bb1b82dacaa6',1,'tvm::tir::Trace::Trace(Array&lt; Instruction &gt; insts, Map&lt; Instruction, ObjectRef &gt; decisions)'],['../classtvm_1_1tir_1_1ScheduleNode.html#a953bca4123b5a758adfdcd65634a5f3b',1,'tvm::tir::ScheduleNode::trace()']]],
+  ['trace',['trace',['../classtvm_1_1tir_1_1ScheduleNode.html#a953bca4123b5a758adfdcd65634a5f3b',1,'tvm::tir::ScheduleNode::trace()'],['../classtvm_1_1tir_1_1Trace.html#a8e09abffd0b9b1afac7b832cf16c142d',1,'tvm::tir::Trace::Trace()'],['../classtvm_1_1tir_1_1Trace.html#af79bccf1bde25efea387bb1b82dacaa6',1,'tvm::tir::Trace::Trace(Array&lt; Instruction &gt; insts, Map&lt; Instruction, ObjectRef &gt; decisions)']]],
   ['traced',['Traced',['../classtvm_1_1tir_1_1Schedule.html#a295d432b86621101f67b20fadb367b91',1,'tvm::tir::Schedule']]],
   ['transform',['Transform',['../classtvm_1_1te_1_1Transform.html#a51422cc2290f6b87fe61edb0db691125',1,'tvm::te::Transform']]],
   ['transform_5flayout',['transform_layout',['../classtvm_1_1te_1_1Stage.html#acec77eca6c9a4f1738a7c119d7ac2c2c',1,'tvm::te::Stage']]],
diff --git a/docs/reference/api/doxygen/search/functions_15.js b/docs/reference/api/doxygen/search/functions_15.js
index 86d8ad114..ac9788707 100644
--- a/docs/reference/api/doxygen/search/functions_15.js
+++ b/docs/reference/api/doxygen/search/functions_15.js
@@ -12,7 +12,7 @@ var searchData=
   ['unionlowerbound',['UnionLowerBound',['../namespacetvm_1_1arith.html#ab22d7fd95abb5fa372843a40e19d80c5',1,'tvm::arith']]],
   ['unionregion',['UnionRegion',['../namespacetvm_1_1arith.html#ad27c4f216e41eb8e81296fb7ec4b9453',1,'tvm::arith']]],
   ['unionregionlowerbound',['UnionRegionLowerBound',['../namespacetvm_1_1arith.html#a4c3dedfa4cba4ad39c953eb51eb83e4d',1,'tvm::arith']]],
-  ['unique',['unique',['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()'],['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()']]],
+  ['unique',['Unique',['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()'],['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()']]],
   ['unmatchedcases',['UnmatchedCases',['../namespacetvm_1_1relay.html#aa3a8cace40f8056fd6412f39c3eaa605',1,'tvm::relay']]],
   ['unravel_5findex',['unravel_index',['../namespacetvm_1_1topi.html#a8811a02532bbe3047986bf1a8449ac0e',1,'tvm::topi']]],
   ['unroll',['unroll',['../classtvm_1_1auto__scheduler_1_1State.html#aa68a9d2e226bae38a36e4be4af1d1ae4',1,'tvm::auto_scheduler::State::unroll()'],['../classtvm_1_1te_1_1Stage.html#af83ad8672660403504f472228b044b33',1,'tvm::te::Stage::unroll()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552',1,'tvm::tir::ScheduleNode::Unroll()']]],
diff --git a/docs/reference/api/doxygen/search/functions_16.js b/docs/reference/api/doxygen/search/functions_16.js
index e95fbd47f..9d33ae36f 100644
--- a/docs/reference/api/doxygen/search/functions_16.js
+++ b/docs/reference/api/doxygen/search/functions_16.js
@@ -8,7 +8,7 @@ var searchData=
   ['vector',['Vector',['../classtvm_1_1arith_1_1IntSet.html#a29b6f1e60f4b328fcfabc514e0c10f17',1,'tvm::arith::IntSet']]],
   ['vectorcombine',['vectorcombine',['../namespacetvm_1_1tir_1_1builtin.html#a30dff65bc2c142b57fae7f60e378ff43',1,'tvm::tir::builtin']]],
   ['vectorhigh',['vectorhigh',['../namespacetvm_1_1tir_1_1builtin.html#a45bf65ca7ca01d2016e0b609117d7e25',1,'tvm::tir::builtin']]],
-  ['vectorize',['Vectorize',['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()'],['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()']]],
+  ['vectorize',['vectorize',['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()']]],
   ['vectorizeloop',['VectorizeLoop',['../namespacetvm_1_1tir_1_1transform.html#af3cecb50a8b8fc8021f6a87bc27587da',1,'tvm::tir::transform']]],
   ['vectorjacobianproduct',['VectorJacobianProduct',['../namespacetvm_1_1te.html#a547183f5a311af53ab598faba423fd64',1,'tvm::te']]],
   ['vectorlow',['vectorlow',['../namespacetvm_1_1tir_1_1builtin.html#a7ed64a9fb0a7f575fc63e1e0395e96a6',1,'tvm::tir::builtin']]],
diff --git a/docs/reference/api/doxygen/search/functions_6.js b/docs/reference/api/doxygen/search/functions_6.js
index 4765c8aea..b940bec7f 100644
--- a/docs/reference/api/doxygen/search/functions_6.js
+++ b/docs/reference/api/doxygen/search/functions_6.js
@@ -68,6 +68,7 @@ var searchData=
   ['fromdlpack',['FromDLPack',['../classtvm_1_1runtime_1_1NDArray.html#abec485628a0ca451b668c42fd8fa691a',1,'tvm::runtime::NDArray']]],
   ['fromexpr',['FromExpr',['../classtvm_1_1IRModule.html#a59099426f65dbeac227e51f8864e322a',1,'tvm::IRModule']]],
   ['fromexprincontext',['FromExprInContext',['../classtvm_1_1IRModule.html#a1cc91fc2b2adaca5a4dcfc14baf28c27',1,'tvm::IRModule']]],
+  ['fromexternaldltensor',['FromExternalDLTensor',['../classtvm_1_1runtime_1_1NDArray.html#a356d1886b24da68c35a0d0b826c9359e',1,'tvm::runtime::NDArray']]],
   ['fromfunc',['FromFunc',['../classtvm_1_1tir_1_1IndexMap.html#afa04f25f10b1dac139df9a1b34598cbb',1,'tvm::tir::IndexMap']]],
   ['fromjson',['FromJSON',['../classtvm_1_1meta__schedule_1_1ArgInfo.html#afc2cfa9fdf0bcedc79e90f07a596f74a',1,'tvm::meta_schedule::ArgInfo::FromJSON()'],['../classtvm_1_1meta__schedule_1_1TensorInfo.html#a1d4166d5ac2777b955c7263ebcaa068a',1,'tvm::meta_schedule::TensorInfo::FromJSON()'],['../classtvm_1_1meta__schedule_1_1Workload.html#a1c3076818c9a20d8e7c675a8ce58f8f3',1,'tvm::meta_schedule::Workload::FromJSON()'],['../classtvm_1_1meta__schedule_1_1TuningRecord.html#aabec8835c7178c808063 [...]
   ['fromminextent',['FromMinExtent',['../classtvm_1_1arith_1_1IntSet.html#a5eaf5c75ebfc33cf04373bc2d0071465',1,'tvm::arith::IntSet::FromMinExtent()'],['../classtvm_1_1Range.html#a91e7301ca1d135ca5f8ed199efbb9818',1,'tvm::Range::FromMinExtent()']]],
diff --git a/docs/reference/api/doxygen/search/functions_d.js b/docs/reference/api/doxygen/search/functions_d.js
index e63ddf166..6b419f55d 100644
--- a/docs/reference/api/doxygen/search/functions_d.js
+++ b/docs/reference/api/doxygen/search/functions_d.js
@@ -31,7 +31,7 @@ var searchData=
   ['matchrange',['MatchRange',['../classtvm_1_1arith_1_1IntSet.html#a2f2999336fbba4f436b66bdddce5c57a',1,'tvm::arith::IntSet']]],
   ['matmul',['matmul',['../namespacetvm_1_1topi.html#adae7dcb7e951109ba72192202d182994',1,'tvm::topi']]],
   ['matrix_5fset_5fdiag',['matrix_set_diag',['../namespacetvm_1_1topi.html#aead477c6c9d4f4589d22b8acff82040c',1,'tvm::topi']]],
-  ['max',['Max',['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
+  ['max',['max',['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
   ['max_5fvalue',['max_value',['../namespacetvm.html#a4f1398024c0af23699447ef910b654b8',1,'tvm']]],
   ['maxconcurrency',['MaxConcurrency',['../namespacetvm_1_1runtime_1_1threading.html#af8c1c389a74e67bcc3680555288219f8',1,'tvm::runtime::threading']]],
   ['maximum',['maximum',['../namespacetvm_1_1topi.html#afd64bc3e27dfc97002d3add5d7ce4174',1,'tvm::topi::maximum(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a5338e9297463bc745027fca67daa2ebb',1,'tvm::topi::maximum(const tvm::te::Tensor &amp;A, const tvm::te::Tensor &amp;B, std::string name=&quot;T_&quot; &quot;maximum&quot;, std::string tag=kBroadcast)'],['../namespacetvm_1_1topi.html#a4076a8d6a2b243c548d741e9f6bcfe69',1,'tvm::topi::maximum(con [...]
@@ -62,7 +62,7 @@ var searchData=
   ['mixedmodemutator',['MixedModeMutator',['../classtvm_1_1relay_1_1MixedModeMutator.html#add94673de95267df385059ac4cc9c519',1,'tvm::relay::MixedModeMutator']]],
   ['mixedmodevisitor',['MixedModeVisitor',['../classtvm_1_1relay_1_1MixedModeVisitor.html#ad98abd46ba60b173c3e33a96aa3ef9b4',1,'tvm::relay::MixedModeVisitor']]],
   ['mixedmodulepassmanager',['MixedModulePassManager',['../namespacetvm.html#abc01352eff102d4902632d097adc0e08',1,'tvm']]],
-  ['mod',['mod',['../classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12',1,'tvm::tir::ScheduleNode::mod()'],['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../namespacetvm_1_1topi.html#aaa95d3ad68932ab206efbe0a326db6a2',1,'tvm::topi::mod(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a4eb4b5a58cf4c5dbbdd4413cfd166882',1,'tvm::topi::mod(const tvm::te::Tensor &amp;A, const tvm::te: [...]
+  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12',1,'tvm::tir::ScheduleNode::mod()'],['../namespacetvm_1_1topi.html#aaa95d3ad68932ab206efbe0a326db6a2',1,'tvm::topi::mod(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a4eb4b5a58cf4c5dbbdd4413cfd166882',1,'tvm::topi::mod(const tvm::te::Tensor &amp;A, const tvm::te: [...]
   ['modularset',['ModularSet',['../classtvm_1_1arith_1_1ModularSet.html#a9f54896d98169246c6a24cc338fde500',1,'tvm::arith::ModularSet']]],
   ['module',['Module',['../classtvm_1_1runtime_1_1Module.html#abfbc619b3b3166d63ec52e399c24bed9',1,'tvm::runtime::Module::Module()'],['../classtvm_1_1runtime_1_1Module.html#abd1380b3f813c2b6acefca3aaef425f4',1,'tvm::runtime::Module::Module(ObjectPtr&lt; Object &gt; n)']]],
   ['move',['Move',['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a162dc8d73dc2306f066c3ee013ff096f',1,'tvm::runtime::vm::Instruction']]],
diff --git a/docs/reference/api/doxygen/search/functions_e.js b/docs/reference/api/doxygen/search/functions_e.js
index 2dcf7a441..8eb037fef 100644
--- a/docs/reference/api/doxygen/search/functions_e.js
+++ b/docs/reference/api/doxygen/search/functions_e.js
@@ -15,6 +15,7 @@ var searchData=
   ['neg',['neg',['../namespacetvm.html#a5cd85b156fb31f75f91c8a5c012f8a66',1,'tvm']]],
   ['negative',['negative',['../namespacetvm_1_1topi.html#af6b3e60333fce92bcf0930e45683a8f6',1,'tvm::topi']]],
   ['new',['New',['../classtvm_1_1runtime_1_1SimpleObjAllocator_1_1Handler.html#afedd0ba3dc8dc82c7566bb9120a7c56d',1,'tvm::runtime::SimpleObjAllocator::Handler::New()'],['../classtvm_1_1runtime_1_1SimpleObjAllocator_1_1ArrayHandler.html#a310471cff82c5d0836f65ec7f199e621',1,'tvm::runtime::SimpleObjAllocator::ArrayHandler::New()']]],
+  ['newfromdltensor',['NewFromDLTensor',['../classtvm_1_1runtime_1_1NDArray.html#afb6060bb96dad082c1deca26e6b58ae2',1,'tvm::runtime::NDArray']]],
   ['nextafter',['nextafter',['../namespacetvm.html#a96d86ba91e4855c84879ba886465cacf',1,'tvm']]],
   ['nexttaskid',['NextTaskId',['../classtvm_1_1meta__schedule_1_1TaskSchedulerNode.html#a079e2964ca86b5c32564140efa3e5626',1,'tvm::meta_schedule::TaskSchedulerNode::NextTaskId()'],['../classtvm_1_1meta__schedule_1_1PyTaskSchedulerNode.html#a23752f62706ef3f0bfac98fb203e5062',1,'tvm::meta_schedule::PyTaskSchedulerNode::NextTaskId()']]],
   ['nll_5floss',['nll_loss',['../namespacetvm_1_1topi.html#aeb1547800d4b7625326a176ca1dec6e0',1,'tvm::topi']]],
diff --git a/docs/reference/api/doxygen/structural__hash_8h_source.html b/docs/reference/api/doxygen/structural__hash_8h_source.html
index 1ebe79500..e53f78e2f 100644
--- a/docs/reference/api/doxygen/structural__hash_8h_source.html
+++ b/docs/reference/api/doxygen/structural__hash_8h_source.html
@@ -84,7 +84,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1DataType_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1DataType.html">tvm::runtime::DataType</a></div><div class="ttdoc">Runtime primitive data type. </div><div class="ttdef"><b>Definition:</b> data_type.h:41</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1DataType_html_acb927f3f6ae5116e8f439159fafbc04e"><div class="ttname"><a href="classtvm_1_1runtime_1_1DataType.html#acb927f3f6ae5116e8f439159fafbc04e">tvm::runtime::DataType::bits</a></div><div class="ttdeci">int bits() const</div><div class="ttdef"><b>Definition:</b> data_type.h:83</div></div>
 <div class="ttc" id="classtvm_1_1BaseValueHash_html"><div class="ttname"><a href="classtvm_1_1BaseValueHash.html">tvm::BaseValueHash</a></div><div class="ttdoc">Hash definition of base value classes. </div><div class="ttdef"><b>Definition:</b> structural_hash.h:38</div></div>
-<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:261</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1NDArray_1_1Container_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1NDArray_1_1Container.html">tvm::runtime::NDArray::Container</a></div><div class="ttdoc">Object container class that backs NDArray. </div><div class="ttdef"><b>Definition:</b> ndarray.h:279</div></div>
 <div class="ttc" id="classtvm_1_1BaseValueHash_html_a142aeac49e5beda0a8fe0ee39b81a5a4"><div class="ttname"><a href="classtvm_1_1BaseValueHash.html#a142aeac49e5beda0a8fe0ee39b81a5a4">tvm::BaseValueHash::operator()</a></div><div class="ttdeci">size_t operator()(const int &amp;key) const</div><div class="ttdef"><b>Definition:</b> structural_hash.h:46</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1DataType_html_a7a67295643b82bbe37cf36e6f69e8323"><div class="ttname"><a href="classtvm_1_1runtime_1_1DataType.html#a7a67295643b82bbe37cf36e6f69e8323">tvm::runtime::DataType::lanes</a></div><div class="ttdeci">int lanes() const</div><div class="ttdef"><b>Definition:</b> data_type.h:87</div></div>
 <div class="ttc" id="namespacetvm_1_1te_html_ae0c71f84710b436cbe0b32289d0838f4"><div class="ttname"><a href="namespacetvm_1_1te.html#ae0c71f84710b436cbe0b32289d0838f4">tvm::te::var</a></div><div class="ttdeci">Var var(std::string name_hint, DataType t=DataType::Int(32))</div><div class="ttdoc">Construct a new Var expression. </div></div>
diff --git a/docs/reference/api/python/auto_scheduler.html b/docs/reference/api/python/auto_scheduler.html
index ca2da0eab..0aa40ce51 100644
--- a/docs/reference/api/python/auto_scheduler.html
+++ b/docs/reference/api/python/auto_scheduler.html
@@ -1713,7 +1713,7 @@ Can be the a function or the function name.</p></li>
 
 <dl class="py function">
 <dt class="sig sig-object py" id="tvm.auto_scheduler.auto_schedule">
-<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
+<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
 <dd><p>THIS API IS DEPRECATED.</p>
 <p>Run auto scheduling search for a task.</p>
 <dl class="field-list simple">
@@ -1750,7 +1750,7 @@ the initial naive schedule (state).</p>
 
 <dl class="py class">
 <dt class="sig sig-object py" id="tvm.auto_scheduler.SketchPolicy">
-<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
+<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
 <dd><p>The search policy that searches in a hierarchical search space defined by sketches.
 The policy randomly samples programs from the space defined by sketches and use evolutionary
 search to fine-tune them.</p>
diff --git a/docs/reference/api/typedoc/classes/bytestreamreader.html b/docs/reference/api/typedoc/classes/bytestreamreader.html
index 47c9ec134..c2501272f 100644
--- a/docs/reference/api/typedoc/classes/bytestreamreader.html
+++ b/docs/reference/api/typedoc/classes/bytestreamreader.html
@@ -119,7 +119,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -141,7 +141,7 @@
 					<div class="tsd-signature tsd-kind-icon">bytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Uint8Array</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -151,7 +151,7 @@
 					<div class="tsd-signature tsd-kind-icon">offset<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span><span class="tsd-signature-symbol"> = 0</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L42">rpc_server.ts:42</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L42">rpc_server.ts:42</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -168,7 +168,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L63">rpc_server.ts:63</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L63">rpc_server.ts:63</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">Uint8Array</span></h4>
@@ -185,7 +185,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L49">rpc_server.ts:49</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L49">rpc_server.ts:49</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -202,7 +202,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L57">rpc_server.ts:57</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L57">rpc_server.ts:57</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
diff --git a/docs/reference/api/typedoc/classes/cachedcallstack.html b/docs/reference/api/typedoc/classes/cachedcallstack.html
index 4fc6a3c8c..a505ab64d 100644
--- a/docs/reference/api/typedoc/classes/cachedcallstack.html
+++ b/docs/reference/api/typedoc/classes/cachedcallstack.html
@@ -144,7 +144,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L223">memory.ts:223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L223">memory.ts:223</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -172,7 +172,7 @@
 					<div class="tsd-signature tsd-kind-icon">temp<wbr>Args<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Array</span><span class="tsd-signature-symbol">&lt;</span><a href="../interfaces/disposable.html" class="tsd-signature-type">Disposable</a><span class="tsd-signature-symbol">&gt;</span><span class="tsd-signature-symbol"> = []</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L208">memory.ts:208</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L208">memory.ts:208</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -194,7 +194,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L312">memory.ts:312</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L312">memory.ts:312</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -226,7 +226,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L284">memory.ts:284</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L284">memory.ts:284</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -262,7 +262,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L388">memory.ts:388</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L388">memory.ts:388</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -300,7 +300,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L376">memory.ts:376</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L376">memory.ts:376</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -340,7 +340,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L267">memory.ts:267</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L267">memory.ts:267</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -373,7 +373,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L243">memory.ts:243</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L243">memory.ts:243</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -390,7 +390,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L321">memory.ts:321</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L321">memory.ts:321</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -422,7 +422,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L252">memory.ts:252</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L252">memory.ts:252</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -444,7 +444,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L359">memory.ts:359</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L359">memory.ts:359</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -470,7 +470,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L342">memory.ts:342</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L342">memory.ts:342</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -496,7 +496,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L350">memory.ts:350</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L350">memory.ts:350</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -522,7 +522,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L326">memory.ts:326</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L326">memory.ts:326</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -548,7 +548,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L363">memory.ts:363</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L363">memory.ts:363</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -574,7 +574,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L346">memory.ts:346</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L346">memory.ts:346</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -600,7 +600,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L334">memory.ts:334</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L334">memory.ts:334</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
diff --git a/docs/reference/api/typedoc/classes/dldatatype.html b/docs/reference/api/typedoc/classes/dldatatype.html
index c6770c89b..838864c6a 100644
--- a/docs/reference/api/typedoc/classes/dldatatype.html
+++ b/docs/reference/api/typedoc/classes/dldatatype.html
@@ -119,7 +119,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L262">runtime.ts:262</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -147,7 +147,7 @@
 					<div class="tsd-signature tsd-kind-icon">bits<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L260">runtime.ts:260</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L260">runtime.ts:260</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -162,7 +162,7 @@
 					<div class="tsd-signature tsd-kind-icon">code<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L258">runtime.ts:258</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L258">runtime.ts:258</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -177,7 +177,7 @@
 					<div class="tsd-signature tsd-kind-icon">lanes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L262">runtime.ts:262</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -199,7 +199,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L279">runtime.ts:279</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L279">runtime.ts:279</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -216,7 +216,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L270">runtime.ts:270</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L270">runtime.ts:270</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/dldevice.html b/docs/reference/api/typedoc/classes/dldevice.html
index 9197daf9f..97d3a3b5f 100644
--- a/docs/reference/api/typedoc/classes/dldevice.html
+++ b/docs/reference/api/typedoc/classes/dldevice.html
@@ -118,7 +118,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L202">runtime.ts:202</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L202">runtime.ts:202</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -146,7 +146,7 @@
 					<div class="tsd-signature tsd-kind-icon">device<wbr>Id<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L200">runtime.ts:200</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L200">runtime.ts:200</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -161,7 +161,7 @@
 					<div class="tsd-signature tsd-kind-icon">device<wbr>Type<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L198">runtime.ts:198</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L198">runtime.ts:198</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -183,7 +183,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L223">runtime.ts:223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L223">runtime.ts:223</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -205,7 +205,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L230">runtime.ts:230</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L230">runtime.ts:230</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/environment.html b/docs/reference/api/typedoc/classes/environment.html
index f34e056d6..033b42910 100644
--- a/docs/reference/api/typedoc/classes/environment.html
+++ b/docs/reference/api/typedoc/classes/environment.html
@@ -125,7 +125,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L86">environment.ts:86</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L86">environment.ts:86</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -169,7 +169,7 @@
 					<aside class="tsd-sources">
 						<p>Implementation of <a href="../interfaces/libraryprovider.html">LibraryProvider</a>.<a href="../interfaces/libraryprovider.html#imports">imports</a></p>
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L70">environment.ts:70</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L70">environment.ts:70</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -179,7 +179,7 @@
 					<div class="tsd-signature tsd-kind-icon">logger<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>msg<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> =&gt; </span><span class="tsd-signature-type">void</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L69">environment.ts:69</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L69">environment.ts:69</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-type-declaration">
@@ -210,7 +210,7 @@
 					<div class="tsd-signature tsd-kind-icon">packedCFunc<wbr>Table<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Array</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">ctypes.FTVMWasmPackedCFunc</span><span class="tsd-signature-symbol"> | </span><span class="tsd-signature-type">undefined</span><span class="tsd-signature-symbol">&gt;</span><span class="tsd-signature-symbol"> = [undefined,]</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L78">environment.ts:78</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L78">environment.ts:78</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -228,7 +228,7 @@
 					<div class="tsd-signature tsd-kind-icon">packedCFunc<wbr>Table<wbr>Free<wbr>Id<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Array</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">number</span><span class="tsd-signature-symbol">&gt;</span><span class="tsd-signature-symbol"> = []</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L84">environment.ts:84</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L84">environment.ts:84</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -250,7 +250,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/environment.ts#L105">environment.ts:105</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/environment.ts#L105">environment.ts:105</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/ffilibrary.html b/docs/reference/api/typedoc/classes/ffilibrary.html
index ba230786c..6d0c29202 100644
--- a/docs/reference/api/typedoc/classes/ffilibrary.html
+++ b/docs/reference/api/typedoc/classes/ffilibrary.html
@@ -131,7 +131,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L49">runtime.ts:49</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L49">runtime.ts:49</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -156,7 +156,7 @@
 					<div class="tsd-signature tsd-kind-icon">exports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">, </span><span class="tsd-signature-type">Function</span><span class="tsd-signature-symbol">&gt;</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L46">runtime.ts:46</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L46">runtime.ts:46</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -166,7 +166,7 @@
 					<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <a href="memory.html" class="tsd-signature-type">Memory</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L45">runtime.ts:45</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L45">runtime.ts:45</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -176,7 +176,7 @@
 					<div class="tsd-signature tsd-kind-icon">wasm32<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">boolean</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L44">runtime.ts:44</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L44">runtime.ts:44</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -186,7 +186,7 @@
 					<div class="tsd-signature tsd-kind-icon">webGPUContext<span class="tsd-signature-symbol">:</span> <a href="webgpucontext.html" class="tsd-signature-type">WebGPUContext</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L47">runtime.ts:47</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L47">runtime.ts:47</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -203,7 +203,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L76">runtime.ts:76</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L76">runtime.ts:76</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -226,7 +226,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L66">runtime.ts:66</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L66">runtime.ts:66</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -243,7 +243,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L84">runtime.ts:84</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L84">runtime.ts:84</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <a href="cachedcallstack.html" class="tsd-signature-type">CachedCallStack</a></h4>
@@ -260,7 +260,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L95">runtime.ts:95</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L95">runtime.ts:95</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -283,7 +283,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L72">runtime.ts:72</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L72">runtime.ts:72</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
diff --git a/docs/reference/api/typedoc/classes/graphexecutor.html b/docs/reference/api/typedoc/classes/graphexecutor.html
index 0c9772682..b0542d95c 100644
--- a/docs/reference/api/typedoc/classes/graphexecutor.html
+++ b/docs/reference/api/typedoc/classes/graphexecutor.html
@@ -130,7 +130,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L583">runtime.ts:583</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L583">runtime.ts:583</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -162,7 +162,7 @@
 					<div class="tsd-signature tsd-kind-icon">module<span class="tsd-signature-symbol">:</span> <a href="module.html" class="tsd-signature-type">Module</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L579">runtime.ts:579</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L579">runtime.ts:579</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -179,7 +179,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L654">runtime.ts:654</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L654">runtime.ts:654</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -224,7 +224,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L597">runtime.ts:597</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L597">runtime.ts:597</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -241,7 +241,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L631">runtime.ts:631</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L631">runtime.ts:631</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -279,7 +279,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L644">runtime.ts:644</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L644">runtime.ts:644</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -310,7 +310,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L621">runtime.ts:621</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L621">runtime.ts:621</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -332,7 +332,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L609">runtime.ts:609</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L609">runtime.ts:609</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/instance.html b/docs/reference/api/typedoc/classes/instance.html
index 4b5b90940..b0786efd7 100644
--- a/docs/reference/api/typedoc/classes/instance.html
+++ b/docs/reference/api/typedoc/classes/instance.html
@@ -139,7 +139,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L692">runtime.ts:692</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L692">runtime.ts:692</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -202,7 +202,7 @@
 					<div class="tsd-signature tsd-kind-icon">exports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">, </span><span class="tsd-signature-type">Function</span><span class="tsd-signature-symbol">&gt;</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L684">runtime.ts:684</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L684">runtime.ts:684</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -212,7 +212,7 @@
 					<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <a href="memory.html" class="tsd-signature-type">Memory</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L683">runtime.ts:683</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L683">runtime.ts:683</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -229,7 +229,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L932">runtime.ts:932</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L932">runtime.ts:932</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -260,7 +260,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L994">runtime.ts:994</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L994">runtime.ts:994</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -303,7 +303,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L924">runtime.ts:924</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L924">runtime.ts:924</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -341,7 +341,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L732">runtime.ts:732</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L732">runtime.ts:732</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -358,7 +358,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L952">runtime.ts:952</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L952">runtime.ts:952</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -402,7 +402,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L816">runtime.ts:816</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L816">runtime.ts:816</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -434,7 +434,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L1033">runtime.ts:1033</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L1033">runtime.ts:1033</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -465,7 +465,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L846">runtime.ts:846</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L846">runtime.ts:846</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -497,7 +497,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L750">runtime.ts:750</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L750">runtime.ts:750</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -520,7 +520,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L1013">runtime.ts:1013</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L1013">runtime.ts:1013</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -568,7 +568,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L789">runtime.ts:789</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L789">runtime.ts:789</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -608,7 +608,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L914">runtime.ts:914</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L914">runtime.ts:914</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -646,7 +646,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L1134">runtime.ts:1134</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L1134">runtime.ts:1134</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -698,7 +698,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L740">runtime.ts:740</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L740">runtime.ts:740</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -722,7 +722,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L868">runtime.ts:868</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L868">runtime.ts:868</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -754,7 +754,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L857">runtime.ts:857</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L857">runtime.ts:857</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -786,7 +786,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L940">runtime.ts:940</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L940">runtime.ts:940</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/memory.html b/docs/reference/api/typedoc/classes/memory.html
index 30f9567ee..075ead122 100644
--- a/docs/reference/api/typedoc/classes/memory.html
+++ b/docs/reference/api/typedoc/classes/memory.html
@@ -130,7 +130,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L40">memory.ts:40</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L40">memory.ts:40</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -152,7 +152,7 @@
 					<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Memory</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L32">memory.ts:32</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L32">memory.ts:32</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -162,7 +162,7 @@
 					<div class="tsd-signature tsd-kind-icon">wasm32<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">boolean</span><span class="tsd-signature-symbol"> = true</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L33">memory.ts:33</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L33">memory.ts:33</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -179,7 +179,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L154">memory.ts:154</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L154">memory.ts:154</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -210,7 +210,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L90">memory.ts:90</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L90">memory.ts:90</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -233,7 +233,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L97">memory.ts:97</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L97">memory.ts:97</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -256,7 +256,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L74">memory.ts:74</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L74">memory.ts:74</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -279,7 +279,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L81">memory.ts:81</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L81">memory.ts:81</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -302,7 +302,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L104">memory.ts:104</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L104">memory.ts:104</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -325,7 +325,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L132">memory.ts:132</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L132">memory.ts:132</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -362,7 +362,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L145">memory.ts:145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L145">memory.ts:145</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -393,7 +393,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L60">memory.ts:60</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L60">memory.ts:60</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -416,7 +416,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L67">memory.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L67">memory.ts:67</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -439,7 +439,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L53">memory.ts:53</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L53">memory.ts:53</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -462,7 +462,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L114">memory.ts:114</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L114">memory.ts:114</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -485,7 +485,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L124">memory.ts:124</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L124">memory.ts:124</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -502,7 +502,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/memory.ts#L175">memory.ts:175</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/memory.ts#L175">memory.ts:175</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/module.html b/docs/reference/api/typedoc/classes/module.html
index e7c139c04..1ca6ce1cc 100644
--- a/docs/reference/api/typedoc/classes/module.html
+++ b/docs/reference/api/typedoc/classes/module.html
@@ -124,7 +124,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L504">runtime.ts:504</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L504">runtime.ts:504</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -170,7 +170,7 @@
 					<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L502">runtime.ts:502</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L502">runtime.ts:502</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -187,7 +187,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L516">runtime.ts:516</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L516">runtime.ts:516</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -204,7 +204,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L530">runtime.ts:530</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L530">runtime.ts:530</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -236,7 +236,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L561">runtime.ts:561</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L561">runtime.ts:561</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/ndarray.html b/docs/reference/api/typedoc/classes/ndarray.html
index 55c4c21d1..19639f44d 100644
--- a/docs/reference/api/typedoc/classes/ndarray.html
+++ b/docs/reference/api/typedoc/classes/ndarray.html
@@ -130,7 +130,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L304">runtime.ts:304</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L304">runtime.ts:304</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -158,7 +158,7 @@
 					<div class="tsd-signature tsd-kind-icon">device<span class="tsd-signature-symbol">:</span> <a href="dldevice.html" class="tsd-signature-type">DLDevice</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L297">runtime.ts:297</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L297">runtime.ts:297</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -173,7 +173,7 @@
 					<div class="tsd-signature tsd-kind-icon">dtype<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L293">runtime.ts:293</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L293">runtime.ts:293</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -188,7 +188,7 @@
 					<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L289">runtime.ts:289</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L289">runtime.ts:289</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -203,7 +203,7 @@
 					<div class="tsd-signature tsd-kind-icon">ndim<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L291">runtime.ts:291</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L291">runtime.ts:291</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -218,7 +218,7 @@
 					<div class="tsd-signature tsd-kind-icon">shape<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Array</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">number</span><span class="tsd-signature-symbol">&gt;</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L295">runtime.ts:295</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L295">runtime.ts:295</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -240,7 +240,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L370">runtime.ts:370</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L370">runtime.ts:370</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -273,7 +273,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L414">runtime.ts:414</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L414">runtime.ts:414</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -305,7 +305,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L355">runtime.ts:355</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L355">runtime.ts:355</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -322,7 +322,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L474">runtime.ts:474</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L474">runtime.ts:474</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -346,7 +346,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L443">runtime.ts:443</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L443">runtime.ts:443</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/packedfunccell.html b/docs/reference/api/typedoc/classes/packedfunccell.html
index 063097767..7a4c997c7 100644
--- a/docs/reference/api/typedoc/classes/packedfunccell.html
+++ b/docs/reference/api/typedoc/classes/packedfunccell.html
@@ -122,7 +122,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L158">runtime.ts:158</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L158">runtime.ts:158</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -147,7 +147,7 @@
 					<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L157">runtime.ts:157</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L157">runtime.ts:157</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -164,7 +164,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L165">runtime.ts:165</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L165">runtime.ts:165</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
diff --git a/docs/reference/api/typedoc/classes/rpcserver.html b/docs/reference/api/typedoc/classes/rpcserver.html
index 28618e272..6b6b9baf5 100644
--- a/docs/reference/api/typedoc/classes/rpcserver.html
+++ b/docs/reference/api/typedoc/classes/rpcserver.html
@@ -115,7 +115,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L92">rpc_server.ts:92</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L92">rpc_server.ts:92</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -176,7 +176,7 @@
 					<div class="tsd-signature tsd-kind-icon">get<wbr>Imports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> =&gt; </span><span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol">&lt;</span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">, </span><span class="tsd-signature-type">unknown</span><span class="tsd-signat [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L82">rpc_server.ts:82</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L82">rpc_server.ts:82</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-type-declaration">
@@ -201,7 +201,7 @@
 					<div class="tsd-signature tsd-kind-icon">key<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L78">rpc_server.ts:78</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L78">rpc_server.ts:78</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -211,7 +211,7 @@
 					<div class="tsd-signature tsd-kind-icon">logger<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>msg<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> =&gt; </span><span class="tsd-signature-type">void</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L81">rpc_server.ts:81</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L81">rpc_server.ts:81</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-type-declaration">
@@ -242,7 +242,7 @@
 					<div class="tsd-signature tsd-kind-icon">socket<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">WebSocket</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L79">rpc_server.ts:79</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L79">rpc_server.ts:79</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -252,7 +252,7 @@
 					<div class="tsd-signature tsd-kind-icon">state<span class="tsd-signature-symbol">:</span> <a href="../enums/rpcserverstate.html" class="tsd-signature-type">RPCServerState</a><span class="tsd-signature-symbol"> = RPCServerState.InitHeader</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L80">rpc_server.ts:80</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L80">rpc_server.ts:80</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -262,7 +262,7 @@
 					<div class="tsd-signature tsd-kind-icon">url<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L77">rpc_server.ts:77</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L77">rpc_server.ts:77</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/classes/scalar.html b/docs/reference/api/typedoc/classes/scalar.html
index 387fa888b..5a7b60bc6 100644
--- a/docs/reference/api/typedoc/classes/scalar.html
+++ b/docs/reference/api/typedoc/classes/scalar.html
@@ -112,7 +112,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L145">runtime.ts:145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L145">runtime.ts:145</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -137,7 +137,7 @@
 					<div class="tsd-signature tsd-kind-icon">dtype<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L145">runtime.ts:145</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L145">runtime.ts:145</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -152,7 +152,7 @@
 					<div class="tsd-signature tsd-kind-icon">value<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L143">runtime.ts:143</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L143">runtime.ts:143</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/webgpucontext.html b/docs/reference/api/typedoc/classes/webgpucontext.html
index 235923c14..a0426c9cc 100644
--- a/docs/reference/api/typedoc/classes/webgpucontext.html
+++ b/docs/reference/api/typedoc/classes/webgpucontext.html
@@ -120,7 +120,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L57">webgpu.ts:57</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L57">webgpu.ts:57</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -145,7 +145,7 @@
 					<div class="tsd-signature tsd-kind-icon">device<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">GPUDevice</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L50">webgpu.ts:50</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L50">webgpu.ts:50</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -155,7 +155,7 @@
 					<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <a href="memory.html" class="tsd-signature-type">Memory</a></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L51">webgpu.ts:51</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L51">webgpu.ts:51</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -172,7 +172,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L84">webgpu.ts:84</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L84">webgpu.ts:84</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -209,7 +209,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L170">webgpu.ts:170</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L170">webgpu.ts:170</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -238,7 +238,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/webgpu.ts#L67">webgpu.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/webgpu.ts#L67">webgpu.ts:67</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/enums/argtypecode.html b/docs/reference/api/typedoc/enums/argtypecode.html
index 78fe5d523..a5d0a19c1 100644
--- a/docs/reference/api/typedoc/enums/argtypecode.html
+++ b/docs/reference/api/typedoc/enums/argtypecode.html
@@ -106,7 +106,7 @@
 					<div class="tsd-signature tsd-kind-icon">DLDevice<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 6</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L220">ctypes.ts:220</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L220">ctypes.ts:220</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -116,7 +116,7 @@
 					<div class="tsd-signature tsd-kind-icon">Float<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 2</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L216">ctypes.ts:216</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L216">ctypes.ts:216</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -126,7 +126,7 @@
 					<div class="tsd-signature tsd-kind-icon">Int<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 0</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L214">ctypes.ts:214</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L214">ctypes.ts:214</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -136,7 +136,7 @@
 					<div class="tsd-signature tsd-kind-icon">Null<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 4</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L218">ctypes.ts:218</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L218">ctypes.ts:218</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -146,7 +146,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMBytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 12</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L226">ctypes.ts:226</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L226">ctypes.ts:226</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -156,7 +156,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMDLTensor<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 7</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L221">ctypes.ts:221</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L221">ctypes.ts:221</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -166,7 +166,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMData<wbr>Type<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 5</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L219">ctypes.ts:219</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L219">ctypes.ts:219</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -176,7 +176,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMModule<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 9</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L223">ctypes.ts:223</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L223">ctypes.ts:223</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -186,7 +186,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMNDArray<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 13</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L227">ctypes.ts:227</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L227">ctypes.ts:227</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -196,7 +196,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMObject<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 8</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L222">ctypes.ts:222</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L222">ctypes.ts:222</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -206,7 +206,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMObjectRValue<wbr>Ref<wbr>Arg<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 14</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L228">ctypes.ts:228</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L228">ctypes.ts:228</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -216,7 +216,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMOpaque<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 3</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L217">ctypes.ts:217</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L217">ctypes.ts:217</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -226,7 +226,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMPacked<wbr>Func<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 10</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L224">ctypes.ts:224</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L224">ctypes.ts:224</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -236,7 +236,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMStr<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 11</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L225">ctypes.ts:225</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L225">ctypes.ts:225</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -246,7 +246,7 @@
 					<div class="tsd-signature tsd-kind-icon">UInt<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 1</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L215">ctypes.ts:215</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L215">ctypes.ts:215</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/enums/aynccallbackcode.html b/docs/reference/api/typedoc/enums/aynccallbackcode.html
index acade5434..8a3bf32e3 100644
--- a/docs/reference/api/typedoc/enums/aynccallbackcode.html
+++ b/docs/reference/api/typedoc/enums/aynccallbackcode.html
@@ -93,7 +93,7 @@
 					<div class="tsd-signature tsd-kind-icon">k<wbr>Exception<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 5</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L676">runtime.ts:676</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L676">runtime.ts:676</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -103,7 +103,7 @@
 					<div class="tsd-signature tsd-kind-icon">k<wbr>Return<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 4</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L675">runtime.ts:675</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L675">runtime.ts:675</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/enums/dldatatypecode.html b/docs/reference/api/typedoc/enums/dldatatypecode.html
index 2032cd9eb..bd5376a0b 100644
--- a/docs/reference/api/typedoc/enums/dldatatypecode.html
+++ b/docs/reference/api/typedoc/enums/dldatatypecode.html
@@ -95,7 +95,7 @@
 					<div class="tsd-signature tsd-kind-icon">Float<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 2</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L242">runtime.ts:242</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L242">runtime.ts:242</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -105,7 +105,7 @@
 					<div class="tsd-signature tsd-kind-icon">Int<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 0</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L240">runtime.ts:240</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L240">runtime.ts:240</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -115,7 +115,7 @@
 					<div class="tsd-signature tsd-kind-icon">Opaque<wbr>Handle<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 3</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L243">runtime.ts:243</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L243">runtime.ts:243</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -125,7 +125,7 @@
 					<div class="tsd-signature tsd-kind-icon">UInt<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 1</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/runtime.ts#L241">runtime.ts:241</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/runtime.ts#L241">runtime.ts:241</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/enums/rpcserverstate.html b/docs/reference/api/typedoc/enums/rpcserverstate.html
index aa420567c..7aefc624c 100644
--- a/docs/reference/api/typedoc/enums/rpcserverstate.html
+++ b/docs/reference/api/typedoc/enums/rpcserverstate.html
@@ -90,7 +90,7 @@
 					<div class="tsd-signature tsd-kind-icon">Init<wbr>Header<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L27">rpc_server.ts:27</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L27">rpc_server.ts:27</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -100,7 +100,7 @@
 					<div class="tsd-signature tsd-kind-icon">Init<wbr>Header<wbr>Key<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L28">rpc_server.ts:28</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L28">rpc_server.ts:28</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -110,7 +110,7 @@
 					<div class="tsd-signature tsd-kind-icon">Init<wbr>Server<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L29">rpc_server.ts:29</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L29">rpc_server.ts:29</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -120,7 +120,7 @@
 					<div class="tsd-signature tsd-kind-icon">Receive<wbr>Packet<wbr>Body<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L32">rpc_server.ts:32</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L32">rpc_server.ts:32</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -130,7 +130,7 @@
 					<div class="tsd-signature tsd-kind-icon">Receive<wbr>Packet<wbr>Header<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L31">rpc_server.ts:31</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L31">rpc_server.ts:31</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -140,7 +140,7 @@
 					<div class="tsd-signature tsd-kind-icon">Wait<wbr>For<wbr>Callback<span class="tsd-signature-symbol">:</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/rpc_server.ts#L30">rpc_server.ts:30</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/rpc_server.ts#L30">rpc_server.ts:30</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/enums/sizeof.html b/docs/reference/api/typedoc/enums/sizeof.html
index be560e1a9..09f6d4d40 100644
--- a/docs/reference/api/typedoc/enums/sizeof.html
+++ b/docs/reference/api/typedoc/enums/sizeof.html
@@ -100,7 +100,7 @@
 					<div class="tsd-signature tsd-kind-icon">DLData<wbr>Type<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = I32</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L206">ctypes.ts:206</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L206">ctypes.ts:206</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -110,7 +110,7 @@
 					<div class="tsd-signature tsd-kind-icon">DLDevice<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = I32 + I32</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L207">ctypes.ts:207</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L207">ctypes.ts:207</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -120,7 +120,7 @@
 					<div class="tsd-signature tsd-kind-icon">F32<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol"> = 4</span></div>
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/351f31b51/web/src/ctypes.ts#L203">ctypes.ts:203</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/fafabc96c/web/src/ctypes.ts#L203">ctypes.ts:203</a></li>
 						</ul>
... 1199 lines suppressed ...