[tvm-site] branch asf-site updated: deploying docs (apache/tvm@271ad4302917011a54b257ca2a78c563a7ba652c)

[tq...@apache.org]

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 bd640f2c11 deploying docs (apache/tvm@271ad4302917011a54b257ca2a78c563a7ba652c)
bd640f2c11 is described below

commit bd640f2c11aa3bbdc89f620b21146f6d91979768
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Wed Nov 16 19:23:46 2022 +0000

    deploying docs (apache/tvm@271ad4302917011a54b257ca2a78c563a7ba652c)
---
 docs/_images/sphx_glr_micro_train_001.png          |  Bin 337505 -> 331892 bytes
 docs/_images/sphx_glr_micro_train_thumb.png        |  Bin 23675 -> 23825 bytes
 .../how_to/compile_models/from_darknet.rst.txt     |    2 +-
 .../how_to/compile_models/from_keras.rst.txt       |    2 +-
 .../how_to/compile_models/from_mxnet.rst.txt       |    2 +-
 .../how_to/compile_models/from_oneflow.rst.txt     |    2 +-
 .../how_to/compile_models/from_pytorch.rst.txt     |    2 +-
 .../how_to/compile_models/from_tensorflow.rst.txt  |    2 +-
 .../compile_models/sg_execution_times.rst.txt      |   22 +-
 .../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   |    8 +-
 .../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     |   14 +-
 .../optimize_operators/sg_execution_times.rst.txt  |    8 +-
 .../sg_execution_times.rst.txt                     |   14 +-
 .../tune_conv2d_layer_cuda.rst.txt                 | 2033 ++------------------
 .../tune_network_arm.rst.txt                       |    6 +-
 .../tune_network_cuda.rst.txt                      |   16 +-
 .../tune_network_mali.rst.txt                      |    6 +-
 .../tune_network_x86.rst.txt                       |   16 +-
 .../tune_sparse_x86.rst.txt                        |  134 +-
 .../tune_with_autotvm/sg_execution_times.rst.txt   |    6 +-
 .../tune_with_autotvm/tune_conv2d_cuda.rst.txt     |  226 +--
 .../work_with_microtvm/micro_autotune.rst.txt      |   16 +-
 .../work_with_microtvm/micro_pytorch.rst.txt       |    4 +-
 .../how_to/work_with_microtvm/micro_train.rst.txt  |   18 +-
 .../work_with_microtvm/sg_execution_times.rst.txt  |   14 +-
 .../work_with_relay/sg_execution_times.rst.txt     |    8 +-
 .../how_to/work_with_schedules/intrin_math.rst.txt |    2 +-
 .../work_with_schedules/sg_execution_times.rst.txt |   12 +-
 .../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     |   11 +-
 docs/_sources/tutorial/autotvm_matmul_x86.rst.txt  |   20 +-
 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       |   46 +-
 docs/commit_hash                                   |    2 +-
 docs/how_to/compile_models/from_darknet.html       |    2 +-
 docs/how_to/compile_models/from_keras.html         |    2 +-
 docs/how_to/compile_models/from_mxnet.html         |    2 +-
 docs/how_to/compile_models/from_oneflow.html       |   18 +-
 docs/how_to/compile_models/from_pytorch.html       |   10 +-
 docs/how_to/compile_models/from_tensorflow.html    |    2 +-
 docs/how_to/compile_models/sg_execution_times.html |   22 +-
 .../deploy_models/deploy_model_on_android.html     |    2 +-
 .../deploy_object_detection_pytorch.html           |   44 +-
 docs/how_to/deploy_models/deploy_prequantized.html |    7 +-
 .../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  |   36 +-
 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     |    8 +-
 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       |   14 +-
 .../optimize_operators/sg_execution_times.html     |    8 +-
 .../sg_execution_times.html                        |   14 +-
 .../tune_conv2d_layer_cuda.html                    | 2029 ++-----------------
 .../tune_with_autoscheduler/tune_network_arm.html  |    6 +-
 .../tune_with_autoscheduler/tune_network_cuda.html |   16 +-
 .../tune_with_autoscheduler/tune_network_mali.html |    6 +-
 .../tune_with_autoscheduler/tune_network_x86.html  |   16 +-
 .../tune_with_autoscheduler/tune_sparse_x86.html   |  134 +-
 .../tune_with_autotvm/sg_execution_times.html      |    6 +-
 .../how_to/tune_with_autotvm/tune_conv2d_cuda.html |  226 +--
 docs/how_to/work_with_microtvm/micro_autotune.html |   16 +-
 docs/how_to/work_with_microtvm/micro_pytorch.html  |    5 +-
 docs/how_to/work_with_microtvm/micro_train.html    |   16 +-
 .../work_with_microtvm/sg_execution_times.html     |   14 +-
 .../how_to/work_with_relay/sg_execution_times.html |    8 +-
 docs/how_to/work_with_schedules/intrin_math.html   |    2 +-
 .../work_with_schedules/sg_execution_times.html    |   12 +-
 docs/how_to/work_with_schedules/tensorize.html     |    2 +-
 .../api/doxygen/nn_2pooling_8h_source.html         |   21 +-
 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       |    7 +-
 docs/tutorial/autotvm_matmul_x86.html              |   20 +-
 docs/tutorial/autotvm_relay_x86.html               |  265 ++-
 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         |   46 +-
 132 files changed, 1537 insertions(+), 4945 deletions(-)

diff --git a/docs/_images/sphx_glr_micro_train_001.png b/docs/_images/sphx_glr_micro_train_001.png
index cdece017f1..9ba43750df 100644
Binary files a/docs/_images/sphx_glr_micro_train_001.png and b/docs/_images/sphx_glr_micro_train_001.png differ
diff --git a/docs/_images/sphx_glr_micro_train_thumb.png b/docs/_images/sphx_glr_micro_train_thumb.png
index 3ea3b2a601..d025dae545 100644
Binary files a/docs/_images/sphx_glr_micro_train_thumb.png and b/docs/_images/sphx_glr_micro_train_thumb.png differ
diff --git a/docs/_sources/how_to/compile_models/from_darknet.rst.txt b/docs/_sources/how_to/compile_models/from_darknet.rst.txt
index 978279bbae..fa2f7b2eb9 100644
--- a/docs/_sources/how_to/compile_models/from_darknet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_darknet.rst.txt
@@ -315,7 +315,7 @@ The process is no different from other examples.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  16.228 seconds)
+   **Total running time of the script:** ( 1 minutes  15.285 seconds)
 
 
 .. _sphx_glr_download_how_to_compile_models_from_darknet.py:
diff --git a/docs/_sources/how_to/compile_models/from_keras.rst.txt b/docs/_sources/how_to/compile_models/from_keras.rst.txt
index 790b80f493..30629a6d70 100644
--- a/docs/_sources/how_to/compile_models/from_keras.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_keras.rst.txt
@@ -228,7 +228,7 @@ Look up prediction top 1 index in 1000 class synset.
  .. code-block:: none
 
     Relay top-1 id: 285, class name: Egyptian cat
-
    1/1 [==============================] - ETA: 0s
    1/1 [==============================] - 1s 955ms/step
+
    1/1 [==============================] - ETA: 0s
    1/1 [==============================] - 1s 1s/step
     Keras top-1 id: 285, class name: Egyptian cat
 
 
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 5eb0d30fd7..98d327b017 100644
--- a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
@@ -115,7 +115,7 @@ In this section, we download a pretrained imagenet model and classify an image.
 
  .. code-block:: none
 
-    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip2b781813-d81c-436a-a775-b37a76c5a5e2 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip9b8baf24-50e2-4883-a416-0d0ac295aea7 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_oneflow.rst.txt b/docs/_sources/how_to/compile_models/from_oneflow.rst.txt
index 43b90986df..7ec005ebb8 100644
--- a/docs/_sources/how_to/compile_models/from_oneflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_oneflow.rst.txt
@@ -116,7 +116,7 @@ Load a pretrained OneFlow model and save model
  .. code-block:: none
 
     Downloading: "https://oneflow-public.oss-cn-beijing.aliyuncs.com/model_zoo/flowvision/classification/ResNet/resnet18.zip" to /workspace/.oneflow/flowvision_cache/resnet18.zip
-
      0%|          | 0.00/41.5M [00:00<?, ?B/s]
     15%|#5        | 6.33M/41.5M [00:00<00:01, 31.4MB/s]
     27%|##6       | 11.0M/41.5M [00:00<00:00, 38.6MB/s]
     39%|###8      | 16.0M/41.5M [00:00<00:00, 33.8MB/s]
     54%|#####3    | 22.3M/41.5M [00:00<00:00, 33.7MB/s]
     62%|######1   | 25.7M/41.5M [00:00<00:00, 31.0MB/s]
     77%|#######7  | 32.0M/41.5M [00:00<00:00, 35.2MB/s]
     85%|########5 | 35.4M/41.5M [00:01<00:00, 33.5MB/s]
     93%|#########3| 38.6M/41.5M [00:01<00:00, 32.8MB/s]
    100%|##########| 41.5M/41.5M [00:01<00:00, 34.3MB/s]
+
      0%|          | 0.00/41.5M [00:00<?, ?B/s]
     15%|#5        | 6.33M/41.5M [00:00<00:00, 47.1MB/s]
     27%|##7       | 11.2M/41.5M [00:00<00:00, 49.2MB/s]
     39%|###8      | 16.0M/41.5M [00:00<00:00, 37.0MB/s]
     48%|####7     | 19.8M/41.5M [00:00<00:00, 33.6MB/s]
     57%|#####7    | 23.7M/41.5M [00:00<00:00, 32.9MB/s]
     65%|######4   | 26.9M/41.5M [00:00<00:00, 30.2MB/s]
     82%|########2 | 34.1M/41.5M [00:00<00:00, 39.4MB/s]
     96%|#########6| 40.0M/41.5M [00:01<00:00, 38.6MB/s]
    100%|##########| 41.5M/41.5M [00:01<00:00, 38.5MB/s]
 
 
 
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 d918e0691f..ad37f69c59 100644
--- a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
@@ -98,7 +98,7 @@ Load a pretrained PyTorch model
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing `weights=ResNet18_Weights.IMAGENET1K_V1`. You can also use `weights=ResNet18_Weights.DEFAULT` to get the most up-to-date weights.
       warnings.warn(msg)
     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]
     18%|#7        | 7.99M/44.7M [00:00<00:00, 68.8MB/s]
     36%|###5      | 16.0M/44.7M [00:00<00:00, 55.9MB/s]
     54%|#####3    | 24.0M/44.7M [00:00<00:00, 62.2MB/s]
     72%|#######1  | 32.0M/44.7M [00:00<00:00, 66.3MB/s]
     90%|########9 | 40.1M/44.7M [00:00<00:00, 70.1MB/s]
    100%|##########| 44.7M/44.7M [00:00<00:00, 71.9MB/s]
+
      0%|          | 0.00/44.7M [00:00<?, ?B/s]
     32%|###1      | 14.1M/44.7M [00:00<00:00, 148MB/s]
     63%|######3   | 28.2M/44.7M [00:00<00:00, 89.5MB/s]
     98%|#########8| 43.9M/44.7M [00:00<00:00, 114MB/s] 
    100%|##########| 44.7M/44.7M [00:00<00:00, 111MB/s]
 
 
 
diff --git a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
index b608418df7..9789fab75a 100644
--- a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
@@ -416,7 +416,7 @@ Run the corresponding model on tensorflow
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  13.077 seconds)
+   **Total running time of the script:** ( 1 minutes  13.573 seconds)
 
 
 .. _sphx_glr_download_how_to_compile_models_from_tensorflow.py:
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 de4e8e5f7a..bb04074062 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
@@ -5,26 +5,26 @@
 
 Computation times
 =================
-**05:58.074** total execution time for **how_to_compile_models** files:
+**06:00.467** total execution time for **how_to_compile_models** files:
 
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)       | 01:16.228 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)       | 01:15.285 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:13.077 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:13.573 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)         | 00:47.284 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)         | 00:48.006 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``)       | 00:33.815 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``)       | 00:34.751 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)           | 00:29.966 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)           | 00:30.889 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)         | 00:27.561 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)         | 00:27.578 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)         | 00:26.150 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)         | 00:25.668 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)       | 00:23.095 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)       | 00:23.593 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)           | 00:18.467 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)           | 00:18.636 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)             | 00:02.433 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)             | 00:02.487 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
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 d0ab064d23..a90e47b39d 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
@@ -433,7 +433,7 @@ Execute on TVM
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      16.1529      16.1286      16.3120      16.0284       0.0906   
+      16.9160      16.8723      17.7639      16.3612       0.4699   
                
 
 
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 c93c63755d..b3aa5f07a2 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
@@ -127,7 +127,7 @@ Load pre-trained maskrcnn from torchvision and do tracing
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing `weights=MaskRCNN_ResNet50_FPN_Weights.COCO_V1`. You can also use `weights=MaskRCNN_ResNet50_FPN_Weights.DEFAULT` to get the most up-to-date weights.
       warnings.warn(msg)
     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]
      5%|4         | 7.99M/170M [00:00<00:04, 39.9MB/s]
      8%|8         | 14.3M/170M [00:00<00:03, 40.8MB/s]
     11%|#         | 18.2M/170M [00:00<00:04, 36.0MB/s]
     14%|#4        | 24.0M/170M [00:00<00:04, 36.7MB/s]
     19%|#8        | 32.0M/170M [00:00<00:03, 38.2MB/s]
     24%|##3       | 40.0M/170M [00:01<00:03, 43.5MB/s]
     28%|##8       | 48.0M/170M [00:01<00:02, 43.9MB/s]
     33%|###2      | 56.0M/170M [00:01<00:02, 48.5MB/s]
     37%|###6      | 62.3M/170M [00:01<00:02, 52.2MB/s]
     40%|###9      | 67.5M/170M [00:01<00:02, 47.1MB/s]
     42%|####2     | 72.2M/170M [00:01<00:02, 41.2MB/s]
     47%|####6     | 79.7M/170M [00:01<00:01, 49.6MB/s]
     50%|####9     | 84.8M/170M [00:02<00:01, 45.0MB/s]
     53%|#####2    | 89.5M/170M [00:02<00:02, 33.8MB/s]
     56%|#####6    | 95.5M/170M [00:02<00:01, 39.6MB/s]
     59%|#####8    | 99.9M/170M [00:02<00:02, 35.7MB/s]
     61%|######1   | 104M/170M [00:02<00:02, 31.7MB/s
 ] 
     66%|######5   | 112M/170M [00:02<00:01, 35.3MB/s]
     71%|#######   | 120M/170M [00:03<00:01, 43.0MB/s]
     75%|#######5  | 128M/170M [00:03<00:00, 44.1MB/s]
     80%|########  | 136M/170M [00:03<00:00, 46.6MB/s]
     85%|########4 | 144M/170M [00:03<00:00, 45.1MB/s]
     89%|########9 | 152M/170M [00:03<00:00, 45.9MB/s]
     93%|#########3| 158M/170M [00:03<00:00, 45.5MB/s]
     96%|#########5| 163M/170M [00:04<00:00, 41.6MB/s]
    100%|##########| 170M/170M [00:04<00:00, 42.9MB/s]
+
      0%|          | 0.00/170M [00:00<?, ?B/s]
      5%|4         | 7.99M/170M [00:00<00:02, 74.6MB/s]
     14%|#4        | 24.0M/170M [00:00<00:01, 124MB/s] 
     21%|##1       | 36.0M/170M [00:00<00:01, 120MB/s]
     28%|##7       | 47.5M/170M [00:00<00:01, 111MB/s]
     34%|###4      | 58.2M/170M [00:00<00:01, 102MB/s]
     41%|####      | 69.4M/170M [00:00<00:00, 106MB/s]
     47%|####6     | 79.6M/170M [00:00<00:00, 104MB/s]
     53%|#####2    | 89.7M/170M [00:00<00:00, 103MB/s]
     59%|#####8    | 99.6M/170M [00:01<00:00, 92.2MB/s]
     66%|######6   | 112M/170M [00:01<00:00, 104MB/s]  
     72%|#######2  | 123M/170M [00:01<00:00, 103MB/s]
     78%|#######8  | 133M/170M [00:01<00:00, 102MB/s]
     84%|########3 | 143M/170M [00:01<00:00, 102MB/s]
     90%|########9 | 152M/170M [00:01<00:00, 94.7MB/s]
     97%|#########6| 164M/170M [00:01<00:00, 103MB/s] 
    100%|##########| 170M/170M [00:01<00:00, 102MB/s]
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torch/nn/functional.py:3897: 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)
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/detection/anchor_utils.py:124: 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').
@@ -296,7 +296,7 @@ Get boxes with score larger than 0.9
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 3 minutes  24.737 seconds)
+   **Total running time of the script:** ( 3 minutes  24.066 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 f4e18c1526..e4d89b2fac 100644
--- a/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
@@ -236,7 +236,7 @@ training. Other models require a full post training calibration.
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing `weights=MobileNet_V2_Weights.IMAGENET1K_V1`. You can also use `weights=MobileNet_V2_Weights.DEFAULT` to get the most up-to-date weights.
       warnings.warn(msg)
     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, 159MB/s]
+
      0%|          | 0.00/13.6M [00:00<?, ?B/s]
     61%|######    | 8.21M/13.6M [00:00<00:00, 86.1MB/s]
    100%|##########| 13.6M/13.6M [00:00<00:00, 115MB/s] 
 
 
 
@@ -418,7 +418,7 @@ 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)  
-      90.6324      90.5561      93.1118      90.2627       0.4082   
+      90.4818      90.3775      93.0215      90.2388       0.3177   
                
 
 
@@ -467,7 +467,7 @@ TODO
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  8.512 seconds)
+   **Total running time of the script:** ( 1 minutes  8.426 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 0ff817b867..f9eee034ab 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
@@ -432,7 +432,7 @@ 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)  
-      121.4938     121.4546     125.6496     120.6399      0.5865   
+      121.9242     121.9387     124.2872     120.9418      0.4936   
                
 
 
@@ -469,7 +469,7 @@ 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:** ( 2 minutes  28.305 seconds)
+   **Total running time of the script:** ( 2 minutes  27.895 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 ce30a8b703..92638aad07 100644
--- a/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
@@ -253,7 +253,7 @@ We create a Relay VM to build and execute the model.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  42.212 seconds)
+   **Total running time of the script:** ( 1 minutes  43.484 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 e87f384719..4614defd58 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
@@ -166,7 +166,7 @@ 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]
      5%|4         | 6163/132723 [00:00<00:02, 61623.30KB/s]
     11%|#1        | 14890/132723 [00:00<00:01, 76706.81KB/s]
     17%|#6        | 22561/132723 [00:00<00:01, 61488.26KB/s]
     24%|##3       | 31336/132723 [00:00<00:01, 70502.66KB/s]
     29%|##9       | 39069/132723 [00:00<00:01, 72748.29KB/s]
     36%|###6      | 47847/132723 [00:00<00:01, 77552.97KB/s]
     42%|####2     | 55779/132723 [00:00<00:01, 60308.08KB/s]
     49%|####8     | 64533/132723 [00:00<00:01, 67303.49KB/s]
     54%|#####4    | 71883/132723 [00:01<00:00, 68966.09KB/s]
     61%|######    | 80613/132723 [00:01<00:00, 74045.10KB/s]
     67%|######7   | 89317/132723 [00:01<00:00, 77730.07KB/s]
     74%|#######3  | 97956/132723 [00:01<00:00, 80224.79KB/s]
     80%|########  | 106638/132723 [00:01<00:00, 82146.77KB/s]
     87%|########6 | 115007/132723 [00:01<00:00, 66534.79KB/s]
     93%|#########3| 123756/132723 [00:01<00:00, 71833.44KB/s]
     99%|########
 #9| 131450/132723 [00:01<00:00, 73184.77KB/s]
    100%|##########| 132723/132723 [00:01<00:00, 71855.54KB/s]
+
      0%|          | 0/132723 [00:00<?, ?KB/s]
      4%|4         | 5804/132723 [00:00<00:02, 58030.23KB/s]
     10%|9         | 13109/132723 [00:00<00:01, 66862.10KB/s]
     15%|#4        | 19796/132723 [00:00<00:01, 64361.03KB/s]
     21%|##1       | 28369/132723 [00:00<00:01, 72601.06KB/s]
     28%|##7       | 36965/132723 [00:00<00:01, 77342.24KB/s]
     34%|###4      | 45592/132723 [00:00<00:01, 80346.29KB/s]
     41%|####      | 54218/132723 [00:00<00:00, 82264.53KB/s]
     47%|####7     | 62867/132723 [00:00<00:00, 83602.05KB/s]
     54%|#####3    | 71483/132723 [00:00<00:00, 84388.91KB/s]
     60%|######    | 80103/132723 [00:01<00:00, 84945.16KB/s]
     67%|######6   | 88603/132723 [00:01<00:00, 75283.38KB/s]
     73%|#######3  | 97213/132723 [00:01<00:00, 78298.82KB/s]
     79%|#######9  | 105211/132723 [00:01<00:00, 73994.42KB/s]
     86%|########5 | 113804/132723 [00:01<00:00, 77293.46KB/s]
     92%|#########1| 121669/132723 [00:01<00:00, 69239.69KB/s]
     98%|########
 #8| 130213/132723 [00:01<00:00, 73531.33KB/s]
    100%|##########| 132723/132723 [00:01<00:00, 75954.56KB/s]
 
 
 
@@ -242,7 +242,7 @@ Display result
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 3 minutes  7.321 seconds)
+   **Total running time of the script:** ( 3 minutes  8.938 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 be5d619c0c..951c4e54e2 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
@@ -5,24 +5,24 @@
 
 Computation times
 =================
-**13:21.255** total execution time for **how_to_deploy_models** files:
+**13:23.321** total execution time for **how_to_deploy_models** files:
 
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:24.737 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:24.066 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)                           | 03:07.321 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)                           | 03:08.938 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)           | 02:28.305 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)           | 02:27.895 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)                               | 01:42.212 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)                               | 01:43.484 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)                         | 01:08.512 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)                         | 01:08.426 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)                 | 00:37.861 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)                 | 00:37.846 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``)                       | 00:26.511 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``)                       | 00:26.691 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)                       | 00:25.790 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)                       | 00:25.969 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)                                     | 00:00.007 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
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 e3c6121d81..c8853ccc67 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
@@ -472,7 +472,7 @@ 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.zip85164043-6ab1-4ce7-9f25-a44176bc77c5 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+    Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip23873bed-17bd-4043-842f-0f79495e7b5f 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 2cfb8109ed..c0f2b2554e 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
@@ -5,14 +5,14 @@
 
 Computation times
 =================
-**00:49.063** total execution time for **how_to_extend_tvm** files:
+**00:49.942** total execution time for **how_to_extend_tvm** files:
 
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:45.485 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:46.316 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)           | 00:02.505 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)           | 00:02.538 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)                     | 00:01.065 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)                     | 00:01.079 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)       | 00:00.008 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
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 64f94f31d2..30e12fd13d 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
@@ -216,10 +216,10 @@ profile the execution time of each passes.
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 7311us [7311us] (46.27%; 46.27%)
-    FoldScaleAxis: 8490us [8us] (53.73%; 53.73%)
-            FoldConstant: 8482us [1698us] (53.68%; 99.90%)
-                    InferType: 6783us [6783us] (42.93%; 79.98%)
+    InferType: 7596us [7596us] (46.51%; 46.51%)
+    FoldScaleAxis: 8737us [8us] (53.49%; 53.49%)
+            FoldConstant: 8729us [1788us] (53.44%; 99.91%)
+                    InferType: 6941us [6941us] (42.50%; 79.52%)
 
 
 
@@ -258,10 +258,10 @@ Refer to following sections and :py:func:`tvm.instrument.pass_instrument` for th
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 6902us [6902us] (44.90%; 44.90%)
-    FoldScaleAxis: 8470us [6us] (55.10%; 55.10%)
-            FoldConstant: 8464us [1732us] (55.06%; 99.93%)
-                    InferType: 6732us [6732us] (43.79%; 79.53%)
+    InferType: 7105us [7105us] (45.22%; 45.22%)
+    FoldScaleAxis: 8607us [6us] (54.78%; 54.78%)
+            FoldConstant: 8600us [1749us] (54.74%; 99.92%)
+                    InferType: 6851us [6851us] (43.61%; 79.67%)
 
 
 
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 4923070246..e71a37c9bd 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
@@ -340,7 +340,7 @@ latency of convolution.
 
  .. code-block:: none
 
-    Convolution: 54.245216 ms
+    Convolution: 54.114494 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 8e9189bc7d..a018306eee 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
@@ -657,7 +657,7 @@ be able to run on our build server
 
  .. code-block:: none
 
-    conv2d with tensor core: 13.002118 ms
+    conv2d with tensor core: 13.367216 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 e0f9d1693a..4d628628ed 100644
--- a/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
@@ -143,8 +143,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
 
  .. code-block:: none
 
-    Numpy running time: 0.019491
-    Baseline: 3.290225
+    Numpy running time: 0.019649
+    Baseline: 3.486395
 
 
 
@@ -238,7 +238,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
 
  .. code-block:: none
 
-    Opt1: 0.328238
+    Opt1: 0.329770
 
 
 
@@ -340,7 +340,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
 
  .. code-block:: none
 
-    Opt2: 0.353639
+    Opt2: 0.349371
 
 
 
@@ -435,7 +435,7 @@ the access pattern for A matrix is more cache friendly.
 
  .. code-block:: none
 
-    Opt3: 0.122369
+    Opt3: 0.120619
 
 
 
@@ -680,7 +680,7 @@ write to C when all the block results are ready.
 
  .. code-block:: none
 
-    Opt5: 0.111313
+    Opt5: 0.112183
 
 
 
@@ -804,7 +804,7 @@ Furthermore, we can also utilize multi-core processors to do the thread-level pa
 
  .. code-block:: none
 
-    Opt6: 0.147555
+    Opt6: 0.147467
 
 
 
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 532377169b..ecb3bfac71 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
@@ -5,12 +5,12 @@
 
 Computation times
 =================
-**00:35.420** total execution time for **how_to_optimize_operators** files:
+**00:36.049** total execution time for **how_to_optimize_operators** files:
 
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)                       | 00:32.731 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)                       | 00:33.260 | 0.0 MB |
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.512 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.554 | 0.0 MB |
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)             | 00:01.176 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)             | 00:01.235 | 0.0 MB |
 +-----------------------------------------------------------------------------------------------+-----------+--------+
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 030d97cdb2..fec2a9e08a 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
@@ -5,18 +5,18 @@
 
 Computation times
 =================
-**09:16.506** total execution time for **how_to_tune_with_autoscheduler** files:
+**09:21.677** total execution time for **how_to_tune_with_autoscheduler** files:
 
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 05:43.754 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 05:48.786 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)             | 01:34.813 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)             | 01:35.031 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)           | 01:04.752 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)           | 01:04.757 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)               | 00:28.883 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)               | 00:29.200 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)             | 00:12.495 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)             | 00:12.289 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)           | 00:11.809 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)           | 00:11.614 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
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 99318a7090..37001e0e9d 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
@@ -206,6 +206,13 @@ file and apply it.
 
 
 
+.. rst-class:: sphx-glr-script-out
+
+ .. code-block:: none
+
+    .T
+
+
 
 
 
@@ -239,952 +246,120 @@ cooperative fetching, unrolling and operator fusion.
                  bias: Buffer(bias_2: Pointer(float32), float32, [1, 512, 1, 1], []),
                  compute: Buffer(compute_2: Pointer(float32), float32, [1, 512, 7, 7], [])}
       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, [16]), storage_scope = local;
-      allocate(pad_temp.shared: Pointer(shared float32), float32, [784]), storage_scope = shared;
-      allocate(kernel.shared: Pointer(shared float32), float32, [256]), storage_scope = shared;
-      attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49 {
-        conv2d_nchw_1: Buffer(conv2d_nchw, float32, [4], [], scope="local", align=8)[0] = 0f32
-        conv2d_nchw_1[2] = 0f32
-        conv2d_nchw_1[4] = 0f32
-        conv2d_nchw_1[6] = 0f32
-        conv2d_nchw_1[8] = 0f32
-        conv2d_nchw_1[10] = 0f32
-        conv2d_nchw_1[12] = 0f32
-        conv2d_nchw_1[14] = 0f32
-        conv2d_nchw_1[1] = 0f32
-        conv2d_nchw_1[3] = 0f32
-        conv2d_nchw_1[5] = 0f32
+      attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 8;
+      allocate(conv2d_nchw: Pointer(local float32), float32, [14]), storage_scope = local;
+      allocate(pad_temp.shared: Pointer(shared float32), float32, [1008]), storage_scope = shared;
+      allocate(kernel.shared: Pointer(shared float32), float32, [3072]), storage_scope = shared;
+      attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224 {
+        conv2d_nchw_1: Buffer(conv2d_nchw, float32, [14], [], scope="local", align=32)[0] = 0f32
         conv2d_nchw_1[7] = 0f32
+        conv2d_nchw_1[1] = 0f32
+        conv2d_nchw_1[8] = 0f32
+        conv2d_nchw_1[2] = 0f32
         conv2d_nchw_1[9] = 0f32
+        conv2d_nchw_1[3] = 0f32
+        conv2d_nchw_1[10] = 0f32
+        conv2d_nchw_1[4] = 0f32
         conv2d_nchw_1[11] = 0f32
+        conv2d_nchw_1[5] = 0f32
+        conv2d_nchw_1[12] = 0f32
+        conv2d_nchw_1[6] = 0f32
         conv2d_nchw_1[13] = 0f32
-        conv2d_nchw_1[15] = 0f32
         for (rc.outer.outer: int32, 0, 32) {
-          for (ry.outer.outer: int32, 0, 3) {
-            attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1: Buffer(pad_temp.shared, float32, [784], [], scope="shared")[threadIdx.x_1] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3: Buffer(data_2, float32, [25088], [])[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 8)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 41)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 90)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 139)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 188)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 237)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 286)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 335)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 384)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 433)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 482)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 531)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 580)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 629)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 678)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (1 <= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 727)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49 {
-              kernel.shared_1: Buffer(kernel.shared, float32, [256], [], scope="shared")[(threadIdx.x_2*4)] = kernel_3: Buffer(kernel_2, float32, [2359296], [])[(((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3))]
-              kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 9)]
-              kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 18)]
-              kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 27)]
-            }
-            attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            if @tir.likely((threadIdx.x_2 < 15), dtype=bool) {
-              let cse_var_1: int32 = (ry.outer.outer*3)
-               {
-                kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_1)]
+          for (rx.outer.outer: int32, 0, 3) {
+            let cse_var_1: int32 = (rc.outer.outer*144)
+             {
+              attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              if @tir.likely((threadIdx.x_1 < 144), dtype=bool) {
+                pad_temp.shared_1: Buffer(pad_temp.shared, float32, [1008], [], scope="shared")[(threadIdx.x_1*7)] = @tir.if_then_else((((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)) && (1 <= rx.outer.outer)), data_3: Buffer(data_2, float32, [25088], [])[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 8)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 1)] = @tir.if_then_else(((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 7)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 2)] = @tir.if_then_else(((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 6)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 3)] = @tir.if_then_else(((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 5)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 4)] = @tir.if_then_else(((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 4)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 5)] = @tir.if_then_else(((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 3)], 0f32, dtype=float32)
+                pad_temp.shared_1[((threadIdx.x_1*7) + 6)] = @tir.if_then_else((((1 <= floormod(threadIdx.x_1, 9)) && (floormod(threadIdx.x_1, 9) < 8)) && (rx.outer.outer < 2)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 2)], 0f32, dtype=float32)
               }
-            }
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[threadIdx.x_1] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 7)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 42)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 91)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 140)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 189)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 238)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 287)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 336)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 385)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 434)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 483)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 532)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 581)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 630)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 679)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else(((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 728)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49 {
-              kernel.shared_1[(threadIdx.x_2*4)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 1)]
-              kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 10)]
-              kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 19)]
-              kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 28)]
-            }
-            attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            if @tir.likely((threadIdx.x_2 < 15), dtype=bool) {
-              let cse_var_2: int32 = (ry.outer.outer*3)
-               {
-                kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_2) + 1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_2) + 1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_2) + 1)]
-                kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_2) + 1)]
+              attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1: Buffer(kernel.shared, float32, [3072], [], scope="shared")[threadIdx.x_2] = kernel_3: Buffer(kernel_2, float32, [2359296], [])[(((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 224)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 224), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 448)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 448), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 672)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 64512)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 896)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 896), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 1120)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1120), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 1344)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 129024)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 1568)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1568), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 1792)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1792), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 2016)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 193536)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 2240)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2240), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 2464)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2464), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              kernel.shared_1[(threadIdx.x_2 + 2688)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 258048)]
+              attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 224;
+              if @tir.likely((threadIdx.x_2 < 160), dtype=bool) {
+                kernel.shared_1[(threadIdx.x_2 + 2912)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2912), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
               }
-            }
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[threadIdx.x_1] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 6)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 43)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 92)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 141)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 190)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 239)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 288)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 337)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 386)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 435)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 484)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 533)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 582)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 631)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 680)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else((((1 <= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) && ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) < 8)) && (floormod(threadIdx.x_1, 7) < 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 729)], 0f32, dtype=float32)
-            attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49 {
-              kernel.shared_1[(threadIdx.x_2*4)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 2)]
-              kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 11)]
-              kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 20)]
-              kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 29)]
-            }
-            attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 49;
-            if @tir.likely((threadIdx.x_2 < 15), dtype=bool) {
-              let cse_var_3: int32 = (ry.outer.outer*3)
-               {
-                kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_3) + 2)]
-                kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_3) + 2)]
-                kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_3) + 2)]
-                kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_3) + 2)]
+              for (rc.outer.inner: int32, 0, 16) {
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rc.outer.inner*63) + floormod(threadIdx.x, 7))]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rc.outer.inner*63) + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+                conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+                conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 56)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+                conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 56)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
               }
             }
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-            conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-            conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
           }
         }
         for (i1.inner: int32, 0, 2) {
-          compute_3: Buffer(compute_2, float32, [25088], [])[(((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x)] = max((conv2d_nchw_1[i1.inner] + bias_3: Buffer(bias_2, float32, [512], [])[((blockIdx.x*16) + i1.inner)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 98)] = max((conv2d_nchw_1[(i1.inner + 2)] + bias_3[(((blockIdx.x*16) + i1.inner) + 2)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 196)] = max((conv2d_nchw_1[(i1.inner + 4)] + bias_3[(((blockIdx.x*16) + i1.inner) + 4)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 294)] = max((conv2d_nchw_1[(i1.inner + 6)] + bias_3[(((blockIdx.x*16) + i1.inner) + 6)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 392)] = max((conv2d_nchw_1[(i1.inner + 8)] + bias_3[(((blockIdx.x*16) + i1.inner) + 8)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 490)] = max((conv2d_nchw_1[(i1.inner + 10)] + bias_3[(((blockIdx.x*16) + i1.inner) + 10)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 588)] = max((conv2d_nchw_1[(i1.inner + 12)] + bias_3[(((blockIdx.x*16) + i1.inner) + 12)]), 0f32)
-          compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 686)] = max((conv2d_nchw_1[(i1.inner + 14)] + bias_3[(((blockIdx.x*16) + i1.inner) + 14)]), 0f32)
+          for (i2.inner: int32, 0, 7) {
+            compute_3: Buffer(compute_2, float32, [25088], [])[(((((blockIdx.x*3136) + (floordiv(threadIdx.x, 7)*98)) + (i1.inner*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[((i1.inner*7) + i2.inner)] + bias_3: Buffer(bias_2, float32, [512], [])[(((blockIdx.x*64) + (floordiv(threadIdx.x, 7)*2)) + i1.inner)]), 0f32)
+          }
         }
       }
     }
@@ -1239,7 +414,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 0.355 ms
+    Execution time of this operator: 0.336 ms
 
 
 
@@ -1289,20 +464,20 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     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=2)
     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_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=1)
-    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=8)
+    conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=32)
+    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=7)
+    conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=7)
+    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_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=8)
-    conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=2)
+    conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=1)
+    conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=16)
     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=1)
+    conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
     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)
     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 [...]
@@ -1310,10 +485,10 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     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=2)
-    compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=1)
-    compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=8)
-    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=7)
+    compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=32)
+    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=7)
+    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_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)
@@ -1334,16 +509,16 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     compute_i0_o_i_i1_o_i_fused_i2_o_i_fused_i3_o_i_fused = s[compute].fuse(compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i)
     s[compute].bind(compute_i0_o_i_i1_o_i_fused_i2_o_i_fused_i3_o_i_fused, te.thread_axis("threadIdx.x"))
     kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused = s[kernel_shared].fuse(kernel_shared_ax0, kernel_shared_ax1, kernel_shared_ax2, kernel_shared_ax3)
-    kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=4)
+    kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=1)
     s[kernel_shared].vectorize(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i)
-    kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=49)
+    kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=224)
     s[kernel_shared].bind(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i, te.thread_axis("threadIdx.x"))
     pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused = s[pad_temp_shared].fuse(pad_temp_shared_ax0, pad_temp_shared_ax1, pad_temp_shared_ax2, pad_temp_shared_ax3)
-    pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=1)
+    pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=7)
     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=49)
+    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=224)
     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", 1024)
+    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:
@@ -1361,891 +536,103 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
       #define int64_t long long
       #define uint64_t unsigned long long
     #endif
-    extern "C" __global__ void __launch_bounds__(49) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
-      float conv2d_nchw[16];
-      __shared__ float pad_temp_shared[784];
-      __shared__ float kernel_shared[256];
+    extern "C" __global__ void __launch_bounds__(224) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+      float conv2d_nchw[14];
+      __shared__ float pad_temp_shared[1008];
+      __shared__ float kernel_shared[3072];
       conv2d_nchw[0] = 0.000000e+00f;
-      conv2d_nchw[2] = 0.000000e+00f;
-      conv2d_nchw[4] = 0.000000e+00f;
-      conv2d_nchw[6] = 0.000000e+00f;
-      conv2d_nchw[8] = 0.000000e+00f;
-      conv2d_nchw[10] = 0.000000e+00f;
-      conv2d_nchw[12] = 0.000000e+00f;
-      conv2d_nchw[14] = 0.000000e+00f;
-      conv2d_nchw[1] = 0.000000e+00f;
-      conv2d_nchw[3] = 0.000000e+00f;
-      conv2d_nchw[5] = 0.000000e+00f;
       conv2d_nchw[7] = 0.000000e+00f;
+      conv2d_nchw[1] = 0.000000e+00f;
+      conv2d_nchw[8] = 0.000000e+00f;
+      conv2d_nchw[2] = 0.000000e+00f;
       conv2d_nchw[9] = 0.000000e+00f;
+      conv2d_nchw[3] = 0.000000e+00f;
+      conv2d_nchw[10] = 0.000000e+00f;
+      conv2d_nchw[4] = 0.000000e+00f;
       conv2d_nchw[11] = 0.000000e+00f;
+      conv2d_nchw[5] = 0.000000e+00f;
+      conv2d_nchw[12] = 0.000000e+00f;
+      conv2d_nchw[6] = 0.000000e+00f;
       conv2d_nchw[13] = 0.000000e+00f;
-      conv2d_nchw[15] = 0.000000e+00f;
       for (int rc_outer_outer = 0; rc_outer_outer < 32; ++rc_outer_outer) {
-        for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
+        for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
           __syncthreads();
-          pad_temp_shared[((int)threadIdx.x)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 8)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 49)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 41)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 98)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 90)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 147)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 139)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 188)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 245)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 237)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 294)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 286)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 343)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 335)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 392)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 384)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 441)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 433)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 490)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 482)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 539)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 531)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 588)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 580)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 637)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 629)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 686)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 678)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 735)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 727)] : 0.000000e+00f);
-          kernel_shared[(((int)threadIdx.x) * 4)] = kernel[(((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3))];
-          kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 9)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 18)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 27)];
-          if (((int)threadIdx.x) < 15) {
-            kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) & 15) * 9)) + (ry_outer_outer * 3))];
-            kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) & 15) * 9)) + (ry_outer_outer * 3))];
-            kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) & 15) * 9)) + (ry_outer_outer * 3))];
-            kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) & 15) * 9)) + (ry_outer_outer * 3))];
+          if (((int)threadIdx.x) < 144) {
+            pad_temp_shared[(((int)threadIdx.x) * 7)] = ((((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) && (1 <= rx_outer_outer)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 8)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 1)] = (((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 7)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 2)] = (((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 6)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 3)] = (((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 5)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 4)] = (((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 4)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 5)] = (((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 3)] : 0.000000e+00f);
+            pad_temp_shared[((((int)threadIdx.x) * 7) + 6)] = ((((1 <= (((int)threadIdx.x) % 9)) && ((((int)threadIdx.x) % 9) < 8)) && (rx_outer_outer < 2)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 2)] : 0.000000e+00f);
           }
-          __syncthreads();
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
-          __syncthreads();
-          pad_temp_shared[((int)threadIdx.x)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 7)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 49)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 42)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 98)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 91)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 147)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 140)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 196)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 189)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 245)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 238)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 294)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 287)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 343)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 336)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 392)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 385)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 441)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 434)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 490)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 483)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 539)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 532)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 588)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 581)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 637)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 630)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 686)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 679)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 735)] = (((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 728)] : 0.000000e+00f);
-          kernel_shared[(((int)threadIdx.x) * 4)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 1)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 10)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 19)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 28)];
-          if (((int)threadIdx.x) < 15) {
-            kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) & 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) & 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) & 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) & 15) * 9)) + (ry_outer_outer * 3)) + 1)];
+          kernel_shared[((int)threadIdx.x)] = kernel[(((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 224) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 448) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 672)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 64512)];
+          kernel_shared[(((int)threadIdx.x) + 896)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 896) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1120) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 129024)];
+          kernel_shared[(((int)threadIdx.x) + 1568)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1568) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1792) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 2016)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 193536)];
+          kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2240) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 2464)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2464) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+          kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 258048)];
+          if (((int)threadIdx.x) < 160) {
+            kernel_shared[(((int)threadIdx.x) + 2912)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2912) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
           }
           __syncthreads();
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
-          __syncthreads();
-          pad_temp_shared[((int)threadIdx.x)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 6)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 49)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 43)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 98)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 92)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 147)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 141)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 190)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 245)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 239)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 294)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 288)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 343)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 337)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 392)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 386)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 441)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 435)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 490)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 484)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 539)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 533)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 588)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 582)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 637)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 631)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 686)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 680)] : 0.000000e+00f);
-          pad_temp_shared[(((int)threadIdx.x) + 735)] = ((((1 <= ((((int)threadIdx.x) / 7) + ry_outer_outer)) && (((((int)threadIdx.x) / 7) + ry_outer_outer) < 8)) && ((((int)threadIdx.x) % 7) < 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 729)] : 0.000000e+00f);
-          kernel_shared[(((int)threadIdx.x) * 4)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 2)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 11)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 20)];
-          kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) & 3) * 36)) + (ry_outer_outer * 3)) + 29)];
-          if (((int)threadIdx.x) < 15) {
-            kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) & 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) & 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) & 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-            kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) >> 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) & 15) * 9)) + (ry_outer_outer * 3)) + 2)];
+          for (int rc_outer_inner = 0; rc_outer_inner < 16; ++rc_outer_inner) {
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+            conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+            conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+            conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
           }
-          __syncthreads();
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-          conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-          conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
         }
       }
       for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
-        compute[(((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x))] = max((conv2d_nchw[i1_inner] + bias[((((int)blockIdx.x) * 16) + i1_inner)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 98)] = max((conv2d_nchw[(i1_inner + 2)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 2)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 196)] = max((conv2d_nchw[(i1_inner + 4)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 4)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 294)] = max((conv2d_nchw[(i1_inner + 6)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 6)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 392)] = max((conv2d_nchw[(i1_inner + 8)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 8)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 490)] = max((conv2d_nchw[(i1_inner + 10)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 10)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 588)] = max((conv2d_nchw[(i1_inner + 12)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 12)]), 0.000000e+00f);
-        compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 686)] = max((conv2d_nchw[(i1_inner + 14)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 14)]), 0.000000e+00f);
+        for (int i2_inner = 0; i2_inner < 7; ++i2_inner) {
+          compute[(((((((int)blockIdx.x) * 3136) + ((((int)threadIdx.x) / 7) * 98)) + (i1_inner * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[((i1_inner * 7) + i2_inner)] + bias[(((((int)blockIdx.x) * 64) + ((((int)threadIdx.x) / 7) * 2)) + i1_inner)]), 0.000000e+00f);
+        }
       }
     }
 
@@ -2307,7 +694,7 @@ In the example below we resume the status and do more 5 trials.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 5 minutes  43.754 seconds)
+   **Total running time of the script:** ( 5 minutes  48.786 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_arm.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_arm.rst.txt
index 191bbba9bf..bf3bacb2b8 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_arm.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_arm.rst.txt
@@ -383,10 +383,10 @@ The task scheduler will just optimize this objective.
     T_softmax_expsum(i0) += T_softmax_exp[i0, k]
     T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-    ========== Task 5  (workload key: ["be3babb9a46e32f66b717a3e2a2d522c", [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
+    ========== Task 5  (workload key: ["8c53ca2904398da2889aa7508082d7bb", [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 1024]
-    tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-    tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+    adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+    adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
     ========== Task 6  (workload key: ["b638843f28bb2f041953d2a091167e8b", [1, 7, 7, 512], [1, 1, 512, 1024], [1, 1, 1, 1024], [1, 7, 7, 1024]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 512]
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 86cdd9f6e4..4cccf45f2b 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
@@ -332,10 +332,10 @@ The task scheduler will just optimize this objective.
     T_softmax_expsum(i0) += T_softmax_exp[i0, k]
     T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-    ========== Task 10  (workload key: ["be3babb9a46e32f66b717a3e2a2d522c", [1, 7, 7, 512], [1, 1, 1, 512]]) ==========
+    ========== Task 10  (workload key: ["8c53ca2904398da2889aa7508082d7bb", [1, 7, 7, 512], [1, 1, 1, 512]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 512]
-    tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-    tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+    adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+    adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
     ========== Task 11  (workload key: ["40b1cf1fd37b0ef111b3cc0247302508", [1, 7, 7, 512], [4, 4, 512, 512], [1, 1, 1, 512], [1, 7, 7, 512]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 512]
@@ -444,12 +444,12 @@ The task scheduler will just optimize this objective.
     p2 = PLACEHOLDER [1, 14, 14, 256]
     T_add(ax0, ax1, ax2, ax3) = (conv2d_winograd[ax0, ax1, ax2, ax3] + p2[ax0, ax1, ax2, ax3])
 
-    ========== Task 20  (workload key: ["affd3c4a65f665e451a06d65bf32750d", [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
+    ========== Task 20  (workload key: ["6d012ba18a086c11ee2b85c7324e16f2", [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
     p0 = PLACEHOLDER [1, 112, 112, 64]
     pad_temp(ax0, ax1, ax2, ax3) = tir.if_then_else(((((ax1 >= 1) && (ax1 < 113)) && (ax2 >= 1)) && (ax2 < 113)), p0[ax0, (ax1 - 1), (ax2 - 1), ax3], -3.40282e+38f)
-    tensor(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
+    pool_max(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
     p1 = PLACEHOLDER [1, 1, 1, 64]
-    T_add(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
+    T_add(ax0, ax1, ax2, ax3) = (pool_max[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
     T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
     ========== Task 21  (workload key: ["0fad1b42d0d33418e0a8d15d3bbad3c9", [1, 14, 14, 256], [1, 1, 256, 512], [1, 7, 7, 512]]) ==========
@@ -643,7 +643,7 @@ 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)  
-       8.2044       8.2046       8.2096       8.1990       0.0043   
+       7.8675       7.8677       7.8690       7.8659       0.0013   
                
 
 
@@ -671,7 +671,7 @@ Other Tips
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  4.752 seconds)
+   **Total running time of the script:** ( 1 minutes  4.757 seconds)
 
 
 .. _sphx_glr_download_how_to_tune_with_autoscheduler_tune_network_cuda.py:
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_mali.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_mali.rst.txt
index 7adefc9772..a4d0f834be 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_mali.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_mali.rst.txt
@@ -275,10 +275,10 @@ The task scheduler will just optimize this objective.
     T_add(ax0, ax1, ax2, ax3) = (conv2d_nhwc[ax0, ax1, ax2, ax3] + p2[ax0, 0, 0, ax3])
     T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
-    ========== Task 5  (workload key: ["be3babb9a46e32f66b717a3e2a2d522c", [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
+    ========== Task 5  (workload key: ["8c53ca2904398da2889aa7508082d7bb", [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 1024]
-    tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-    tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+    adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+    adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
     ========== Task 6  (workload key: ["7d79c516e212fe1d73f5dbb90eaca2cf", [1, 1000], [1, 1000]]) ==========
     p0 = PLACEHOLDER [1, 1000]
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 631d6a07db..badf7095c3 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
@@ -326,10 +326,10 @@ The task scheduler will just optimize this objective.
     T_softmax_expsum(i0) += T_softmax_exp[i0, k]
     T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-    ========== Task 11  (workload key: ["be3babb9a46e32f66b717a3e2a2d522c", [1, 7, 7, 2048], [1, 1, 1, 2048]]) ==========
+    ========== Task 11  (workload key: ["8c53ca2904398da2889aa7508082d7bb", [1, 7, 7, 2048], [1, 1, 1, 2048]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 2048]
-    tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-    tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+    adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+    adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
     ========== Task 12  (workload key: ["3060808fc5c74e18b1276729071fbae0", [1, 7, 7, 512], [1, 1, 512, 2048], [1, 7, 7, 2048], [1, 7, 7, 2048]]) ==========
     p0 = PLACEHOLDER [1, 7, 7, 512]
@@ -465,12 +465,12 @@ The task scheduler will just optimize this objective.
     p2 = PLACEHOLDER [1, 1000]
     T_add(ax0, ax1) = (T_matmul_NT[ax0, ax1] + p2[ax0, ax1])
 
-    ========== Task 27  (workload key: ["affd3c4a65f665e451a06d65bf32750d", [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
+    ========== Task 27  (workload key: ["6d012ba18a086c11ee2b85c7324e16f2", [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
     p0 = PLACEHOLDER [1, 112, 112, 64]
     pad_temp(ax0, ax1, ax2, ax3) = tir.if_then_else(((((ax1 >= 1) && (ax1 < 113)) && (ax2 >= 1)) && (ax2 < 113)), p0[ax0, (ax1 - 1), (ax2 - 1), ax3], -3.40282e+38f)
-    tensor(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
+    pool_max(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
     p1 = PLACEHOLDER [1, 1, 1, 64]
-    T_add(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
+    T_add(ax0, ax1, ax2, ax3) = (pool_max[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
     T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
     ========== Task 28  (workload key: ["6d628209072e3e3dd8f49359935acea6", [1, 56, 56, 64], [1, 1, 64, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
@@ -662,7 +662,7 @@ 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)  
-      763.1550     763.9276     764.1947     761.3427      1.2861   
+      757.8686     757.7418     758.5651     757.2990      0.5246   
                
 
 
@@ -690,7 +690,7 @@ Other Tips
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  34.813 seconds)
+   **Total running time of the script:** ( 1 minutes  35.031 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 eae3ac0f0b..f1632b8b93 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
@@ -386,77 +386,77 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
                  placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [128, 512], []),
                  compute: Buffer(compute_2: Pointer(float32), float32, [128, 512], [])}
       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.i1.outer.fused: int32, 0, 256) "parallel" {
-        allocate(compute_3: Pointer(global float32), float32, [256]), storage_scope = global {
-          for (nb_j.inner: int32, 0, 2) {
-            for (i.inner.init: int32, 0, 8) {
-              let cse_var_1: int32 = ((i.inner.init*32) + (nb_j.inner*16))
-               {
-                compute_4: Buffer(compute_3, float32, [256], [])[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
-              }
-            }
-            for (elem_idx: int32, 0, let cse_var_2: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner) in (placeholder_15: Buffer(placeholder_13, int32, [33], [])[(cse_var_2 + 1)] - placeholder_15[cse_var_2])) {
-              for (i.inner: int32, 0, 8) {
-                let cse_var_21: int32 = (elem_idx*16)
-                let cse_var_20: int32 = ((i.inner*32) + (nb_j.inner*16))
-                let cse_var_19: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner)
-                let cse_var_18: int32 = ((floordiv(i0.outer.i1.outer.fused, 16)*2048) + (i.inner*256))
-                let cse_var_17: int32 = (cse_var_20 + 9)
-                let cse_var_16: int32 = (cse_var_20 + 8)
-                let cse_var_15: int32 = (cse_var_20 + 7)
-                let cse_var_14: int32 = (cse_var_20 + 6)
-                let cse_var_13: int32 = (cse_var_20 + 5)
-                let cse_var_12: int32 = (cse_var_20 + 4)
-                let cse_var_11: int32 = (cse_var_20 + 3)
-                let cse_var_10: int32 = (cse_var_20 + 2)
-                let cse_var_9: int32 = (cse_var_20 + 15)
-                let cse_var_8: int32 = (cse_var_20 + 14)
-                let cse_var_7: int32 = (cse_var_20 + 13)
-                let cse_var_6: int32 = (cse_var_20 + 12)
-                let cse_var_5: int32 = (cse_var_20 + 11)
-                let cse_var_4: int32 = (cse_var_20 + 10)
-                let cse_var_3: int32 = (cse_var_20 + 1)
+      for (i0.outer.i1.outer.fused: int32, 0, 32) "parallel" {
+        allocate(compute_3: Pointer(global float32), float32, [2048]), storage_scope = global {
+          for (i.outer.inner: int32, 0, 4) {
+            for (nb_j.inner: int32, 0, 2) {
+              for (i.inner.init: int32, 0, 16) {
+                let cse_var_1: int32 = (((i.outer.inner*512) + (i.inner.init*32)) + (nb_j.inner*16))
                  {
-                  compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_16: Buffer(placeholder_11, float32, [78656], [])[((placeholder_15[cse_var_19]*16) + cse_var_21)]*max(placeholder_17: Buffer(placeholder_10, float32, [32768], [])[(cse_var_18 + placeholder_18: Buffer(placeholder_12, int32, [4916], [])[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 1)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 2)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 3)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 4)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 5)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 6)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 7)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 8)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 9)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 10)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 11)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 12)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 13)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 14)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-                  compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 15)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
+                  compute_4: Buffer(compute_3, float32, [2048], [])[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
+                }
+              }
+              for (elem_idx: int32, 0, let cse_var_2: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner) in (placeholder_15: Buffer(placeholder_13, int32, [33], [])[(cse_var_2 + 1)] - placeholder_15[cse_var_2])) {
+                for (i.inner: int32, 0, 16) {
+                  let cse_var_21: int32 = (elem_idx*16)
+                  let cse_var_20: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner)
+                  let cse_var_19: int32 = (((i.outer.inner*512) + (i.inner*32)) + (nb_j.inner*16))
+                  let cse_var_18: int32 = (((floordiv(i0.outer.i1.outer.fused, 16)*16384) + (i.outer.inner*4096)) + (i.inner*256))
+                  let cse_var_17: int32 = (cse_var_19 + 9)
+                  let cse_var_16: int32 = (cse_var_19 + 8)
+                  let cse_var_15: int32 = (cse_var_19 + 7)
+                  let cse_var_14: int32 = (cse_var_19 + 6)
+                  let cse_var_13: int32 = (cse_var_19 + 5)
+                  let cse_var_12: int32 = (cse_var_19 + 4)
+                  let cse_var_11: int32 = (cse_var_19 + 3)
+                  let cse_var_10: int32 = (cse_var_19 + 2)
+                  let cse_var_9: int32 = (cse_var_19 + 15)
+                  let cse_var_8: int32 = (cse_var_19 + 14)
+                  let cse_var_7: int32 = (cse_var_19 + 13)
+                  let cse_var_6: int32 = (cse_var_19 + 12)
+                  let cse_var_5: int32 = (cse_var_19 + 11)
+                  let cse_var_4: int32 = (cse_var_19 + 10)
+                  let cse_var_3: int32 = (cse_var_19 + 1)
+                   {
+                    compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_16: Buffer(placeholder_11, float32, [78656], [])[((placeholder_15[cse_var_20]*16) + cse_var_21)]*max(placeholder_17: Buffer(placeholder_10, float32, [32768], [])[(cse_var_18 + placeholder_18: Buffer(placeholder_12, int32, [4916], [])[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 1)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 2)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 3)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 4)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 5)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 6)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 7)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 8)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 9)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 10)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 11)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 12)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 13)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 14)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                    compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 15)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                  }
                 }
               }
             }
           }
-          for (i0.inner: int32, 0, 8) {
-            for (i1.inner: int32, 0, 32) {
-              let cse_var_22: int32 = ((((floordiv(i0.outer.i1.outer.fused, 16)*4096) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 16)*32)) + i1.inner)
-              compute_5: Buffer(compute_2, float32, [65536], [])[cse_var_22] = max((compute_4[((i0.inner*32) + i1.inner)] + placeholder_19: Buffer(placeholder_14, float32, [65536], [])[cse_var_22]), 0f32)
-            }
+          for (i0.inner: int32, 0, 64) {
+            let cse_var_22: int32 = (((floordiv(i0.outer.i1.outer.fused, 16)*32768) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 16)*32))
+            compute_5: Buffer(compute_2, float32, [65536], [])[ramp(cse_var_22, 1, 32)] = max((compute_4[ramp((i0.inner*32), 1, 32)] + placeholder_19: Buffer(placeholder_14, float32, [65536], [])[ramp(cse_var_22, 1, 32)]), broadcast(0f32, 32))
           }
         }
       }
@@ -512,7 +512,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 1.906 ms
+    Execution time of this operator: 1.724 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 25f8446b3e..cfa297d834 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
@@ -5,12 +5,12 @@
 
 Computation times
 =================
-**00:35.755** total execution time for **how_to_tune_with_autotvm** files:
+**00:46.368** total execution time for **how_to_tune_with_autotvm** files:
 
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)           | 00:35.720 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)           | 00:46.333 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)               | 00:00.019 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)               | 00:00.020 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``)             | 00:00.005 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
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 948d1f6800..5ecb3559db 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
@@ -265,8 +265,7 @@ for this template
     waiting for device...
     device available
     Get devices for measurement successfully!
-    No: 1   GFLOPS: 2.32/2.32       result: MeasureResult(costs=(0.099728897,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.013670921325684, timestamp=1668620461.9634821) [('tile_f', [-1, 8, 8, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1781024
-    No: 2   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+    No: 1   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -388,9 +387,26 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 16, 1, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10440020
-    No: 3   GFLOPS: 1.98/2.32       result: MeasureResult(costs=(0.11707182299999999,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.8137314319610596, timestamp=1668620464.7798214)        [('tile_f', [-1, 128, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2561687
-    No: 4   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 1, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2927506
+    No: 2   GFLOPS: 0.00/0.00       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
+        return self.__get_result()
+      File "/usr/lib/python3.7/concurrent/futures/_base.py", line 384, in __get_result
+        raise self._exception
+      File "/usr/lib/python3.7/concurrent/futures/thread.py", line 57, in run
+        result = self.fn(*self.args, **self.kwargs)
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 432, in <lambda>
+        worker = lambda *args: self._worker_run(*args)
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 401, in _worker_run
+        return proc.recv()
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 309, in recv
+        raise TimeoutError()
+    TimeoutError
+
+            [('tile_f', [-1, 1, 1, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4310425
+    No: 3   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -512,8 +528,9 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3355501
-    No: 5   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 64, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5607743
+    No: 4   GFLOPS: 397.48/397.48   result: MeasureResult(costs=(0.0005824169812206572,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2404870986938477, timestamp=1668621352.333588)       [('tile_f', [-1, 4, 1, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8337662
+    No: 5   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -635,8 +652,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 2, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3036631
-    No: 6   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 256, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 64, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7765172
+    No: 6   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -758,9 +775,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 512, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7778477
-    No: 7   GFLOPS: 10.83/10.83     result: MeasureResult(costs=(0.021368804833333335,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.6941542625427246, timestamp=1668620468.6158936)       [('tile_f', [-1, 32, 8, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3326432
-    No: 8   GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 256, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7259219
+    No: 7   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -882,8 +898,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 256, 2, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6304998
-    No: 9   GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 64, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5392113
+    No: 8   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1005,8 +1021,9 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 4, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,609645
-    No: 10  GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 16, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1216947
+    No: 9   GFLOPS: 851.04/851.04   result: MeasureResult(costs=(0.0002720207644110276,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.3329086303710938, timestamp=1668621356.4669878)      [('tile_f', [-1, 1, 8, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3497859
+    No: 10  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1128,10 +1145,9 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9569677
-    No: 11  GFLOPS: 160.31/160.31   result: MeasureResult(costs=(0.0014441192545454544,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.183053970336914, timestamp=1668620471.017138)        [('tile_f', [-1, 2, 4, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9412773
-    No: 12  GFLOPS: 15.38/160.31    result: MeasureResult(costs=(0.015050043285714285,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.1918892860412598, timestamp=1668620471.784751)        [('tile_f', [-1, 8, 16, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2520449
-    No: 13  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 512, 1, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6432369
+    No: 11  GFLOPS: 40.76/851.04    result: MeasureResult(costs=(0.005679272050000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4551396369934082, timestamp=1668621357.2073452)       [('tile_f', [-1, 4, 1, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6203782
+    No: 12  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1253,8 +1269,9 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 2, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4276066
-    No: 14  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 16, 2, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6857028
+    No: 13  GFLOPS: 2.45/851.04     result: MeasureResult(costs=(0.09452290325,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.229509353637695, timestamp=1668621369.0406008)       [('tile_f', [-1, 32, 2, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7025275
+    No: 14  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1376,8 +1393,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 4, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3255932
-    No: 15  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 4, 64, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4547496
+    No: 15  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1499,8 +1516,26 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 32, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3803233
-    No: 16  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1284555
+    No: 16  GFLOPS: 0.00/851.04     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
+        return self.__get_result()
+      File "/usr/lib/python3.7/concurrent/futures/_base.py", line 384, in __get_result
+        raise self._exception
+      File "/usr/lib/python3.7/concurrent/futures/thread.py", line 57, in run
+        result = self.fn(*self.args, **self.kwargs)
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 432, in <lambda>
+        worker = lambda *args: self._worker_run(*args)
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 401, in _worker_run
+        return proc.recv()
+      File "/workspace/python/tvm/contrib/popen_pool.py", line 309, in recv
+        raise TimeoutError()
+    TimeoutError
+
+            [('tile_f', [-1, 8, 1, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7284038
+    No: 17  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1622,9 +1657,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 1, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5578288
-    No: 17  GFLOPS: 36.23/160.31    result: MeasureResult(costs=(0.006389735090909091,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.769231081008911, timestamp=1668620476.9480226)        [('tile_f', [-1, 1, 4, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4576115
-    No: 18  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 4, 32, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9604632
+    No: 18  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1746,8 +1780,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 1, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7713998
-    No: 19  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 16, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5827079
+    No: 19  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
@@ -1869,130 +1903,8 @@ for this template
       File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 2, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8664231
-    No: 20  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 588, in __call__
-        func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 540, in _build_func_common
-        func = build(s, args, target_host=task.target_host, runtime=runtime)
-      File "/workspace/python/tvm/driver/build_module.py", line 227, in build
-        input_mod = lower(inputs, args, name=name, binds=binds)
-      File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
-        return ffi.lower_schedule(inp, args, name, binds, simple_mode)
-      File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
-      File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
-      File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-    tvm._ffi.base.TVMError: Traceback (most recent call last):
-      24: TVMFuncCall
-            at ../src/runtime/c_runtime_api.cc:477
-      23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-            at ../include/tvm/runtime/packed_func.h:1217
-      22: Call
-            at ../include/tvm/runtime/packed_func.h:1213
-      21: operator()
-            at ../include/tvm/runtime/packed_func.h:1731
-      20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
-            at ../include/tvm/runtime/packed_func.h:1671
-      19: run<>
-            at ../include/tvm/runtime/packed_func.h:1631
-      18: run<tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1646
-      13: operator()
-            at ../src/driver/driver_api.cc:388
-      12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
-            at ../src/driver/driver_api.cc:374
-      11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
-            at ../src/driver/driver_api.cc:269
-      10: tvm::transform::Pass::operator()(tvm::IRModule) const
-            at ../src/ir/transform.cc:258
-      9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:274
-      8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:453
-      7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:274
-      6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/tir/ir/transform.cc:100
-      5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
-            at ../include/tvm/runtime/packed_func.h:1750
-      4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
-            at ../include/tvm/runtime/packed_func.h:1694
-      3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
-            at ../include/tvm/runtime/packed_func.h:1618
-      2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-            at ../include/tvm/runtime/packed_func.h:1217
-      1: Call
-            at ../include/tvm/runtime/packed_func.h:1213
-      0: operator()
-            at ../src/runtime/c_runtime_api.cc:534
-      File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, in verify_pass
-        raise InstantiationError("Skipped because of invalid gpu kernel")
-    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-    Traceback (most recent call last):
-      24: TVMFuncCall
-            at ../src/runtime/c_runtime_api.cc:477
-      23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-            at ../include/tvm/runtime/packed_func.h:1217
-      22: Call
-            at ../include/tvm/runtime/packed_func.h:1213
-      21: operator()
-            at ../include/tvm/runtime/packed_func.h:1731
-      20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
-            at ../include/tvm/runtime/packed_func.h:1671
-      19: run<>
-            at ../include/tvm/runtime/packed_func.h:1631
-      18: run<tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1631
-      14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1646
-      13: operator()
-            at ../src/driver/driver_api.cc:388
-      12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
-            at ../src/driver/driver_api.cc:374
-      11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
-            at ../src/driver/driver_api.cc:269
-      10: tvm::transform::Pass::operator()(tvm::IRModule) const
-            at ../src/ir/transform.cc:258
-      9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:274
-      8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:453
-      7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/ir/transform.cc:274
-      6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
-            at ../src/tir/ir/transform.cc:100
-      5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
-            at ../include/tvm/runtime/packed_func.h:1750
-      4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
-            at ../include/tvm/runtime/packed_func.h:1694
-      3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
-            at ../include/tvm/runtime/packed_func.h:1618
-      2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-            at ../include/tvm/runtime/packed_func.h:1217
-      1: Call
-            at ../include/tvm/runtime/packed_func.h:1213
-      0: operator()
-            at ../src/runtime/c_runtime_api.cc:534
-      File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 871, 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, 256, 1]), ('tile_y', [-1, 7, 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,943632
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 64, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7500508
+    No: 20  GFLOPS: 28.37/851.04    result: MeasureResult(costs=(0.008161008923076923,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.737098217010498, timestamp=1668621372.0138779)        [('tile_f', [-1, 1, 4, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6845812
 
 
 
@@ -2047,9 +1959,9 @@ and measure running time.
     Finish loading 20 records
 
     Best config:
-    [('tile_f', [-1, 2, 4, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9412773
+    [('tile_f', [-1, 1, 8, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3497859
     Finish loading 20 records
-    Time cost of this operator: 0.001880
+    Time cost of this operator: 0.000728
 
 
 
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 9a0fc4929c..acd98e7eb8 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
@@ -327,10 +327,10 @@ Timing the untuned program
     ########## Build without Autotuning ##########
     Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs  Measurements(us)  
     ---------                                     ---                                           --------  -------  -----              ------  -------  ----------------  
-    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  309.4     98.728   (1, 2, 10, 10, 3)  2       1        [309.4]           
-    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.023     0.965    (1, 6, 10, 10)     1       1        [3.023]           
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.962     0.307    (1, 1, 10, 10, 3)  1       1        [0.962]           
-    Total_time                                    -                                             313.386   -        -                  -       -        -                 
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  311.1     98.715   (1, 2, 10, 10, 3)  2       1        [311.1]           
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.082     0.978    (1, 6, 10, 10)     1       1        [3.082]           
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.969     0.307    (1, 1, 10, 10, 3)  1       1        [0.969]           
+    Total_time                                    -                                             315.151   -        -                  -       -        -                 
 
 
 
@@ -394,10 +394,10 @@ Timing the tuned program
     ########## Build with Autotuning ##########
     Node Name                                     Ops                                           Time(us)  Time(%)  Shape              Inputs  Outputs  Measurements(us)  
     ---------                                     ---                                           --------  -------  -----              ------  -------  ----------------  
-    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  101.8     97.407   (1, 6, 10, 10, 1)  2       1        [101.8]           
-    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.756     1.68     (1, 6, 10, 10)     1       1        [1.756]           
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.955     0.913    (1, 1, 10, 10, 3)  1       1        [0.955]           
-    Total_time                                    -                                             104.51    -        -                  -       -        -                 
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  102.9     97.488   (1, 6, 10, 10, 1)  2       1        [102.9]           
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.808     1.713    (1, 6, 10, 10)     1       1        [1.808]           
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.843     0.799    (1, 3, 10, 10, 1)  1       1        [0.843]           
+    Total_time                                    -                                             105.552   -        -                  -       -        -                 
 
 
 
diff --git a/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt b/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
index 9da436b52b..cfaa927c01 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
@@ -109,7 +109,7 @@ download a cat image and preprocess it to use as the model input.
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torch/ao/quantization/utils.py:281: UserWarning: must run observer before calling calculate_qparams. Returning default values.
       "must run observer before calling calculate_qparams. " +
     Downloading: "https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2_qnnpack_37f702c5.pth
-
      0%|          | 0.00/3.42M [00:00<?, ?B/s]
     76%|#######6  | 2.61M/3.42M [00:00<00:00, 27.3MB/s]
    100%|##########| 3.42M/3.42M [00:00<00:00, 27.9MB/s]
+
      0%|          | 0.00/3.42M [00:00<?, ?B/s]
    100%|##########| 3.42M/3.42M [00:00<00:00, 65.2MB/s]
     /workspace/python/tvm/relay/frontend/pytorch_utils.py:47: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
       return LooseVersion(torch_ver) > ver
     /venv/apache-tvm-py3.7/lib/python3.7/site-packages/setuptools/_distutils/version.py:346: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
@@ -314,7 +314,7 @@ Look up prediction top 1 index in 1000 class synset.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  5.341 seconds)
+   **Total running time of the script:** ( 1 minutes  5.355 seconds)
 
 
 .. _sphx_glr_download_how_to_work_with_microtvm_micro_pytorch.py:
diff --git a/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt b/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt
index 594787074d..dad11d5b67 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt
@@ -225,7 +225,7 @@ take about **2 minutes** to download the Stanford Cars, while COCO 2017 validati
  .. code-block:: none
 
 
-    '/tmp/tmpe0nb0235/images/random'
+    '/tmp/tmpa9mo0yuu/images/random'
 
 
 
@@ -316,7 +316,7 @@ objects to other stuff? We can display some examples from our datasets using ``m
 
 
 .. image-sg:: /how_to/work_with_microtvm/images/sphx_glr_micro_train_001.png
-   :alt: [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]
+   :alt: [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]
    :srcset: /how_to/work_with_microtvm/images/sphx_glr_micro_train_001.png
    :class: sphx-glr-single-img
 
@@ -325,8 +325,8 @@ objects to other stuff? We can display some examples from our datasets using ``m
 
  .. code-block:: none
 
-    /tmp/tmpe0nb0235/images/target contains 8144 images
-    /tmp/tmpe0nb0235/images/random contains 5000 images
+    /tmp/tmpa9mo0yuu/images/target contains 8144 images
+    /tmp/tmpa9mo0yuu/images/random contains 5000 images
 
 
 
@@ -501,13 +501,13 @@ the time on our validation set).
  .. code-block:: none
 
     Epoch 1/3
-    328/328 - 47s - loss: 0.2109 - accuracy: 0.9260 - val_loss: 0.1198 - val_accuracy: 0.9569 - 47s/epoch - 144ms/step
+    328/328 - 47s - loss: 0.2387 - accuracy: 0.9199 - val_loss: 0.1355 - val_accuracy: 0.9494 - 47s/epoch - 144ms/step
     Epoch 2/3
-    328/328 - 44s - loss: 0.1042 - accuracy: 0.9624 - val_loss: 0.1374 - val_accuracy: 0.9498 - 44s/epoch - 133ms/step
+    328/328 - 44s - loss: 0.1044 - accuracy: 0.9632 - val_loss: 0.1209 - val_accuracy: 0.9581 - 44s/epoch - 133ms/step
     Epoch 3/3
-    328/328 - 44s - loss: 0.0650 - accuracy: 0.9738 - val_loss: 0.1059 - val_accuracy: 0.9660 - 44s/epoch - 133ms/step
+    328/328 - 44s - loss: 0.0669 - accuracy: 0.9757 - val_loss: 0.1102 - val_accuracy: 0.9619 - 44s/epoch - 133ms/step
 
-    <keras.callbacks.History object at 0x7f34dd15a3d0>
+    <keras.callbacks.History object at 0x7f1545f86750>
 
 
 
@@ -864,7 +864,7 @@ Arduino tutorial for how to do that `on GitHub <https://github.com/guberti/tvm-a
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 4 minutes  42.129 seconds)
+   **Total running time of the script:** ( 4 minutes  50.868 seconds)
 
 
 .. _sphx_glr_download_how_to_work_with_microtvm_micro_train.py:
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 e1265c5993..30815463d0 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
@@ -5,20 +5,20 @@
 
 Computation times
 =================
-**06:51.708** total execution time for **how_to_work_with_microtvm** files:
+**07:00.828** total execution time for **how_to_work_with_microtvm** files:
 
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``)               | 04:42.129 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``)               | 04:50.868 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``)           | 01:05.341 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``)           | 01:05.355 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)         | 00:51.638 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)         | 00:51.852 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``)                   | 00:08.628 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``)                   | 00:08.831 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)             | 00:03.969 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)             | 00:03.920 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``) | 00:00.001 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``) | 00:00.002 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)             | 00:00.001 | 0.0 MB |
 +---------------------------------------------------------------------------------------------+-----------+--------+
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 2aca469e19..4c98d5b47d 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
@@ -5,14 +5,14 @@
 
 Computation times
 =================
-**00:44.784** total execution time for **how_to_work_with_relay** files:
+**00:44.968** total execution time for **how_to_work_with_relay** files:
 
 +----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:32.784 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:33.233 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)           | 00:10.272 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)           | 00:10.284 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)                             | 00:01.721 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)                             | 00:01.444 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)                 | 00:00.007 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/work_with_schedules/intrin_math.rst.txt b/docs/_sources/how_to/work_with_schedules/intrin_math.rst.txt
index 83ed44ba4f..2d8c3d806e 100644
--- a/docs/_sources/how_to/work_with_schedules/intrin_math.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/intrin_math.rst.txt
@@ -261,7 +261,7 @@ The following example customizes CUDA lowering rule for :code:`exp`.
  .. code-block:: none
 
 
-    <function my_cuda_math_rule at 0x7f330c705e60>
+    <function my_cuda_math_rule at 0x7f15942b48c0>
 
 
 
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 db4b13a3c6..fec75d256c 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
@@ -5,20 +5,20 @@
 
 Computation times
 =================
-**00:06.448** total execution time for **how_to_work_with_schedules** files:
+**00:06.163** total execution time for **how_to_work_with_schedules** files:
 
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)                 | 00:03.974 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)                 | 00:03.763 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)                     | 00:01.086 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)                     | 00:01.058 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)                     | 00:00.594 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)                     | 00:00.571 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)                               | 00:00.577 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)                               | 00:00.553 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)                     | 00:00.118 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.050 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.051 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)                               | 00:00.030 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
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 3d98da4987..7de06914ac 100644
--- a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
@@ -343,7 +343,7 @@ The importing needs to happen before the tensorized GEMV being executed.
                  B: Buffer(B_2: Pointer(float32), float32, [512, 64], []),
                  C: Buffer(C_2: Pointer(float32), float32, [1024, 512], [])}
       buffer_map = {A_1: A, B_1: B, C_1: C} {
-      attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmp8j_p51zl/input0.cc'\nsource_filename = \"/tmp/tmp8j_p51zl/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/tmpkzqa97qs/input0.cc'\nsource_filename = \"/tmp/tmpkzqa97qs/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 a0018ca40d..e3164507da 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
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**00:27.412** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:27.644** total execution time for **topic_vta_tutorials_autotvm** files:
 
 +---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:27.406 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:27.637 | 0.0 MB |
 +---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)     | 00:00.007 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)     | 00:00.006 | 0.0 MB |
 +---------------------------------------------------------------------------------------+-----------+--------+
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 2731db312e..c25a51399f 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
@@ -289,7 +289,7 @@ 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 30.35s!
+    resnet18_v1 inference graph built in 30.75s!
 
 
 
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 582dbfbde3..904137c534 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
@@ -333,7 +333,7 @@ The compilation steps are:
 
     /workspace/python/tvm/relay/build_module.py:348: DeprecationWarning: Please use input parameter mod (tvm.IRModule) instead of deprecated parameter mod (tvm.relay.function.Function)
       DeprecationWarning,
-    yolov3-tiny inference graph built in 20.43s!
+    yolov3-tiny inference graph built in 20.50s!
 
 
 
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 6f102ab2a3..3ea3f5fce9 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
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**01:43.048** total execution time for **topic_vta_tutorials_frontend** files:
+**01:43.810** total execution time for **topic_vta_tutorials_frontend** files:
 
 +------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)           | 00:52.739 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)           | 00:53.116 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.309 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.694 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------+-----------+--------+
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 1db2181846..cec547e11f 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
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**00:03.235** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.177** total execution time for **topic_vta_tutorials_optimize** files:
 
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)         | 00:02.763 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)         | 00:02.730 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.472 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.447 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
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 bf6716f4f4..1cb807ea58 100644
--- a/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
 
 Computation times
 =================
-**00:00.837** total execution time for **topic_vta_tutorials** files:
+**00:00.816** total execution time for **topic_vta_tutorials** files:
 
 +---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.445 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.444 | 0.0 MB |
 +---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.392 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.372 | 0.0 MB |
 +---------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
index 1e6e2c2f4a..abad01db10 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -203,6 +203,13 @@ trials, we can load the best schedule from the log file and apply it.
 
 
 
+.. rst-class:: sphx-glr-script-out
+
+ .. code-block:: none
+
+    .T*E
+
+
 
 
 
@@ -325,7 +332,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 100.349 ms
+    Execution time of this operator: 95.411 ms
 
 
 
@@ -443,7 +450,7 @@ operations.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  21.583 seconds)
+   **Total running time of the script:** ( 1 minutes  49.022 seconds)
 
 
 .. _sphx_glr_download_tutorial_auto_scheduler_matmul_x86.py:
diff --git a/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt b/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt
index b7f1e492b6..3e9b26d845 100644
--- a/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt
@@ -450,16 +450,16 @@ reduce variance, we take 5 measurements and average them.
     waiting for device...
     device available
     Get devices for measurement successfully!
-    No: 1   GFLOPS: 2.59/2.59       result: MeasureResult(costs=(0.10359291479999999,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.8303396701812744, timestamp=1668619046.14139)  [('tile_y', [-1, 8]), ('tile_x', [-1, 4])],None,23
-    No: 2   GFLOPS: 0.86/2.59       result: MeasureResult(costs=(0.31212432100000004,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.175588369369507, timestamp=1668619051.3432844) [('tile_y', [-1, 512]), ('tile_x', [-1, 2])],None,19
-    No: 3   GFLOPS: 12.48/12.48     result: MeasureResult(costs=(0.021505876200000003,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.48230600357055664, timestamp=1668619052.6419113)      [('tile_y', [-1, 128]), ('tile_x', [-1, 512])],None,97
-    No: 4   GFLOPS: 11.90/12.48     result: MeasureResult(costs=(0.0225636806,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5249507427215576, timestamp=1668619053.9455721)       [('tile_y', [-1, 256]), ('tile_x', [-1, 256])],None,88
-    No: 5   GFLOPS: 1.08/12.48      result: MeasureResult(costs=(0.24850028739999996,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.030014753341675, timestamp=1668619058.1830983) [('tile_y', [-1, 16]), ('tile_x', [-1, 1])],None,4
-    No: 6   GFLOPS: 1.63/12.48      result: MeasureResult(costs=(0.16475953640000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.791689157485962, timestamp=1668619061.7600245) [('tile_y', [-1, 64]), ('tile_x', [-1, 4])],None,26
-    No: 7   GFLOPS: 10.96/12.48     result: MeasureResult(costs=(0.0245022136,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6264443397521973, timestamp=1668619062.3191602)       [('tile_y', [-1, 1]), ('tile_x', [-1, 64])],None,60
-    No: 8   GFLOPS: 10.91/12.48     result: MeasureResult(costs=(0.0245933216,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5864930152893066, timestamp=1668619062.924451)        [('tile_y', [-1, 2]), ('tile_x', [-1, 256])],None,81
-    No: 9   GFLOPS: 4.13/12.48      result: MeasureResult(costs=(0.065029372,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.1726446151733398, timestamp=1668619064.210647) [('tile_y', [-1, 16]), ('tile_x', [-1, 16])],None,44
-    No: 10  GFLOPS: 2.12/12.48      result: MeasureResult(costs=(0.1264244674,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.1329433917999268, timestamp=1668619066.3955705)       [('tile_y', [-1, 512]), ('tile_x', [-1, 8])],None,39
+    No: 1   GFLOPS: 0.90/0.90       result: MeasureResult(costs=(0.2998461346,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.952210903167725, timestamp=1668619894.0839715)        [('tile_y', [-1, 256]), ('tile_x', [-1, 2])],None,18
+    No: 2   GFLOPS: 11.55/11.55     result: MeasureResult(costs=(0.0232333256,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5519640445709229, timestamp=1668619894.62132) [('tile_y', [-1, 32]), ('tile_x', [-1, 32])],None,55
+    No: 3   GFLOPS: 1.77/11.55      result: MeasureResult(costs=(0.1519943126,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.574821949005127, timestamp=1668619897.9824896)        [('tile_y', [-1, 512]), ('tile_x', [-1, 4])],None,29
+    No: 4   GFLOPS: 2.70/11.55      result: MeasureResult(costs=(0.0992638146,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7829089164733887, timestamp=1668619899.7848358)       [('tile_y', [-1, 4]), ('tile_x', [-1, 8])],None,32
+    No: 5   GFLOPS: 0.89/11.55      result: MeasureResult(costs=(0.3008661622,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.962441921234131, timestamp=1668619904.8722544)        [('tile_y', [-1, 64]), ('tile_x', [-1, 2])],None,16
+    No: 6   GFLOPS: 2.99/11.55      result: MeasureResult(costs=(0.0898423854,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.602233648300171, timestamp=1668619907.2355342)        [('tile_y', [-1, 2]), ('tile_x', [-1, 16])],None,41
+    No: 7   GFLOPS: 3.22/11.55      result: MeasureResult(costs=(0.0834807634,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4857280254364014, timestamp=1668619909.5006485)       [('tile_y', [-1, 512]), ('tile_x', [-1, 16])],None,49
+    No: 8   GFLOPS: 1.56/11.55      result: MeasureResult(costs=(0.1725276412,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.8960962295532227, timestamp=1668619912.4255238)       [('tile_y', [-1, 8]), ('tile_x', [-1, 1])],None,3
+    No: 9   GFLOPS: 9.52/11.55      result: MeasureResult(costs=(0.028189924999999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.575998067855835, timestamp=1668619913.115027) [('tile_y', [-1, 4]), ('tile_x', [-1, 32])],None,52
+    No: 10  GFLOPS: 1.48/11.55      result: MeasureResult(costs=(0.18186832039999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.056715726852417, timestamp=1668619916.2314937) [('tile_y', [-1, 1]), ('tile_x', [-1, 1])],None,0
 
 
 
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index c8075c8422..2cb8f95281 100644
--- a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
@@ -320,7 +320,7 @@ standard deviation.
 
  .. code-block:: none
 
-    {'mean': 521.135307909999, 'median': 521.1086226000134, 'std': 2.045413370611934}
+    {'mean': 520.7545591599956, 'median': 520.3844177999827, 'std': 2.606245863239369}
 
 
 
@@ -554,32 +554,30 @@ the tuning data to.
 
  .. code-block:: none
 
-
    [Task  1/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  1/25]  Current/Best:   15.05/  23.80 GFLOPS | Progress: (4/20) | 6.92 s
    [Task  1/25]  Current/Best:   21.97/  23.80 GFLOPS | Progress: (8/20) | 10.22 s
    [Task  1/25]  Current/Best:   23.34/  23.80 GFLOPS | Progress: (12/20) | 16.23 s
    [Task  1/25]  Current/Best:   15.60/  23.80 GFLOPS | Progress: (16/20) | 18.93 s
    [Task  1/25]  Current/Best:   11.13/  23.80 GFLOPS | Progress: (20/20) | 20.94 s Done.
-
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  2/25]  Current/Best:    7.90/  17.91 GFLOPS | Progress: (4/20) | 2.90 s
    [Task  2/25]  Current/Best:   18.50/  18.50 GFLOPS | Progress: (8/20) | 3.85 s
    [Task  2/25]  Current/Best:    3.32/  18.50 GFLOPS | Progress: (12/20) | 5.01 s
    [Task  2/25]  Current/Best:   14.53/  18.50 GFLOPS | Progress: (16/20) | 7.80 s
    [Task  2/25]  Current/Best:    8.66/  18.50 GFLOPS | Progress: (20/20) | 9.17 s Done.
-
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  3/25]  Current/Best:   14.96/  14.96 GFLOPS | Progress: (4/20) | 3.55 s
    [Task  3/25]  Current/Best:   13.00/  14.96 GFLOPS | Progress: (8/20) | 5.83 s
    [Task  3/25]  Current/Best:   13.74/  17.98 GFLOPS | Progress: (12/20) | 8.25 s
    [Task  3/25]  Current/Best:    9.13/  23.69 GFLOPS | Progress: (16/20) | 10.97 s
    [Task  3/25]  Current/Best:   14.67/  23.69 GFLOPS | Progress: (20/20) | 12.98 s Done.
-
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  4/25]  Current/Best:   12.29/  18.72 GFLOPS | Progress: (4/20) | 6.03 s
    [Task  4/25]  Current/Best:    3.38/  18.72 GFLOPS | Progress: (8/20) | 8.95 s
    [Task  4/25]  Current/Best:    7.92/  19.09 GFLOPS | Progress: (12/20) | 11.03 s
    [Task  4/25]  Current/Best:   19.22/  19.22 GFLOPS | Progress: (16/20) | 17.10 s
    [Task  4/25]  Current/Best:   17.34/  19.22 GFLOPS | Progress: (20/20) | 18.37 s Done.
-
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  5/25]  Current/Best:    9.49/  19.34 GFLOPS | Progress: (4/20) | 3.50 s
    [Task  5/25]  Current/Best:   15.98/  19.34 GFLOPS | Progress: (8/20) | 5.22 s
    [Task  5/25]  Current/Best:   11.01/  20.06 GFLOPS | Progress: (12/20) | 8.19 s
    [Task  5/25]  Current/Best:   19.88/  20.06 GFLOPS | Progress: (16/20) | 10.39 s
    [Task  5/25]  Current/Best:   14.48/  20.06 GFLOPS | Progress: (20/20) | 12.38 s Done.
-
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  6/25]  Current/Best:    7.16/  12.92 GFLOPS | Progress: (4/20) | 4.31 s
    [Task  6/25]  Current/Best:    7.94/  14.62 GFLOPS | Progress: (8/20) | 6.69 s
    [Task  6/25]  Current/Best:   14.69/  14.69 GFLOPS | Progress: (12/20) | 10.87 s
    [Task  6/25]  Current/Best:    8.36/  14.69 GFLOPS | Progress: (16/20) | 13.61 s
    [Task  6/25]  Current/Best:   13.55/  14.69 GFLOPS | Progress: (20/20) | 16.12 s Done.
-
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  7/25]  Current/Best:    5.30/  19.41 GFLOPS | Progress: (4/20) | 3.88 s
    [Task  7/25]  Current/Best:   17.39/  19.41 GFLOPS | Progress: (8/20) | 5.92 s
    [Task  7/25]  Current/Best:   14.12/  19.41 GFLOPS | Progress: (12/20) | 9.25 s
    [Task  7/25]  Current/Best:   14.46/  20.93 GFLOPS | Progress: (16/20) | 12.96 s
    [Task  7/25]  Current/Best:   11.59/  20.93 GFLOPS | Progress: (20/20) | 15.61 s Done.
-
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  8/25]  Current/Best:    9.95/  17.55 GFLOPS | Progress: (4/20) | 4.78 s
    [Task  8/25]  Current/Best:    9.71/  17.55 GFLOPS | Progress: (8/20) | 6.96 s
    [Task  8/25]  Current/Best:   11.13/  18.83 GFLOPS | Progress: (12/20) | 9.31 s
    [Task  8/25]  Current/Best:   16.02/  18.83 GFLOPS | Progress: (16/20) | 11.50 s
    [Task  8/25]  Current/Best:   11.55/  18.83 GFLOPS | Progress: (20/20) | 13.76 s Done.
-
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  9/25]  Current/Best:   12.43/  18.23 GFLOPS | Progress: (4/20) | 2.95 s
    [Task  9/25]  Current/Best:   12.34/  18.23 GFLOPS | Progress: (8/20) | 4.85 s
    [Task  9/25]  Current/Best:   17.50/  18.23 GFLOPS | Progress: (12/20) | 8.82 s
    [Task  9/25]  Current/Best:    4.53/  18.23 GFLOPS | Progress: (16/20) | 10.48 s
    [Task  9/25]  Current/Best:   17.16/  18.23 GFLOPS | Progress: (20/20) | 12.97 s Done.
-
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 10/25]  Current/Best:   13.34/  13.34 GFLOPS | Progress: (4/20) | 3.18 s
    [Task 10/25]  Current/Best:   18.36/  18.36 GFLOPS | Progress: (8/20) | 4.76 s
    [Task 10/25]  Current/Best:   10.32/  19.02 GFLOPS | Progress: (12/20) | 6.81 s
    [Task 10/25]  Current/Best:    4.68/  19.02 GFLOPS | Progress: (16/20) | 8.48 s
    [Task 10/25]  Current/Best:   10.74/  19.02 GFLOPS | Progress: (20/20) | 11.13 s Done.
-
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 11/25]  Current/Best:   18.27/  18.27 GFLOPS | Progress: (4/20) | 3.58 s
    [Task 11/25]  Current/Best:    7.78/  18.27 GFLOPS | Progress: (8/20) | 6.25 s
    [Task 11/25]  Current/Best:   11.59/  18.98 GFLOPS | Progress: (12/20) | 9.74 s
    [Task 11/25]  Current/Best:   19.05/  19.05 GFLOPS | Progress: (16/20) | 11.89 s
    [Task 11/25]  Current/Best:   10.89/  19.05 GFLOPS | Progress: (20/20) | 14.33 s Done.
-
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 12/25]  Current/Best:   14.58/  14.58 GFLOPS | Progress: (4/20) | 3.72 s
    [Task 12/25]  Current/Best:   18.08/  18.08 GFLOPS | Progress: (8/20) | 6.95 s
    [Task 12/25]  Current/Best:    8.90/  18.08 GFLOPS | Progress: (12/20) | 10.81 s
    [Task 12/25]  Current/Best:    8.18/  22.41 GFLOPS | Progress: (16/20) | 16.04 s
    [Task 12/25]  Current/Best:   11.51/  22.41 GFLOPS | Progress: (20/20) | 18.62 s Done.
-
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 13/25]  Current/Best:    6.98/  12.20 GFLOPS | Progress: (4/20) | 4.76 s
    [Task 13/25]  Current/Best:   17.33/  17.33 GFLOPS | Progress: (8/20) | 8.29 s
    [Task 13/25]  Current/Best:   16.65/  17.33 GFLOPS | Progress: (12/20) | 11.16 s
    [Task 13/25]  Current/Best:   11.19/  21.89 GFLOPS | Progress: (16/20) | 14.44 s
    [Task 13/25]  Current/Best:   19.44/  21.89 GFLOPS | Progress: (20/20) | 17.26 s Done.
-
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 14/25]  Current/Best:   19.91/  19.91 GFLOPS | Progress: (4/20) | 4.73 s
    [Task 14/25]  Current/Best:    4.86/  19.91 GFLOPS | Progress: (8/20) | 7.75 s
    [Task 14/25]  Current/Best:   11.85/  19.91 GFLOPS | Progress: (12/20) | 10.42 s
    [Task 14/25]  Current/Best:   12.03/  21.32 GFLOPS | Progress: (16/20) | 14.02 s
    [Task 14/25]  Current/Best:   11.51/  21.32 GFLOPS | Progress: (20/20) | 16.04 s
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
-
    [Task 15/25]  Current/Best:    3.20/  17.75 GFLOPS | Progress: (4/20) | 3.77 s
    [Task 15/25]  Current/Best:   18.27/  21.63 GFLOPS | Progress: (8/20) | 6.45 s
    [Task 15/25]  Current/Best:   11.55/  21.63 GFLOPS | Progress: (12/20) | 10.22 s
    [Task 15/25]  Current/Best:   19.05/  21.63 GFLOPS | Progress: (16/20) | 12.13 s
    [Task 15/25]  Current/Best:   12.32/  21.63 GFLOPS | Progress: (20/20) | 14.32 s
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 16/25]  Current/Best:   15.80/  17.62 GFLOPS | Progress: (4/20) | 3.71 s
    [Task 16/25]  Current/Best:   20.37/  20.37 GFLOPS | Progress: (8/20) | 5.51 s
    [Task 16/25]  Current/Best:   22.67/  22.67 GFLOPS | Progress: (12/20) | 7.03 s
    [Task 16/25]  Current/Best:   16.72/  22.67 GFLOPS | Progress: (16/20) | 9.52 s
    [Task 16/25]  Current/Best:   17.86/  22.67 GFLOPS | Progress: (20/20) | 11.82 s Done.
-
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 17/25]  Current/Best:    6.18/  15.78 GFLOPS | Progress: (4/20) | 4.35 s
    [Task 17/25]  Current/Best:    9.64/  15.78 GFLOPS | Progress: (8/20) | 7.96 s
    [Task 17/25]  Current/Best:    6.03/  19.53 GFLOPS | Progress: (12/20) | 11.37 s
    [Task 17/25]  Current/Best:   21.62/  21.62 GFLOPS | Progress: (16/20) | 14.65 s
    [Task 17/25]  Current/Best:   13.06/  21.62 GFLOPS | Progress: (20/20) | 16.92 s Done.
-
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 18/25]  Current/Best:    2.99/  16.50 GFLOPS | Progress: (4/20) | 3.83 s
    [Task 18/25]  Current/Best:   10.49/  20.92 GFLOPS | Progress: (8/20) | 10.24 s
    [Task 18/25]  Current/Best:    9.57/  20.92 GFLOPS | Progress: (12/20) | 17.95 s
    [Task 18/25]  Current/Best:   14.32/  20.92 GFLOPS | Progress: (16/20) | 19.68 s
    [Task 18/25]  Current/Best:   15.02/  20.92 GFLOPS | Progress: (20/20) | 22.34 s Done.
-
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 19/25]  Current/Best:    7.39/  17.00 GFLOPS | Progress: (4/20) | 4.71 s
    [Task 19/25]  Current/Best:   18.49/  18.49 GFLOPS | Progress: (8/20) | 8.50 s
    [Task 19/25]  Current/Best:    9.84/  18.49 GFLOPS | Progress: (12/20) | 12.85 s
    [Task 19/25]  Current/Best:   15.24/  18.49 GFLOPS | Progress: (16/20) | 17.12 s
    [Task 19/25]  Current/Best:    7.83/  18.49 GFLOPS | Progress: (20/20) | 21.34 s Done.
-
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 20/25]  Current/Best:   11.56/  17.55 GFLOPS | Progress: (4/20) | 4.13 s
    [Task 20/25]  Current/Best:    2.71/  19.22 GFLOPS | Progress: (8/20) | 6.47 s
    [Task 20/25]  Current/Best:   15.37/  19.22 GFLOPS | Progress: (12/20) | 9.01 s
    [Task 20/25]  Current/Best:   13.83/  20.14 GFLOPS | Progress: (16/20) | 11.23 s
    [Task 20/25]  Current/Best:   12.42/  20.14 GFLOPS | Progress: (20/20) | 14.07 s
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 21/25]  Current/Best:    7.06/  15.55 GFLOPS | Progress: (4/20) | 2.98 s
    [Task 21/25]  Current/Best:    9.11/  15.55 GFLOPS | Progress: (8/20) | 6.48 s
    [Task 21/25]  Current/Best:   20.61/  20.61 GFLOPS | Progress: (12/20) | 9.92 s
    [Task 21/25]  Current/Best:   18.12/  21.48 GFLOPS | Progress: (16/20) | 11.09 s Done.
+
    [Task  1/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  1/25]  Current/Best:    9.62/  14.90 GFLOPS | Progress: (4/20) | 8.90 s
    [Task  1/25]  Current/Best:   23.14/  23.14 GFLOPS | Progress: (8/20) | 12.28 s
    [Task  1/25]  Current/Best:   21.53/  23.14 GFLOPS | Progress: (12/20) | 15.57 s
    [Task  1/25]  Current/Best:   15.32/  23.14 GFLOPS | Progress: (16/20) | 18.25 s
    [Task  1/25]  Current/Best:   17.88/  23.14 GFLOPS | Progress: (20/20) | 20.75 s Done.
+
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  2/25]  Current/Best:   12.05/  17.05 GFLOPS | Progress: (4/20) | 4.41 s
    [Task  2/25]  Current/Best:   15.60/  17.05 GFLOPS | Progress: (8/20) | 6.08 s
    [Task  2/25]  Current/Best:   11.84/  17.05 GFLOPS | Progress: (12/20) | 8.08 s
    [Task  2/25]  Current/Best:   12.94/  18.03 GFLOPS | Progress: (16/20) | 9.98 s
    [Task  2/25]  Current/Best:   13.74/  18.03 GFLOPS | Progress: (20/20) | 11.22 s Done.
+
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  3/25]  Current/Best:   17.23/  20.05 GFLOPS | Progress: (4/20) | 3.66 s
    [Task  3/25]  Current/Best:   22.38/  22.38 GFLOPS | Progress: (8/20) | 5.40 s
    [Task  3/25]  Current/Best:   19.38/  22.38 GFLOPS | Progress: (12/20) | 8.10 s
    [Task  3/25]  Current/Best:   19.37/  23.48 GFLOPS | Progress: (16/20) | 9.76 s
    [Task  3/25]  Current/Best:   22.73/  23.48 GFLOPS | Progress: (20/20) | 11.75 s Done.
+
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  4/25]  Current/Best:    9.18/  19.03 GFLOPS | Progress: (4/20) | 7.58 s
    [Task  4/25]  Current/Best:   18.36/  19.03 GFLOPS | Progress: (8/20) | 10.77 s
    [Task  4/25]  Current/Best:    6.12/  19.60 GFLOPS | Progress: (12/20) | 12.15 s
    [Task  4/25]  Current/Best:    4.90/  19.60 GFLOPS | Progress: (16/20) | 23.34 s
    [Task  4/25]  Current/Best:   12.96/  19.60 GFLOPS | Progress: (20/20) | 26.15 s Done.
+
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  5/25]  Current/Best:   11.50/  19.06 GFLOPS | Progress: (4/20) | 2.96 s
    [Task  5/25]  Current/Best:   10.87/  21.70 GFLOPS | Progress: (8/20) | 4.96 s
    [Task  5/25]  Current/Best:    7.09/  21.70 GFLOPS | Progress: (12/20) | 6.67 s
    [Task  5/25]  Current/Best:   12.72/  21.70 GFLOPS | Progress: (16/20) | 8.38 s
    [Task  5/25]  Current/Best:   14.25/  21.70 GFLOPS | Progress: (20/20) | 10.58 s Done.
+
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  6/25]  Current/Best:   16.51/  16.51 GFLOPS | Progress: (4/20) | 5.04 s
    [Task  6/25]  Current/Best:    6.01/  16.51 GFLOPS | Progress: (8/20) | 8.63 s
    [Task  6/25]  Current/Best:   12.65/  16.51 GFLOPS | Progress: (12/20) | 15.93 s
    [Task  6/25]  Current/Best:    4.39/  16.51 GFLOPS | Progress: (16/20) | 18.74 s
    [Task  6/25]  Current/Best:    4.28/  16.51 GFLOPS | Progress: (20/20) | 22.37 s Done.
+
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  7/25]  Current/Best:   11.61/  18.06 GFLOPS | Progress: (4/20) | 4.80 s
    [Task  7/25]  Current/Best:   10.21/  18.06 GFLOPS | Progress: (8/20) | 8.54 s
    [Task  7/25]  Current/Best:   14.63/  18.06 GFLOPS | Progress: (12/20) | 10.66 s
    [Task  7/25]  Current/Best:    9.57/  18.06 GFLOPS | Progress: (16/20) | 12.65 s
    [Task  7/25]  Current/Best:   14.78/  18.06 GFLOPS | Progress: (20/20) | 14.68 s Done.
+
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  8/25]  Current/Best:   11.79/  14.69 GFLOPS | Progress: (4/20) | 6.20 s
    [Task  8/25]  Current/Best:   14.58/  15.90 GFLOPS | Progress: (8/20) | 7.88 s
    [Task  8/25]  Current/Best:   18.03/  18.03 GFLOPS | Progress: (12/20) | 19.60 s
    [Task  8/25]  Current/Best:    4.99/  18.03 GFLOPS | Progress: (16/20) | 26.14 s
    [Task  8/25]  Current/Best:    2.35/  18.03 GFLOPS | Progress: (20/20) | 31.59 s
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  9/25]  Current/Best:    5.81/  19.11 GFLOPS | Progress: (4/20) | 3.12 s
    [Task  9/25]  Current/Best:    8.68/  19.11 GFLOPS | Progress: (8/20) | 8.67 s
    [Task  9/25]  Current/Best:   19.04/  21.98 GFLOPS | Progress: (12/20) | 10.09 s
    [Task  9/25]  Current/Best:   12.43/  21.98 GFLOPS | Progress: (16/20) | 11.48 s
    [Task  9/25]  Current/Best:   15.45/  21.98 GFLOPS | Progress: (20/20
 ) | 12.98 s Done.
+
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 10/25]  Current/Best:   13.71/  19.82 GFLOPS | Progress: (4/20) | 3.53 s
    [Task 10/25]  Current/Best:   12.44/  19.82 GFLOPS | Progress: (8/20) | 5.05 s
    [Task 10/25]  Current/Best:   12.85/  19.82 GFLOPS | Progress: (12/20) | 7.12 s
    [Task 10/25]  Current/Best:    8.34/  19.82 GFLOPS | Progress: (16/20) | 10.03 s
    [Task 10/25]  Current/Best:    4.83/  19.82 GFLOPS | Progress: (20/20) | 13.54 s Done.
+
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 11/25]  Current/Best:   16.61/  18.52 GFLOPS | Progress: (4/20) | 3.84 s
    [Task 11/25]  Current/Best:    9.63/  22.60 GFLOPS | Progress: (8/20) | 5.68 s
    [Task 11/25]  Current/Best:   12.28/  22.60 GFLOPS | Progress: (12/20) | 8.20 s
    [Task 11/25]  Current/Best:    9.41/  22.60 GFLOPS | Progress: (16/20) | 10.15 s
    [Task 11/25]  Current/Best:   17.22/  22.60 GFLOPS | Progress: (20/20) | 12.35 s Done.
+
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 12/25]  Current/Best:   20.07/  20.07 GFLOPS | Progress: (4/20) | 3.49 s
    [Task 12/25]  Current/Best:   10.32/  20.07 GFLOPS | Progress: (8/20) | 10.83 s
    [Task 12/25]  Current/Best:    9.54/  20.07 GFLOPS | Progress: (12/20) | 13.98 s
    [Task 12/25]  Current/Best:   14.36/  20.07 GFLOPS | Progress: (16/20) | 16.10 s
    [Task 12/25]  Current/Best:   15.11/  20.07 GFLOPS | Progress: (20/20) | 18.37 s Done.
+
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 13/25]  Current/Best:    8.67/  11.71 GFLOPS | Progress: (4/20) | 5.08 s
    [Task 13/25]  Current/Best:   10.87/  16.72 GFLOPS | Progress: (8/20) | 7.77 s
    [Task 13/25]  Current/Best:    6.45/  21.20 GFLOPS | Progress: (12/20) | 10.87 s
    [Task 13/25]  Current/Best:   20.36/  21.20 GFLOPS | Progress: (16/20) | 13.83 s
    [Task 13/25]  Current/Best:   21.75/  21.75 GFLOPS | Progress: (20/20) | 15.85 s Done.
+
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 14/25]  Current/Best:   13.22/  18.97 GFLOPS | Progress: (4/20) | 3.22 s
    [Task 14/25]  Current/Best:   16.28/  22.21 GFLOPS | Progress: (8/20) | 5.52 s
    [Task 14/25]  Current/Best:   15.84/  22.21 GFLOPS | Progress: (12/20) | 8.95 s
    [Task 14/25]  Current/Best:   14.36/  22.21 GFLOPS | Progress: (16/20) | 15.59 s
    [Task 14/25]  Current/Best:    4.99/  22.21 GFLOPS | Progress: (20/20) | 22.81 s
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
      Done.
-
    [Task 21/25]  Current/Best:   10.95/  21.69 GFLOPS | Progress: (20/20) | 13.57 s Done.
-
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 22/25]  Current/Best:   12.03/  17.70 GFLOPS | Progress: (4/20) | 2.87 s
    [Task 22/25]  Current/Best:    9.89/  17.70 GFLOPS | Progress: (8/20) | 4.33 s
    [Task 22/25]  Current/Best:   18.35/  18.35 GFLOPS | Progress: (12/20) | 5.95 s
    [Task 22/25]  Current/Best:    6.14/  18.35 GFLOPS | Progress: (16/20) | 8.24 s
    [Task 22/25]  Current/Best:   12.32/  18.35 GFLOPS | Progress: (20/20) | 10.35 s Done.
-
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 23/25]  Current/Best:   21.15/  21.15 GFLOPS | Progress: (4/20) | 3.56 s
    [Task 23/25]  Current/Best:    3.08/  21.15 GFLOPS | Progress: (8/20) | 7.43 s
    [Task 23/25]  Current/Best:   12.91/  21.15 GFLOPS | Progress: (12/20) | 9.98 s
    [Task 23/25]  Current/Best:   14.52/  21.45 GFLOPS | Progress: (16/20) | 12.58 s
    [Task 23/25]  Current/Best:    9.85/  21.45 GFLOPS | Progress: (20/20) | 15.09 s Done.
-
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 24/25]  Current/Best:    5.72/   9.83 GFLOPS | Progress: (4/20) | 4.45 s
    [Task 24/25]  Current/Best:    3.19/   9.83 GFLOPS | Progress: (8/20) | 8.93 s
    [Task 24/25]  Current/Best:    3.25/   9.83 GFLOPS | Progress: (12/20) | 10.79 s
    [Task 24/25]  Current/Best:    1.20/   9.83 GFLOPS | Progress: (16/20) | 21.51 s
    [Task 24/25]  Current/Best:    2.94/   9.83 GFLOPS | Progress: (20/20) | 32.38 s
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 25/25]  Current/Best:    1.55/   3.42 GFLOPS | Progress: (4/20) | 12.33 s
    [Task 25/25]  Current/Best:    5.52/   5.52 GFLOPS | Progress: (8/20) | 17.66 s
    [Task 25/25]  Current/Best:    7.17/   7.76 GFLOPS | Progress: (12/20) | 21.97 s
    [Task 25/25]  Current/Best:    2.73/   7.76 GFLOPS | Progress: (16/20) | 32.72 s Done.
-
    [Task 25/25]  Current/Best:    4.83/   7.76 GFLOPS | Progress: (20/20) | 36.69 s Done.
-
+
    [Task 15/25]  Current/Best:   11.33/  18.65 GFLOPS | Progress: (4/20) | 3.23 s
    [Task 15/25]  Current/Best:   19.27/  19.27 GFLOPS | Progress: (8/20) | 5.49 s
    [Task 15/25]  Current/Best:   16.15/  19.27 GFLOPS | Progress: (12/20) | 8.47 s
    [Task 15/25]  Current/Best:   21.13/  21.13 GFLOPS | Progress: (16/20) | 9.79 s
    [Task 15/25]  Current/Best:   19.04/  23.72 GFLOPS | Progress: (20/20) | 12.23 s
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 16/25]  Current/Best:    4.97/  14.54 GFLOPS | Progress: (4/20) | 3.18 s
    [Task 16/25]  Current/Best:   10.55/  16.11 GFLOPS | Progress: (8/20) | 5.15 s
    [Task 16/25]  Current/Best:   18.55/  18.55 GFLOPS | Progress: (12/20) | 7.36 s
    [Task 16/25]  Current/Best:   11.59/  18.55 GFLOPS | Progress: (16/20) | 9.12 s
    [Task 16/25]  Current/Best:   11.38/  18.55 GFLOPS | Progress: (20/20) | 12.40 s Done.
+
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 17/25]  Current/Best:    5.07/  11.98 GFLOPS | Progress: (4/20) | 4.61 s
    [Task 17/25]  Current/Best:   16.35/  18.57 GFLOPS | Progress: (8/20) | 6.65 s
    [Task 17/25]  Current/Best:    9.66/  21.79 GFLOPS | Progress: (12/20) | 9.10 s
    [Task 17/25]  Current/Best:    9.31/  21.79 GFLOPS | Progress: (16/20) | 11.76 s
    [Task 17/25]  Current/Best:    8.94/  22.03 GFLOPS | Progress: (20/20) | 14.33 s Done.
+
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 18/25]  Current/Best:   13.50/  15.97 GFLOPS | Progress: (4/20) | 5.20 s
    [Task 18/25]  Current/Best:    9.55/  19.57 GFLOPS | Progress: (8/20) | 10.48 s
    [Task 18/25]  Current/Best:   18.74/  19.58 GFLOPS | Progress: (12/20) | 13.30 s
    [Task 18/25]  Current/Best:    6.78/  19.58 GFLOPS | Progress: (16/20) | 15.87 s
    [Task 18/25]  Current/Best:   18.74/  19.58 GFLOPS | Progress: (20/20) | 17.58 s Done.
+
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 19/25]  Current/Best:    8.07/  11.37 GFLOPS | Progress: (4/20) | 6.79 s
    [Task 19/25]  Current/Best:   10.20/  20.44 GFLOPS | Progress: (8/20) | 10.60 s
    [Task 19/25]  Current/Best:   18.01/  20.44 GFLOPS | Progress: (12/20) | 12.45 s
    [Task 19/25]  Current/Best:   11.74/  20.44 GFLOPS | Progress: (16/20) | 14.92 s
    [Task 19/25]  Current/Best:   10.51/  20.77 GFLOPS | Progress: (20/20) | 17.22 s Done.
+
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 20/25]  Current/Best:    9.04/  11.56 GFLOPS | Progress: (4/20) | 4.28 s
    [Task 20/25]  Current/Best:   12.56/  14.58 GFLOPS | Progress: (8/20) | 6.25 s
    [Task 20/25]  Current/Best:   10.37/  16.45 GFLOPS | Progress: (12/20) | 8.66 s
    [Task 20/25]  Current/Best:   11.70/  16.45 GFLOPS | Progress: (16/20) | 13.52 s
    [Task 20/25]  Current/Best:   13.12/  18.40 GFLOPS | Progress: (20/20) | 15.26 s
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 21/25]  Current/Best:    8.91/  10.50 GFLOPS | Progress: (4/20) | 4.49 s
    [Task 21/25]  Current/Best:    4.67/  10.50 GFLOPS | Progress: (8/20) | 8.14 s
    [Task 21/25]  Current/Best:    1.61/  12.45 GFLOPS | Progress: (12/20) | 10.58 s Done.
+
    [Task 21/25]  Current/Best:   13.02/  17.12 GFLOPS | Progress: (16/20) | 13.00 s
    [Task 21/25]  Current/Best:    6.29/  17.12 GFLOPS | Progress: (20/20) | 20.55 s Done.
+
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 22/25]  Current/Best:    5.26/  17.77 GFLOPS | Progress: (4/20) | 3.20 s
    [Task 22/25]  Current/Best:    6.13/  18.76 GFLOPS | Progress: (8/20) | 4.59 s
    [Task 22/25]  Current/Best:    8.24/  18.76 GFLOPS | Progress: (12/20) | 6.36 s
    [Task 22/25]  Current/Best:   16.09/  18.76 GFLOPS | Progress: (16/20) | 9.50 s
    [Task 22/25]  Current/Best:   17.83/  18.76 GFLOPS | Progress: (20/20) | 11.69 s Done.
+
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 23/25]  Current/Best:   17.31/  18.32 GFLOPS | Progress: (4/20) | 3.84 s
    [Task 23/25]  Current/Best:   16.90/  19.20 GFLOPS | Progress: (8/20) | 6.36 s
    [Task 23/25]  Current/Best:   11.95/  21.37 GFLOPS | Progress: (12/20) | 8.17 s
    [Task 23/25]  Current/Best:   17.97/  21.37 GFLOPS | Progress: (16/20) | 10.44 s
    [Task 23/25]  Current/Best:   10.59/  21.37 GFLOPS | Progress: (20/20) | 12.97 s Done.
+
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 24/25]  Current/Best:    7.16/   7.16 GFLOPS | Progress: (4/20) | 12.31 s
    [Task 24/25]  Current/Best:    0.00/   7.16 GFLOPS | Progress: (8/20) | 23.54 s
    [Task 24/25]  Current/Best:    6.13/   7.16 GFLOPS | Progress: (12/20) | 35.27 s
    [Task 24/25]  Current/Best:    1.97/   9.61 GFLOPS | Progress: (16/20) | 37.28 s
    [Task 24/25]  Current/Best:    2.22/   9.61 GFLOPS | Progress: (20/20) | 48.07 s
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
+
    [Task 25/25]  Current/Best:    2.96/   5.60 GFLOPS | Progress: (4/20) | 12.33 s
    [Task 25/25]  Current/Best:    5.63/   7.75 GFLOPS | Progress: (8/20) | 24.04 s
    [Task 25/25]  Current/Best:    5.47/   7.98 GFLOPS | Progress: (12/20) | 34.17 s
    [Task 25/25]  Current/Best:    8.73/   8.73 GFLOPS | Progress: (16/20) | 44.66 s
    [Task 25/25]  Current/Best:    3.44/   8.73 GFLOPS | Progress: (20/20) | 55.45 s
 
 
 
@@ -733,8 +731,8 @@ improvement in comparing the optimized model to the unoptimized model.
 
  .. code-block:: none
 
-    optimized: {'mean': 429.7182228099973, 'median': 431.4217151000321, 'std': 4.16696520745253}
-    unoptimized: {'mean': 521.135307909999, 'median': 521.1086226000134, 'std': 2.045413370611934}
+    optimized: {'mean': 419.3073919599965, 'median': 419.3220671000063, 'std': 0.8234036066356031}
+    unoptimized: {'mean': 520.7545591599956, 'median': 520.3844177999827, 'std': 2.606245863239369}
 
 
 
@@ -757,7 +755,7 @@ profiling/benchmarking.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 10 minutes  33.992 seconds)
+   **Total running time of the script:** ( 11 minutes  37.977 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 3997ce2d64..e663074443 100644
--- a/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
+++ b/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
@@ -270,7 +270,7 @@ device and returns the measured cost. Network overhead is excluded.
 
  .. code-block:: none
 
-    1.264e-07 secs/op
+    1.265e-07 secs/op
 
 
 
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index d5dc89805f..a37b66f0b3 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -260,7 +260,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
 
  .. code-block:: none
 
-    [stage(a, placeholder(a, 0x20961e90)), stage(b, placeholder(b, 0x21e64170)), 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, 0x225864f0)), stage(b, placeholder(b, 0x13397270)), 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 8e97fff97f..c7e58f01d3 100644
--- a/docs/_sources/tutorial/sg_execution_times.rst.txt
+++ b/docs/_sources/tutorial/sg_execution_times.rst.txt
@@ -5,32 +5,32 @@
 
 Computation times
 =================
-**14:00.891** total execution time for **tutorial** files:
+**15:38.537** total execution time for **tutorial** files:
 
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)                 | 10:33.992 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)                 | 11:37.977 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:21.583 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:49.022 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)     | 01:00.939 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)     | 01:02.200 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)                 | 00:36.630 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)                 | 00:36.681 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)               | 00:26.073 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)               | 00:31.065 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)                               | 00:00.799 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)                               | 00:00.782 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)       | 00:00.668 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)       | 00:00.643 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.194 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.159 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)                           | 00:00.007 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)                           | 00:00.005 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_uma.py` (``uma.py``)                                             | 00:00.002 | 0.0 MB |
-+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)   | 00:00.001 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_uma.py` (``uma.py``)                                             | 00:00.001 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)                             | 00:00.001 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)   | 00:00.001 | 0.0 MB |
++------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_install.py` (``install.py``)                                     | 00:00.001 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
index dd0da46ae0..5912551596 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -294,8 +294,8 @@ helper function to run a profile of the TVM generated code.
 
  .. code-block:: none
 
-    Numpy running time: 0.000008
-    naive: 0.000010
+    Numpy running time: 0.000007
+    naive: 0.000007
 
 
 
@@ -393,7 +393,7 @@ compile and run this new schedule with the parallel operation applied:
 
  .. code-block:: none
 
-    parallel: 0.000009
+    parallel: 0.000008
 
 
 
@@ -499,10 +499,10 @@ We can now compare the different schedules
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                   numpy    7.852259996070643e-06                    1.0
-                   naive              9.7594e-06       1.242877847254639
-                parallel              9.3415e-06      1.1896575004743335
-                  vector    2.4529200000000002e-05    3.1238395076417085
+                   numpy    6.7453399969963355e-06                   1.0
+                   naive    6.6364000000000005e-06     0.983849591414985
+                parallel              8.1158e-06      1.2031713751440138
+                  vector             2.45018e-05       3.632404002008872
 
 
 
@@ -923,7 +923,7 @@ matrix multiplication.
 
  .. code-block:: none
 
-    Numpy running time: 0.019926
+    Numpy running time: 0.019089
 
 
 
@@ -981,7 +981,7 @@ optimizations.
 
  .. code-block:: none
 
-    none: 3.330825
+    none: 3.445382
 
 
 
@@ -1083,7 +1083,7 @@ schedule.
 
  .. code-block:: none
 
-    blocking: 0.330132
+    blocking: 0.325926
 
 
 
@@ -1178,7 +1178,7 @@ already cache friendly from our previous optimizations.
 
  .. code-block:: none
 
-    vectorization: 0.363457
+    vectorization: 0.357039
     @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, [1024, 1024], []),
@@ -1251,7 +1251,7 @@ more cache friendly.
 
  .. code-block:: none
 
-    loop permutation: 0.133229
+    loop permutation: 0.123395
     @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, [1024, 1024], []),
@@ -1349,7 +1349,7 @@ optimized schedule.
 
  .. code-block:: none
 
-    array packing: 0.107371
+    array packing: 0.110126
     @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, [1024, 1024], []),
@@ -1441,7 +1441,7 @@ to `C` when all the block results are ready.
 
  .. code-block:: none
 
-    block caching: 0.111066
+    block caching: 0.111337
     @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, [1024, 1024], []),
@@ -1526,7 +1526,7 @@ of thread-level parallelization.
 
  .. code-block:: none
 
-    parallelization: 0.146867
+    parallelization: 0.146905
     @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, [1024, 1024], []),
@@ -1606,13 +1606,13 @@ working, we can compare the results.
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                    none            3.3308246923                     1.0
-                blocking             0.330131613     0.09911407639170515
-           vectorization            0.3634572827     0.10911930716144223
-        loop permutation            0.1332293653     0.03999891246392872
-           array packing            0.1073706838     0.03223546530329654
-           block caching     0.11106572819999998    0.033344813510226175
-         parallelization            0.1468670843     0.04409330957570913
+                    none            3.4453823653                     1.0
+                blocking            0.3259262646     0.09459799524213929
+           vectorization            0.3570393058     0.10362835469174739
+        loop permutation            0.1233947181     0.03581452071699323
+           array packing            0.1101264246     0.03196348414304691
+           block caching     0.11133696910000002    0.032314836873063746
+         parallelization            0.1469047951    0.042638168866116134
 
 
 
@@ -1654,7 +1654,7 @@ the computation for specific platforms.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  0.939 seconds)
+   **Total running time of the script:** ( 1 minutes  2.200 seconds)
 
 
 .. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
diff --git a/docs/commit_hash b/docs/commit_hash
index 78287c1c83..c2d4e2aa43 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-52739ef8cd4499b555453d7159991316afc44301
+271ad4302917011a54b257ca2a78c563a7ba652c
diff --git a/docs/how_to/compile_models/from_darknet.html b/docs/how_to/compile_models/from_darknet.html
index 1e3e30ba57..25dedb5007 100644
--- a/docs/how_to/compile_models/from_darknet.html
+++ b/docs/how_to/compile_models/from_darknet.html
@@ -585,7 +585,7 @@ class:[&#39;truck 0.9266&#39;] left:471 top:83 right:689 bottom:169
 class:[&#39;bicycle 0.9984&#39;] left:111 top:113 right:577 bottom:447
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  16.228 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  15.285 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-darknet-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/7716f96385bd5abb6e822041e285be54/from_darknet.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_darknet.py</span></code></a></p>
diff --git a/docs/how_to/compile_models/from_keras.html b/docs/how_to/compile_models/from_keras.html
index 6852fe7b67..71f55e8c3e 100644
--- a/docs/how_to/compile_models/from_keras.html
+++ b/docs/how_to/compile_models/from_keras.html
@@ -506,7 +506,7 @@ pip install -U tensorflow --user
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Relay top-1 id: 285, class name: Egyptian cat
 
 1/1 [==============================] - ETA: 0s
-1/1 [==============================] - 1s 955ms/step
+1/1 [==============================] - 1s 1s/step
 Keras top-1 id: 285, class name: Egyptian cat
 </pre></div>
 </div>
diff --git a/docs/how_to/compile_models/from_mxnet.html b/docs/how_to/compile_models/from_mxnet.html
index b2eade0ba9..96c7c996fe 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -440,7 +440,7 @@ to download the full example code</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;x&quot;</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#tuple" title="builtins.tuple" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">x</span><span class="o">.</span><span class="n">shape</span></a><span class="p">)</span>
 </pre></div>
 </div>
-<img src="../../_images/sphx_glr_from_mxnet_001.png" srcset="../../_images/sphx_glr_from_mxnet_001.png" alt="from mxnet" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip2b781813-d81c-436a-a775-b37a76c5a5e2 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+<img src="../../_images/sphx_glr_from_mxnet_001.png" srcset="../../_images/sphx_glr_from_mxnet_001.png" alt="from mxnet" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip9b8baf24-50e2-4883-a416-0d0ac295aea7 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_oneflow.html b/docs/how_to/compile_models/from_oneflow.html
index 0b7745e302..d394f1212d 100644
--- a/docs/how_to/compile_models/from_oneflow.html
+++ b/docs/how_to/compile_models/from_oneflow.html
@@ -448,15 +448,15 @@ Deprecated in NumPy 1.20; for more details and guidance: https://numpy.org/devdo
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading: &quot;https://oneflow-public.oss-cn-beijing.aliyuncs.com/model_zoo/flowvision/classification/ResNet/resnet18.zip&quot; to /workspace/.oneflow/flowvision_cache/resnet18.zip
 
   0%|          | 0.00/41.5M [00:00&lt;?, ?B/s]
- 15%|#5        | 6.33M/41.5M [00:00&lt;00:01, 31.4MB/s]
- 27%|##6       | 11.0M/41.5M [00:00&lt;00:00, 38.6MB/s]
- 39%|###8      | 16.0M/41.5M [00:00&lt;00:00, 33.8MB/s]
- 54%|#####3    | 22.3M/41.5M [00:00&lt;00:00, 33.7MB/s]
- 62%|######1   | 25.7M/41.5M [00:00&lt;00:00, 31.0MB/s]
- 77%|#######7  | 32.0M/41.5M [00:00&lt;00:00, 35.2MB/s]
- 85%|########5 | 35.4M/41.5M [00:01&lt;00:00, 33.5MB/s]
- 93%|#########3| 38.6M/41.5M [00:01&lt;00:00, 32.8MB/s]
-100%|##########| 41.5M/41.5M [00:01&lt;00:00, 34.3MB/s]
+ 15%|#5        | 6.33M/41.5M [00:00&lt;00:00, 47.1MB/s]
+ 27%|##7       | 11.2M/41.5M [00:00&lt;00:00, 49.2MB/s]
+ 39%|###8      | 16.0M/41.5M [00:00&lt;00:00, 37.0MB/s]
+ 48%|####7     | 19.8M/41.5M [00:00&lt;00:00, 33.6MB/s]
+ 57%|#####7    | 23.7M/41.5M [00:00&lt;00:00, 32.9MB/s]
+ 65%|######4   | 26.9M/41.5M [00:00&lt;00:00, 30.2MB/s]
+ 82%|########2 | 34.1M/41.5M [00:00&lt;00:00, 39.4MB/s]
+ 96%|#########6| 40.0M/41.5M [00:01&lt;00:00, 38.6MB/s]
+100%|##########| 41.5M/41.5M [00:01&lt;00:00, 38.5MB/s]
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/compile_models/from_pytorch.html b/docs/how_to/compile_models/from_pytorch.html
index 29671fb9f3..ed113dddf7 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -431,12 +431,10 @@ be unstable.</p>
 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]
- 18%|#7        | 7.99M/44.7M [00:00&lt;00:00, 68.8MB/s]
- 36%|###5      | 16.0M/44.7M [00:00&lt;00:00, 55.9MB/s]
- 54%|#####3    | 24.0M/44.7M [00:00&lt;00:00, 62.2MB/s]
- 72%|#######1  | 32.0M/44.7M [00:00&lt;00:00, 66.3MB/s]
- 90%|########9 | 40.1M/44.7M [00:00&lt;00:00, 70.1MB/s]
-100%|##########| 44.7M/44.7M [00:00&lt;00:00, 71.9MB/s]
+ 32%|###1      | 14.1M/44.7M [00:00&lt;00:00, 148MB/s]
+ 63%|######3   | 28.2M/44.7M [00:00&lt;00:00, 89.5MB/s]
+ 98%|#########8| 43.9M/44.7M [00:00&lt;00:00, 114MB/s]
+100%|##########| 44.7M/44.7M [00:00&lt;00:00, 111MB/s]
 </pre></div>
 </div>
 </div>
diff --git a/docs/how_to/compile_models/from_tensorflow.html b/docs/how_to/compile_models/from_tensorflow.html
index 3977441680..5e95529952 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -645,7 +645,7 @@ banana (score = 0.00022)
 desk (score = 0.00019)
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  13.077 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  13.573 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-tensorflow-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/7f1d3d1b878694c201c614c807cdebc8/from_tensorflow.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_tensorflow.py</span></code></a></p>
diff --git a/docs/how_to/compile_models/sg_execution_times.html b/docs/how_to/compile_models/sg_execution_times.html
index 77e074abe4..20784ebe67 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>05:58.074</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>06:00.467</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 81%" />
@@ -349,43 +349,43 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>01:16.228</p></td>
+<td><p>01:15.285</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>01:13.077</p></td>
+<td><p>01:13.573</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:47.284</p></td>
+<td><p>00:48.006</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="from_oneflow.html#sphx-glr-how-to-compile-models-from-oneflow-py"><span class="std std-ref">Compile OneFlow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_oneflow.py</span></code>)</p></td>
-<td><p>00:33.815</p></td>
+<td><p>00:34.751</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:29.966</p></td>
+<td><p>00:30.889</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:27.561</p></td>
+<td><p>00:27.578</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:26.150</p></td>
+<td><p>00:25.668</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:23.095</p></td>
+<td><p>00:23.593</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:18.467</p></td>
+<td><p>00:18.636</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:02.433</p></td>
+<td><p>00:02.487</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
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 b823d3f3e9..efe44471a4 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -661,7 +661,7 @@ to the remote android device.</p>
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  16.1529      16.1286      16.3120      16.0284       0.0906
+  16.9160      16.8723      17.7639      16.3612       0.4699
 </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 54bddae2f0..c5d36d2a61 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -453,32 +453,22 @@ be unstable.</p>
 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]
-  5%|4         | 7.99M/170M [00:00&lt;00:04, 39.9MB/s]
-  8%|8         | 14.3M/170M [00:00&lt;00:03, 40.8MB/s]
- 11%|#         | 18.2M/170M [00:00&lt;00:04, 36.0MB/s]
- 14%|#4        | 24.0M/170M [00:00&lt;00:04, 36.7MB/s]
- 19%|#8        | 32.0M/170M [00:00&lt;00:03, 38.2MB/s]
- 24%|##3       | 40.0M/170M [00:01&lt;00:03, 43.5MB/s]
- 28%|##8       | 48.0M/170M [00:01&lt;00:02, 43.9MB/s]
- 33%|###2      | 56.0M/170M [00:01&lt;00:02, 48.5MB/s]
- 37%|###6      | 62.3M/170M [00:01&lt;00:02, 52.2MB/s]
- 40%|###9      | 67.5M/170M [00:01&lt;00:02, 47.1MB/s]
- 42%|####2     | 72.2M/170M [00:01&lt;00:02, 41.2MB/s]
- 47%|####6     | 79.7M/170M [00:01&lt;00:01, 49.6MB/s]
- 50%|####9     | 84.8M/170M [00:02&lt;00:01, 45.0MB/s]
- 53%|#####2    | 89.5M/170M [00:02&lt;00:02, 33.8MB/s]
- 56%|#####6    | 95.5M/170M [00:02&lt;00:01, 39.6MB/s]
- 59%|#####8    | 99.9M/170M [00:02&lt;00:02, 35.7MB/s]
- 61%|######1   | 104M/170M [00:02&lt;00:02, 31.7MB/s]
- 66%|######5   | 112M/170M [00:02&lt;00:01, 35.3MB/s]
- 71%|#######   | 120M/170M [00:03&lt;00:01, 43.0MB/s]
- 75%|#######5  | 128M/170M [00:03&lt;00:00, 44.1MB/s]
- 80%|########  | 136M/170M [00:03&lt;00:00, 46.6MB/s]
- 85%|########4 | 144M/170M [00:03&lt;00:00, 45.1MB/s]
- 89%|########9 | 152M/170M [00:03&lt;00:00, 45.9MB/s]
- 93%|#########3| 158M/170M [00:03&lt;00:00, 45.5MB/s]
- 96%|#########5| 163M/170M [00:04&lt;00:00, 41.6MB/s]
-100%|##########| 170M/170M [00:04&lt;00:00, 42.9MB/s]
+  5%|4         | 7.99M/170M [00:00&lt;00:02, 74.6MB/s]
+ 14%|#4        | 24.0M/170M [00:00&lt;00:01, 124MB/s]
+ 21%|##1       | 36.0M/170M [00:00&lt;00:01, 120MB/s]
+ 28%|##7       | 47.5M/170M [00:00&lt;00:01, 111MB/s]
+ 34%|###4      | 58.2M/170M [00:00&lt;00:01, 102MB/s]
+ 41%|####      | 69.4M/170M [00:00&lt;00:00, 106MB/s]
+ 47%|####6     | 79.6M/170M [00:00&lt;00:00, 104MB/s]
+ 53%|#####2    | 89.7M/170M [00:00&lt;00:00, 103MB/s]
+ 59%|#####8    | 99.6M/170M [00:01&lt;00:00, 92.2MB/s]
+ 66%|######6   | 112M/170M [00:01&lt;00:00, 104MB/s]
+ 72%|#######2  | 123M/170M [00:01&lt;00:00, 103MB/s]
+ 78%|#######8  | 133M/170M [00:01&lt;00:00, 102MB/s]
+ 84%|########3 | 143M/170M [00:01&lt;00:00, 102MB/s]
+ 90%|########9 | 152M/170M [00:01&lt;00:00, 94.7MB/s]
+ 97%|#########6| 164M/170M [00:01&lt;00:00, 103MB/s]
+100%|##########| 170M/170M [00:01&lt;00:00, 102MB/s]
 /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torch/nn/functional.py:3897: 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)
 /venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/detection/anchor_utils.py:124: 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=& [...]
@@ -576,7 +566,7 @@ 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> ( 3 minutes  24.737 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes  24.066 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-object-detection-pytorch-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 089b356433..bb24f6868a 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -497,7 +497,8 @@ training. Other models require a full post training calibration.</p>
 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, 159MB/s]
+ 61%|######    | 8.21M/13.6M [00:00&lt;00:00, 86.1MB/s]
+100%|##########| 13.6M/13.6M [00:00&lt;00:00, 115MB/s]
 </pre></div>
 </div>
 </div>
@@ -588,7 +589,7 @@ output values are identical out of 1000 outputs from mobilenet v2.</p>
 </div>
 <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)
-  90.6324      90.5561      93.1118      90.2627       0.4082
+  90.4818      90.3775      93.0215      90.2388       0.3177
 </pre></div>
 </div>
 <div class="admonition note">
@@ -627,7 +628,7 @@ 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  8.512 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  8.426 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 4fd6a01496..31d41d2afc 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -582,7 +582,7 @@ TFLite Top-5 labels: [387 102 386 341 349]
 </div>
 <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)
-  121.4938     121.4546     125.6496     120.6399      0.5865
+  121.9242     121.9387     124.2872     120.9418      0.4936
 </pre></div>
 </div>
 <div class="admonition note">
@@ -610,7 +610,7 @@ 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> ( 2 minutes  28.305 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  27.895 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-tflite-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 f15f121577..574dfa156f 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -520,7 +520,7 @@ 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  42.212 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  43.484 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-quantized-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 c158302807..f9c366cc7a 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -462,23 +462,23 @@ 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]
-  5%|4         | 6163/132723 [00:00&lt;00:02, 61623.30KB/s]
- 11%|#1        | 14890/132723 [00:00&lt;00:01, 76706.81KB/s]
- 17%|#6        | 22561/132723 [00:00&lt;00:01, 61488.26KB/s]
- 24%|##3       | 31336/132723 [00:00&lt;00:01, 70502.66KB/s]
- 29%|##9       | 39069/132723 [00:00&lt;00:01, 72748.29KB/s]
- 36%|###6      | 47847/132723 [00:00&lt;00:01, 77552.97KB/s]
- 42%|####2     | 55779/132723 [00:00&lt;00:01, 60308.08KB/s]
- 49%|####8     | 64533/132723 [00:00&lt;00:01, 67303.49KB/s]
- 54%|#####4    | 71883/132723 [00:01&lt;00:00, 68966.09KB/s]
- 61%|######    | 80613/132723 [00:01&lt;00:00, 74045.10KB/s]
- 67%|######7   | 89317/132723 [00:01&lt;00:00, 77730.07KB/s]
- 74%|#######3  | 97956/132723 [00:01&lt;00:00, 80224.79KB/s]
- 80%|########  | 106638/132723 [00:01&lt;00:00, 82146.77KB/s]
- 87%|########6 | 115007/132723 [00:01&lt;00:00, 66534.79KB/s]
- 93%|#########3| 123756/132723 [00:01&lt;00:00, 71833.44KB/s]
- 99%|#########9| 131450/132723 [00:01&lt;00:00, 73184.77KB/s]
-100%|##########| 132723/132723 [00:01&lt;00:00, 71855.54KB/s]
+  4%|4         | 5804/132723 [00:00&lt;00:02, 58030.23KB/s]
+ 10%|9         | 13109/132723 [00:00&lt;00:01, 66862.10KB/s]
+ 15%|#4        | 19796/132723 [00:00&lt;00:01, 64361.03KB/s]
+ 21%|##1       | 28369/132723 [00:00&lt;00:01, 72601.06KB/s]
+ 28%|##7       | 36965/132723 [00:00&lt;00:01, 77342.24KB/s]
+ 34%|###4      | 45592/132723 [00:00&lt;00:01, 80346.29KB/s]
+ 41%|####      | 54218/132723 [00:00&lt;00:00, 82264.53KB/s]
+ 47%|####7     | 62867/132723 [00:00&lt;00:00, 83602.05KB/s]
+ 54%|#####3    | 71483/132723 [00:00&lt;00:00, 84388.91KB/s]
+ 60%|######    | 80103/132723 [00:01&lt;00:00, 84945.16KB/s]
+ 67%|######6   | 88603/132723 [00:01&lt;00:00, 75283.38KB/s]
+ 73%|#######3  | 97213/132723 [00:01&lt;00:00, 78298.82KB/s]
+ 79%|#######9  | 105211/132723 [00:01&lt;00:00, 73994.42KB/s]
+ 86%|########5 | 113804/132723 [00:01&lt;00:00, 77293.46KB/s]
+ 92%|#########1| 121669/132723 [00:01&lt;00:00, 69239.69KB/s]
+ 98%|#########8| 130213/132723 [00:01&lt;00:00, 73531.33KB/s]
+100%|##########| 132723/132723 [00:01&lt;00:00, 75954.56KB/s]
 </pre></div>
 </div>
 <p>Create TVM runtime and do inference
@@ -517,7 +517,7 @@ Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from h
 <span class="n">plt</span><span class="o">.</span><span class="n">show</span><span class="p">()</span>
 </pre></div>
 </div>
-<img src="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" srcset="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" alt="deploy ssd gluoncv" class = "sphx-glr-single-img"/><p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes  7.321 seconds)</p>
+<img src="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" srcset="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" alt="deploy ssd gluoncv" class = "sphx-glr-single-img"/><p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes  8.938 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-ssd-gluoncv-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 67e6391281..80fca9545e 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>13:21.255</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>13:23.321</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 86%" />
@@ -349,35 +349,35 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>03:24.737</p></td>
+<td><p>03:24.066</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>03:07.321</p></td>
+<td><p>03:08.938</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>02:28.305</p></td>
+<td><p>02:27.895</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>01:42.212</p></td>
+<td><p>01:43.484</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>01:08.512</p></td>
+<td><p>01:08.426</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:37.861</p></td>
+<td><p>00:37.846</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><a class="reference internal" href="deploy_model_on_nano.html#sphx-glr-how-to-deploy-models-deploy-model-on-nano-py"><span class="std std-ref">Deploy the Pretrained Model on Jetson Nano</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_nano.py</span></code>)</p></td>
-<td><p>00:26.511</p></td>
+<td><p>00:26.691</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:25.790</p></td>
+<td><p>00:25.969</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
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 b957d82b27..f55d27dd36 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -621,7 +621,7 @@ In this alpha state of the Bring Your Own Datatypes framework, we have not imple
 <span class="n">module</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a> <span class="o">=</span> <span class="n">get_mobilenet</span><span class="p">()</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip85164043-6ab1-4ce7-9f25-a44176bc77c5 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.zip23873bed-17bd-4043-842f-0f79495e7b5f 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 47425a31d8..686e682116 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>00:49.063</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:49.942</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -349,15 +349,15 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:45.485</p></td>
+<td><p>00:46.316</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:02.505</p></td>
+<td><p>00:02.538</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:01.065</p></td>
+<td><p>00:01.079</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index 4b42d9094f..9326145970 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -525,10 +525,10 @@ profile the execution time of each passes.</p>
 </pre></div>
 </div>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 7311us [7311us] (46.27%; 46.27%)
-FoldScaleAxis: 8490us [8us] (53.73%; 53.73%)
-        FoldConstant: 8482us [1698us] (53.68%; 99.90%)
-                InferType: 6783us [6783us] (42.93%; 79.98%)
+InferType: 7596us [7596us] (46.51%; 46.51%)
+FoldScaleAxis: 8737us [8us] (53.49%; 53.49%)
+        FoldConstant: 8729us [1788us] (53.44%; 99.91%)
+                InferType: 6941us [6941us] (42.50%; 79.52%)
 </pre></div>
 </div>
 </div>
@@ -550,10 +550,10 @@ Refer to following sections and <a class="reference internal" href="../../refere
 </pre></div>
 </div>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 6902us [6902us] (44.90%; 44.90%)
-FoldScaleAxis: 8470us [6us] (55.10%; 55.10%)
-        FoldConstant: 8464us [1732us] (55.06%; 99.93%)
-                InferType: 6732us [6732us] (43.79%; 79.53%)
+InferType: 7105us [7105us] (45.22%; 45.22%)
+FoldScaleAxis: 8607us [6us] (54.78%; 54.78%)
+        FoldConstant: 8600us [1749us] (54.74%; 99.92%)
+                InferType: 6851us [6851us] (43.61%; 79.67%)
 </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 f4c67ed43e..bab4ef0aad 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -577,7 +577,7 @@ latency of convolution.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Convolution: </span><span class="si">%f</span><span class="s2"> ms&quot;</span> <span class="o">%</span> <span class="p">(</span><span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">w</span><span class="p">,</span> <span class="n">b</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span> <span class="o">*</span> <span cl [...]
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.245216 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.114494 ms
 </pre></div>
 </div>
 <div class="sphx-glr-footer 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 3082283a9f..14bbad5579 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -914,7 +914,7 @@ be able to run on our build server</p>
     <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;conv2d with tensor core: </span><span class="si">%f</span><span class="s2"> ms&quot;</span> <span class="o">%</span> <span class="p">(</span><span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">w</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span> <span class="o">* [...]
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 13.002118 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 13.367216 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 588d1d4676..606379323d 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -474,8 +474,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Baseline: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019491
-Baseline: 3.290225
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019649
+Baseline: 3.486395
 </pre></div>
 </div>
 <p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -534,7 +534,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt1: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.328238
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.329770
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -600,7 +600,7 @@ vastly.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt2: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.353639
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.349371
 </pre></div>
 </div>
 <p>Here is the generated IR after vectorization.</p>
@@ -660,7 +660,7 @@ the access pattern for A matrix is more cache friendly.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt3: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.122369
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.120619
 </pre></div>
 </div>
 <p>Here is the generated IR after loop permutation.</p>
@@ -827,7 +827,7 @@ write to C when all the block results are ready.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt5: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">evaluator</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="n">b</span><span class="p">,</span> <span class="n">c</span><span class="p">)</span><span class="o">.</span><span class="n">mean</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.111313
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.112183
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -916,7 +916,7 @@ write to C when all the block results are ready.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt6: </span><span class="si">%f</span><span class="s2">&quot;</span> <span class="o">%</span> <span class="n">opt6_time</span><span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.147555
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.147467
 </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 d32706e36a..26c9e8c2bc 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>00:35.420</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:36.049</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 83%" />
@@ -349,15 +349,15 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:32.731</p></td>
+<td><p>00:33.260</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:01.512</p></td>
+<td><p>00:01.554</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:01.176</p></td>
+<td><p>00:01.235</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
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 99d6075a32..30bef2974a 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>09:16.506</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>09:21.677</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 85%" />
@@ -349,27 +349,27 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>05:43.754</p></td>
+<td><p>05:48.786</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>01:34.813</p></td>
+<td><p>01:35.031</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>01:04.752</p></td>
+<td><p>01:04.757</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:28.883</p></td>
+<td><p>00:29.200</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:12.495</p></td>
+<td><p>00:12.289</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:11.809</p></td>
+<td><p>00:11.614</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
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 8df73536b8..53ba5f5aa0 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
@@ -488,6 +488,9 @@ file and apply it.</p>
 <span class="k">del</span> <span class="n">measure_ctx</span>
 </pre></div>
 </div>
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>.T
+</pre></div>
+</div>
 <p>We can lower the schedule to see the IR after auto-scheduling.
 The auto-scheduler correctly performs optimizations including multi-level tiling,
 cooperative fetching, unrolling and operator fusion.</p>
@@ -503,952 +506,120 @@ cooperative fetching, unrolling and operator fusion.</p>
              bias: Buffer(bias_2: Pointer(float32), float32, [1, 512, 1, 1], []),
              compute: Buffer(compute_2: Pointer(float32), float32, [1, 512, 7, 7], [])}
   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, [16]), storage_scope = local;
-  allocate(pad_temp.shared: Pointer(shared float32), float32, [784]), storage_scope = shared;
-  allocate(kernel.shared: Pointer(shared float32), float32, [256]), storage_scope = shared;
-  attr [IterVar(threadIdx.x: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49 {
-    conv2d_nchw_1: Buffer(conv2d_nchw, float32, [4], [], scope=&quot;local&quot;, align=8)[0] = 0f32
-    conv2d_nchw_1[2] = 0f32
-    conv2d_nchw_1[4] = 0f32
-    conv2d_nchw_1[6] = 0f32
-    conv2d_nchw_1[8] = 0f32
-    conv2d_nchw_1[10] = 0f32
-    conv2d_nchw_1[12] = 0f32
-    conv2d_nchw_1[14] = 0f32
-    conv2d_nchw_1[1] = 0f32
-    conv2d_nchw_1[3] = 0f32
-    conv2d_nchw_1[5] = 0f32
+  attr [IterVar(blockIdx.x: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;blockIdx.x&quot;)] &quot;thread_extent&quot; = 8;
+  allocate(conv2d_nchw: Pointer(local float32), float32, [14]), storage_scope = local;
+  allocate(pad_temp.shared: Pointer(shared float32), float32, [1008]), storage_scope = shared;
+  allocate(kernel.shared: Pointer(shared float32), float32, [3072]), storage_scope = shared;
+  attr [IterVar(threadIdx.x: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224 {
+    conv2d_nchw_1: Buffer(conv2d_nchw, float32, [14], [], scope=&quot;local&quot;, align=32)[0] = 0f32
     conv2d_nchw_1[7] = 0f32
+    conv2d_nchw_1[1] = 0f32
+    conv2d_nchw_1[8] = 0f32
+    conv2d_nchw_1[2] = 0f32
     conv2d_nchw_1[9] = 0f32
+    conv2d_nchw_1[3] = 0f32
+    conv2d_nchw_1[10] = 0f32
+    conv2d_nchw_1[4] = 0f32
     conv2d_nchw_1[11] = 0f32
+    conv2d_nchw_1[5] = 0f32
+    conv2d_nchw_1[12] = 0f32
+    conv2d_nchw_1[6] = 0f32
     conv2d_nchw_1[13] = 0f32
-    conv2d_nchw_1[15] = 0f32
     for (rc.outer.outer: int32, 0, 32) {
-      for (ry.outer.outer: int32, 0, 3) {
-        attr [IterVar(threadIdx.x_1: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1: Buffer(pad_temp.shared, float32, [784], [], scope=&quot;shared&quot;)[threadIdx.x_1] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3: Buffer(data_2, float32, [25088], [])[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 8)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 41)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 90)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 139)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 188)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 237)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 286)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 335)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 384)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 433)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 482)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 531)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 580)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 629)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 678)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (1 &lt;= floormod(threadIdx.x_1, 7))), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 727)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_2: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49 {
-          kernel.shared_1: Buffer(kernel.shared, float32, [256], [], scope=&quot;shared&quot;)[(threadIdx.x_2*4)] = kernel_3: Buffer(kernel_2, float32, [2359296], [])[(((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3))]
-          kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 9)]
-          kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 18)]
-          kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 27)]
-        }
-        attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        if @tir.likely((threadIdx.x_2 &lt; 15), dtype=bool) {
-          let cse_var_1: int32 = (ry.outer.outer*3)
-           {
-            kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[(((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_1)]
+      for (rx.outer.outer: int32, 0, 3) {
+        let cse_var_1: int32 = (rc.outer.outer*144)
+         {
+          attr [IterVar(threadIdx.x_1: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          if @tir.likely((threadIdx.x_1 &lt; 144), dtype=bool) {
+            pad_temp.shared_1: Buffer(pad_temp.shared, float32, [1008], [], scope=&quot;shared&quot;)[(threadIdx.x_1*7)] = @tir.if_then_else((((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)) &amp;&amp; (1 &lt;= rx.outer.outer)), data_3: Buffer(data_2, float32, [25088], [])[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 8)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 1)] = @tir.if_then_else(((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 7)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 2)] = @tir.if_then_else(((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 6)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 3)] = @tir.if_then_else(((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 5)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 4)] = @tir.if_then_else(((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 4)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 5)] = @tir.if_then_else(((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 3)], 0f32, dtype=float32)
+            pad_temp.shared_1[((threadIdx.x_1*7) + 6)] = @tir.if_then_else((((1 &lt;= floormod(threadIdx.x_1, 9)) &amp;&amp; (floormod(threadIdx.x_1, 9) &lt; 8)) &amp;&amp; (rx.outer.outer &lt; 2)), data_3[(((((rc.outer.outer*784) + (floordiv(threadIdx.x_1, 9)*49)) + (floormod(threadIdx.x_1, 9)*7)) + rx.outer.outer) - 2)], 0f32, dtype=float32)
           }
-        }
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[threadIdx.x_1] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 7)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 42)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 91)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 140)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 189)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 238)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 287)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 336)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 385)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 434)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 483)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 532)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 581)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 630)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 679)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else(((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 728)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49 {
-          kernel.shared_1[(threadIdx.x_2*4)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 1)]
-          kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 10)]
-          kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 19)]
-          kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 28)]
-        }
-        attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        if @tir.likely((threadIdx.x_2 &lt; 15), dtype=bool) {
-          let cse_var_2: int32 = (ry.outer.outer*3)
-           {
-            kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_2) + 1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_2) + 1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_2) + 1)]
-            kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_2) + 1)]
+          attr [IterVar(threadIdx.x_2: int32, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1: Buffer(kernel.shared, float32, [3072], [], scope=&quot;shared&quot;)[threadIdx.x_2] = kernel_3: Buffer(kernel_2, float32, [2359296], [])[(((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 224)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 224), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 448)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 448), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 672)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 64512)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 896)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 896), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 1120)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1120), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 1344)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 129024)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 1568)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1568), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 1792)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 1792), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 2016)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 193536)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 2240)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2240), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 2464)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2464), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 16), 48), 3)*9)) + (floormod((threadIdx.x_2 + 1), 3)*3)) + rx.outer.outer)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          kernel.shared_1[(threadIdx.x_2 + 2688)] = kernel_3[((((((blockIdx.x*294912) + (floordiv(threadIdx.x_2, 48)*4608)) + cse_var_1) + (floormod(threadIdx.x_2, 48)*3)) + rx.outer.outer) + 258048)]
+          attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 224;
+          if @tir.likely((threadIdx.x_2 &lt; 160), dtype=bool) {
+            kernel.shared_1[(threadIdx.x_2 + 2912)] = kernel_3[((((((blockIdx.x*294912) + (floordiv((threadIdx.x_2 + 2912), 48)*4608)) + cse_var_1) + (floordiv(floormod((threadIdx.x_2 + 32), 48), 3)*9)) + (floormod((threadIdx.x_2 + 2), 3)*3)) + rx.outer.outer)]
           }
-        }
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[threadIdx.x_1] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) - 6)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 49)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 43)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 98)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 92)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 147)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 141)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 196)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 190)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 245)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 239)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 294)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 288)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 343)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 337)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 392)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 386)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 441)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 435)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 490)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 484)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 539)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 533)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 588)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 582)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 637)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 631)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 686)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 680)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_1, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        pad_temp.shared_1[(threadIdx.x_1 + 735)] = @tir.if_then_else((((1 &lt;= (floordiv(threadIdx.x_1, 7) + ry.outer.outer)) &amp;&amp; ((floordiv(threadIdx.x_1, 7) + ry.outer.outer) &lt; 8)) &amp;&amp; (floormod(threadIdx.x_1, 7) &lt; 6)), data_3[((((rc.outer.outer*784) + (ry.outer.outer*7)) + threadIdx.x_1) + 729)], 0f32, dtype=float32)
-        attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49 {
-          kernel.shared_1[(threadIdx.x_2*4)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 2)]
-          kernel.shared_1[((threadIdx.x_2*4) + 1)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 11)]
-          kernel.shared_1[((threadIdx.x_2*4) + 2)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 20)]
-          kernel.shared_1[((threadIdx.x_2*4) + 3)] = kernel_3[((((((blockIdx.x*73728) + (floordiv(threadIdx.x_2, 4)*4608)) + (rc.outer.outer*144)) + (floormod(threadIdx.x_2, 4)*36)) + (ry.outer.outer*3)) + 29)]
-        }
-        attr [IterVar(threadIdx.x_2, (nullptr), &quot;ThreadIndex&quot;, &quot;threadIdx.x&quot;)] &quot;thread_extent&quot; = 49;
-        if @tir.likely((threadIdx.x_2 &lt; 15), dtype=bool) {
-          let cse_var_3: int32 = (ry.outer.outer*3)
-           {
-            kernel.shared_1[((threadIdx.x_2*4) + 196)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 4), 16)*9)) + cse_var_3) + 2)]
-            kernel.shared_1[((threadIdx.x_2*4) + 197)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 5), 16)*9)) + cse_var_3) + 2)]
-            kernel.shared_1[((threadIdx.x_2*4) + 198)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 6), 16)*9)) + cse_var_3) + 2)]
-            kernel.shared_1[((threadIdx.x_2*4) + 199)] = kernel_3[((((((blockIdx.x*73728) + (floordiv((threadIdx.x_2 + 49), 4)*4608)) + (rc.outer.outer*144)) + (floormod(((threadIdx.x_2*4) + 7), 16)*9)) + cse_var_3) + 2)]
+          for (rc.outer.inner: int32, 0, 16) {
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[((rc.outer.inner*63) + floormod(threadIdx.x, 7))]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[((rc.outer.inner*63) + floormod(threadIdx.x, 7))]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3))]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 48)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 7)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 1)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 49)]))
+            conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 14)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 21)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 28)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 35)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 42)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 49)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
+            conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 56)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 2)]))
+            conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(((rc.outer.inner*63) + floormod(threadIdx.x, 7)) + 56)]*kernel.shared_1[(((floordiv(threadIdx.x, 7)*96) + (rc.outer.inner*3)) + 50)]))
           }
         }
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[0]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[32]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[64]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[96]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[128]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[160]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[192]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[224]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[16]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[48]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[80]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[112]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[144]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[176]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[208]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[threadIdx.x]*kernel.shared_1[240]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[1]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[33]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[65]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[97]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[129]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[161]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[193]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[225]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[17]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[49]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[81]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[113]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[145]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[177]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[209]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 49)]*kernel.shared_1[241]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[2]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[34]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[66]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[98]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[130]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[162]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[194]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[226]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[18]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[50]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[82]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[114]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[146]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[178]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[210]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 98)]*kernel.shared_1[242]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[3]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[35]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[67]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[99]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[131]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[163]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[195]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[227]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[19]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[51]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[83]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[115]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[147]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[179]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[211]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 147)]*kernel.shared_1[243]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[4]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[36]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[68]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[100]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[132]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[164]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[196]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[228]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[20]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[52]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[84]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[116]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[148]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[180]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[212]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 196)]*kernel.shared_1[244]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[5]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[37]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[69]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[101]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[133]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[165]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[197]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[229]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[21]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[53]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[85]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[117]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[149]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[181]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[213]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 245)]*kernel.shared_1[245]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[6]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[38]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[70]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[102]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[134]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[166]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[198]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[230]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[22]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[54]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[86]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[118]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[150]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[182]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[214]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 294)]*kernel.shared_1[246]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[7]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[39]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[71]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[103]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[135]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[167]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[199]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[231]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[23]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[55]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[87]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[119]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[151]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[183]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[215]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 343)]*kernel.shared_1[247]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[8]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[40]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[72]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[104]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[136]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[168]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[200]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[232]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[24]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[56]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[88]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[120]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[152]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[184]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[216]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 392)]*kernel.shared_1[248]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[9]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[41]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[73]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[105]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[137]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[169]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[201]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[233]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[25]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[57]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[89]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[121]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[153]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[185]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[217]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 441)]*kernel.shared_1[249]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[10]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[42]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[74]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[106]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[138]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[170]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[202]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[234]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[26]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[58]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[90]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[122]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[154]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[186]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[218]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 490)]*kernel.shared_1[250]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[11]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[43]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[75]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[107]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[139]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[171]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[203]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[235]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[27]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[59]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[91]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[123]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[155]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[187]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[219]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 539)]*kernel.shared_1[251]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[12]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[44]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[76]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[108]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[140]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[172]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[204]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[236]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[28]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[60]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[92]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[124]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[156]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[188]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[220]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 588)]*kernel.shared_1[252]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[13]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[45]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[77]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[109]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[141]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[173]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[205]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[237]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[29]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[61]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[93]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[125]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[157]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[189]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[221]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 637)]*kernel.shared_1[253]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[14]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[46]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[78]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[110]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[142]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[174]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[206]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[238]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[30]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[62]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[94]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[126]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[158]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[190]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[222]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 686)]*kernel.shared_1[254]))
-        conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[15]))
-        conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[47]))
-        conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[79]))
-        conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[111]))
-        conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[143]))
-        conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[175]))
-        conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[207]))
-        conv2d_nchw_1[14] = (conv2d_nchw_1[14] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[239]))
-        conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[31]))
-        conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[63]))
-        conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[95]))
-        conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[127]))
-        conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[159]))
-        conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[191]))
-        conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[223]))
-        conv2d_nchw_1[15] = (conv2d_nchw_1[15] + (pad_temp.shared_1[(threadIdx.x + 735)]*kernel.shared_1[255]))
       }
     }
     for (i1.inner: int32, 0, 2) {
-      compute_3: Buffer(compute_2, float32, [25088], [])[(((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x)] = max((conv2d_nchw_1[i1.inner] + bias_3: Buffer(bias_2, float32, [512], [])[((blockIdx.x*16) + i1.inner)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 98)] = max((conv2d_nchw_1[(i1.inner + 2)] + bias_3[(((blockIdx.x*16) + i1.inner) + 2)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 196)] = max((conv2d_nchw_1[(i1.inner + 4)] + bias_3[(((blockIdx.x*16) + i1.inner) + 4)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 294)] = max((conv2d_nchw_1[(i1.inner + 6)] + bias_3[(((blockIdx.x*16) + i1.inner) + 6)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 392)] = max((conv2d_nchw_1[(i1.inner + 8)] + bias_3[(((blockIdx.x*16) + i1.inner) + 8)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 490)] = max((conv2d_nchw_1[(i1.inner + 10)] + bias_3[(((blockIdx.x*16) + i1.inner) + 10)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 588)] = max((conv2d_nchw_1[(i1.inner + 12)] + bias_3[(((blockIdx.x*16) + i1.inner) + 12)]), 0f32)
-      compute_3[((((blockIdx.x*784) + (i1.inner*49)) + threadIdx.x) + 686)] = max((conv2d_nchw_1[(i1.inner + 14)] + bias_3[(((blockIdx.x*16) + i1.inner) + 14)]), 0f32)
+      for (i2.inner: int32, 0, 7) {
+        compute_3: Buffer(compute_2, float32, [25088], [])[(((((blockIdx.x*3136) + (floordiv(threadIdx.x, 7)*98)) + (i1.inner*49)) + (i2.inner*7)) + floormod(threadIdx.x, 7))] = max((conv2d_nchw_1[((i1.inner*7) + i2.inner)] + bias_3: Buffer(bias_2, float32, [512], [])[(((blockIdx.x*64) + (floordiv(threadIdx.x, 7)*2)) + i1.inner)]), 0f32)
+      }
     }
   }
 }
@@ -1485,7 +656,7 @@ cooperative fetching, unrolling and operator fusion.</p>
 <span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.355 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.336 ms
 </pre></div>
 </div>
 </div>
@@ -1516,20 +687,20 @@ conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_
 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=2)
 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_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=1)
-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=8)
+conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=32)
+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=7)
+conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=7)
+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_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=8)
-conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=2)
+conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=1)
+conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=16)
 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=1)
+conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
 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)
 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 [...]
@@ -1537,10 +708,10 @@ 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=2)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=1)
-compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=8)
-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=7)
+compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=32)
+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=7)
+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_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)
@@ -1561,16 +732,16 @@ s[compute].bind(compute_i0_o_o_i_i1_o_o_i_fused_i2_o_o_i_fused_i3_o_o_i_fused, t
 compute_i0_o_i_i1_o_i_fused_i2_o_i_fused_i3_o_i_fused = s[compute].fuse(compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i)
 s[compute].bind(compute_i0_o_i_i1_o_i_fused_i2_o_i_fused_i3_o_i_fused, te.thread_axis(&quot;threadIdx.x&quot;))
 kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused = s[kernel_shared].fuse(kernel_shared_ax0, kernel_shared_ax1, kernel_shared_ax2, kernel_shared_ax3)
-kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=4)
+kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=1)
 s[kernel_shared].vectorize(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i)
-kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=49)
+kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_o, kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i = s[kernel_shared].split(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, factor=224)
 s[kernel_shared].bind(kernel_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o_i, te.thread_axis(&quot;threadIdx.x&quot;))
 pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused = s[pad_temp_shared].fuse(pad_temp_shared_ax0, pad_temp_shared_ax1, pad_temp_shared_ax2, pad_temp_shared_ax3)
-pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=1)
+pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_o, pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused_i = s[pad_temp_shared].split(pad_temp_shared_ax0_ax1_fused_ax2_fused_ax3_fused, factor=7)
 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=49)
+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=224)
 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;, 1024)
+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:
@@ -1588,891 +759,103 @@ CUDA source code:
   #define int64_t long long
   #define uint64_t unsigned long long
 #endif
-extern &quot;C&quot; __global__ void __launch_bounds__(49) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
-  float conv2d_nchw[16];
-  __shared__ float pad_temp_shared[784];
-  __shared__ float kernel_shared[256];
+extern &quot;C&quot; __global__ void __launch_bounds__(224) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+  float conv2d_nchw[14];
+  __shared__ float pad_temp_shared[1008];
+  __shared__ float kernel_shared[3072];
   conv2d_nchw[0] = 0.000000e+00f;
-  conv2d_nchw[2] = 0.000000e+00f;
-  conv2d_nchw[4] = 0.000000e+00f;
-  conv2d_nchw[6] = 0.000000e+00f;
-  conv2d_nchw[8] = 0.000000e+00f;
-  conv2d_nchw[10] = 0.000000e+00f;
-  conv2d_nchw[12] = 0.000000e+00f;
-  conv2d_nchw[14] = 0.000000e+00f;
-  conv2d_nchw[1] = 0.000000e+00f;
-  conv2d_nchw[3] = 0.000000e+00f;
-  conv2d_nchw[5] = 0.000000e+00f;
   conv2d_nchw[7] = 0.000000e+00f;
+  conv2d_nchw[1] = 0.000000e+00f;
+  conv2d_nchw[8] = 0.000000e+00f;
+  conv2d_nchw[2] = 0.000000e+00f;
   conv2d_nchw[9] = 0.000000e+00f;
+  conv2d_nchw[3] = 0.000000e+00f;
+  conv2d_nchw[10] = 0.000000e+00f;
+  conv2d_nchw[4] = 0.000000e+00f;
   conv2d_nchw[11] = 0.000000e+00f;
+  conv2d_nchw[5] = 0.000000e+00f;
+  conv2d_nchw[12] = 0.000000e+00f;
+  conv2d_nchw[6] = 0.000000e+00f;
   conv2d_nchw[13] = 0.000000e+00f;
-  conv2d_nchw[15] = 0.000000e+00f;
   for (int rc_outer_outer = 0; rc_outer_outer &lt; 32; ++rc_outer_outer) {
-    for (int ry_outer_outer = 0; ry_outer_outer &lt; 3; ++ry_outer_outer) {
+    for (int rx_outer_outer = 0; rx_outer_outer &lt; 3; ++rx_outer_outer) {
       __syncthreads();
-      pad_temp_shared[((int)threadIdx.x)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 8)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 49)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 41)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 98)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 90)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 147)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 139)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 188)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 245)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 237)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 294)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 286)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 343)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 335)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 392)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 384)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 441)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 433)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 490)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 482)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 539)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 531)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 588)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 580)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 637)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 629)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 686)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 678)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 735)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; (1 &lt;= (((int)threadIdx.x) % 7))) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 727)] : 0.000000e+00f);
-      kernel_shared[(((int)threadIdx.x) * 4)] = kernel[(((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3))];
-      kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 9)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 18)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 27)];
-      if (((int)threadIdx.x) &lt; 15) {
-        kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) &amp; 15) * 9)) + (ry_outer_outer * 3))];
-        kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) &amp; 15) * 9)) + (ry_outer_outer * 3))];
-        kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) &amp; 15) * 9)) + (ry_outer_outer * 3))];
-        kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[(((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) &amp; 15) * 9)) + (ry_outer_outer * 3))];
+      if (((int)threadIdx.x) &lt; 144) {
+        pad_temp_shared[(((int)threadIdx.x) * 7)] = ((((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) &amp;&amp; (1 &lt;= rx_outer_outer)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 8)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 1)] = (((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 7)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 2)] = (((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 6)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 3)] = (((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 5)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 4)] = (((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 4)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 5)] = (((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 3)] : 0.000000e+00f);
+        pad_temp_shared[((((int)threadIdx.x) * 7) + 6)] = ((((1 &lt;= (((int)threadIdx.x) % 9)) &amp;&amp; ((((int)threadIdx.x) % 9) &lt; 8)) &amp;&amp; (rx_outer_outer &lt; 2)) ? data[(((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 9) * 49)) + ((((int)threadIdx.x) % 9) * 7)) + rx_outer_outer) - 2)] : 0.000000e+00f);
       }
-      __syncthreads();
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
-      __syncthreads();
-      pad_temp_shared[((int)threadIdx.x)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 7)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 49)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 42)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 98)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 91)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 147)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 140)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 196)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 189)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 245)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 238)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 294)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 287)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 343)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 336)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 392)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 385)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 441)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 434)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 490)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 483)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 539)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 532)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 588)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 581)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 637)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 630)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 686)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 679)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 735)] = (((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 728)] : 0.000000e+00f);
-      kernel_shared[(((int)threadIdx.x) * 4)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 1)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 10)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 19)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 28)];
-      if (((int)threadIdx.x) &lt; 15) {
-        kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 1)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 1)];
+      kernel_shared[((int)threadIdx.x)] = kernel[(((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 224) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 448) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 672)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 64512)];
+      kernel_shared[(((int)threadIdx.x) + 896)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 896) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1120) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 129024)];
+      kernel_shared[(((int)threadIdx.x) + 1568)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1568) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1792) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 2016)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 193536)];
+      kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2240) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 2464)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2464) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 16) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 1) % 3) * 3)) + rx_outer_outer)];
+      kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) % 48) * 3)) + rx_outer_outer) + 258048)];
+      if (((int)threadIdx.x) &lt; 160) {
+        kernel_shared[(((int)threadIdx.x) + 2912)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 2912) / 48) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) + 32) % 48) / 3) * 9)) + (((((int)threadIdx.x) + 2) % 3) * 3)) + rx_outer_outer)];
       }
       __syncthreads();
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
-      __syncthreads();
-      pad_temp_shared[((int)threadIdx.x)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) - 6)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 49)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 43)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 98)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 92)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 147)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 141)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 196)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 190)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 245)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 239)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 294)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 288)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 343)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 337)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 392)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 386)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 441)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 435)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 490)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 484)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 539)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 533)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 588)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 582)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 637)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 631)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 686)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 680)] : 0.000000e+00f);
-      pad_temp_shared[(((int)threadIdx.x) + 735)] = ((((1 &lt;= ((((int)threadIdx.x) / 7) + ry_outer_outer)) &amp;&amp; (((((int)threadIdx.x) / 7) + ry_outer_outer) &lt; 8)) &amp;&amp; ((((int)threadIdx.x) % 7) &lt; 6)) ? data[((((rc_outer_outer * 784) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + 729)] : 0.000000e+00f);
-      kernel_shared[(((int)threadIdx.x) * 4)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 2)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 1)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 11)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 2)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 20)];
-      kernel_shared[((((int)threadIdx.x) * 4) + 3)] = kernel[((((((((int)blockIdx.x) * 73728) + ((((int)threadIdx.x) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((int)threadIdx.x) &amp; 3) * 36)) + (ry_outer_outer * 3)) + 29)];
-      if (((int)threadIdx.x) &lt; 15) {
-        kernel_shared[((((int)threadIdx.x) * 4) + 196)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 4) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 197)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 5) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 198)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 6) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 2)];
-        kernel_shared[((((int)threadIdx.x) * 4) + 199)] = kernel[((((((((int)blockIdx.x) * 73728) + (((((int)threadIdx.x) + 49) &gt;&gt; 2) * 4608)) + (rc_outer_outer * 144)) + ((((((int)threadIdx.x) * 4) + 7) &amp; 15) * 9)) + (ry_outer_outer * 3)) + 2)];
+      for (int rc_outer_inner = 0; rc_outer_inner &lt; 16; ++rc_outer_inner) {
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3))]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 48)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 1)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 49)]));
+        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
+        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 2)]));
+        conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 96) + (rc_outer_inner * 3)) + 50)]));
       }
-      __syncthreads();
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[0]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[32]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[64]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[96]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[128]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[160]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[192]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[224]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[16]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[48]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[80]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[112]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[144]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[176]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[208]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[((int)threadIdx.x)] * kernel_shared[240]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[1]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[33]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[65]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[97]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[129]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[161]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[193]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[225]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[17]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[49]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[81]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[113]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[145]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[177]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[209]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 49)] * kernel_shared[241]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[2]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[34]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[66]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[98]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[130]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[162]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[194]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[226]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[18]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[50]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[82]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[114]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[146]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[178]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[210]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 98)] * kernel_shared[242]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[3]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[35]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[67]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[99]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[131]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[163]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[195]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[227]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[19]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[51]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[83]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[115]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[147]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[179]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[211]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 147)] * kernel_shared[243]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[4]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[36]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[68]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[100]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[132]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[164]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[196]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[228]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[20]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[52]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[84]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[116]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[148]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[180]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[212]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 196)] * kernel_shared[244]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[5]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[37]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[69]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[101]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[133]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[165]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[197]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[229]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[21]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[53]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[85]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[117]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[149]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[181]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[213]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 245)] * kernel_shared[245]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[6]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[38]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[70]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[102]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[134]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[166]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[198]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[230]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[22]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[54]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[86]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[118]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[150]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[182]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[214]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 294)] * kernel_shared[246]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[7]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[39]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[71]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[103]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[135]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[167]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[199]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[231]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[23]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[55]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[87]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[119]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[151]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[183]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[215]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 343)] * kernel_shared[247]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[8]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[40]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[72]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[104]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[136]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[168]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[200]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[232]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[24]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[56]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[88]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[120]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[152]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[184]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[216]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 392)] * kernel_shared[248]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[9]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[41]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[73]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[105]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[137]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[169]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[201]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[233]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[25]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[57]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[89]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[121]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[153]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[185]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[217]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 441)] * kernel_shared[249]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[10]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[42]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[74]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[106]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[138]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[170]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[202]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[234]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[26]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[58]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[90]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[122]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[154]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[186]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[218]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 490)] * kernel_shared[250]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[11]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[43]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[75]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[107]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[139]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[171]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[203]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[235]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[27]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[59]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[91]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[123]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[155]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[187]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[219]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 539)] * kernel_shared[251]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[12]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[44]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[76]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[108]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[140]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[172]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[204]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[236]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[28]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[60]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[92]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[124]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[156]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[188]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[220]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 588)] * kernel_shared[252]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[13]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[45]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[77]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[109]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[141]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[173]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[205]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[237]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[29]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[61]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[93]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[125]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[157]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[189]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[221]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 637)] * kernel_shared[253]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[14]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[46]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[78]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[110]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[142]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[174]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[206]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[238]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[30]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[62]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[94]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[126]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[158]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[190]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[222]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 686)] * kernel_shared[254]));
-      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[15]));
-      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[47]));
-      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[79]));
-      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[111]));
-      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[143]));
-      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[175]));
-      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[207]));
-      conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[239]));
-      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[31]));
-      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[63]));
-      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[95]));
-      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[127]));
-      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[159]));
-      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[191]));
-      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[223]));
-      conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((int)threadIdx.x) + 735)] * kernel_shared[255]));
     }
   }
   for (int i1_inner = 0; i1_inner &lt; 2; ++i1_inner) {
-    compute[(((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x))] = max((conv2d_nchw[i1_inner] + bias[((((int)blockIdx.x) * 16) + i1_inner)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 98)] = max((conv2d_nchw[(i1_inner + 2)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 2)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 196)] = max((conv2d_nchw[(i1_inner + 4)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 4)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 294)] = max((conv2d_nchw[(i1_inner + 6)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 6)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 392)] = max((conv2d_nchw[(i1_inner + 8)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 8)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 490)] = max((conv2d_nchw[(i1_inner + 10)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 10)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 588)] = max((conv2d_nchw[(i1_inner + 12)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 12)]), 0.000000e+00f);
-    compute[((((((int)blockIdx.x) * 784) + (i1_inner * 49)) + ((int)threadIdx.x)) + 686)] = max((conv2d_nchw[(i1_inner + 14)] + bias[(((((int)blockIdx.x) * 16) + i1_inner) + 14)]), 0.000000e+00f);
+    for (int i2_inner = 0; i2_inner &lt; 7; ++i2_inner) {
+      compute[(((((((int)blockIdx.x) * 3136) + ((((int)threadIdx.x) / 7) * 98)) + (i1_inner * 49)) + (i2_inner * 7)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[((i1_inner * 7) + i2_inner)] + bias[(((((int)blockIdx.x) * 64) + ((((int)threadIdx.x) / 7) * 2)) + i1_inner)]), 0.000000e+00f);
+    }
   }
 }
 </pre></div>
@@ -2509,7 +892,7 @@ 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> ( 5 minutes  43.754 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 5 minutes  48.786 seconds)</p>
 <div class="sphx-glr-footer 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 sphx-glr-download-python 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_arm.html b/docs/how_to/tune_with_autoscheduler/tune_network_arm.html
index d3095eb887..c67290ed21 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_arm.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_arm.html
@@ -667,10 +667,10 @@ T_softmax_exp(i0, i1) = tir.exp((p0[i0, i1] - T_softmax_maxelem[i0]))
 T_softmax_expsum(i0) += T_softmax_exp[i0, k]
 T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-========== Task 5  (workload key: [&quot;be3babb9a46e32f66b717a3e2a2d522c&quot;, [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
+========== Task 5  (workload key: [&quot;8c53ca2904398da2889aa7508082d7bb&quot;, [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 1024]
-tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
 ========== Task 6  (workload key: [&quot;b638843f28bb2f041953d2a091167e8b&quot;, [1, 7, 7, 512], [1, 1, 512, 1024], [1, 1, 1, 1024], [1, 7, 7, 1024]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 512]
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 43e614a31c..314b80bc29 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -644,10 +644,10 @@ T_softmax_exp(i0, i1) = tir.exp((p0[i0, i1] - T_softmax_maxelem[i0]))
 T_softmax_expsum(i0) += T_softmax_exp[i0, k]
 T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-========== Task 10  (workload key: [&quot;be3babb9a46e32f66b717a3e2a2d522c&quot;, [1, 7, 7, 512], [1, 1, 1, 512]]) ==========
+========== Task 10  (workload key: [&quot;8c53ca2904398da2889aa7508082d7bb&quot;, [1, 7, 7, 512], [1, 1, 1, 512]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 512]
-tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
 ========== Task 11  (workload key: [&quot;40b1cf1fd37b0ef111b3cc0247302508&quot;, [1, 7, 7, 512], [4, 4, 512, 512], [1, 1, 1, 512], [1, 7, 7, 512]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 512]
@@ -756,12 +756,12 @@ conv2d_winograd(n, h, w, co) = inverse[floormod(h, 2), floormod(w, 2), ((((n*7)*
 p2 = PLACEHOLDER [1, 14, 14, 256]
 T_add(ax0, ax1, ax2, ax3) = (conv2d_winograd[ax0, ax1, ax2, ax3] + p2[ax0, ax1, ax2, ax3])
 
-========== Task 20  (workload key: [&quot;affd3c4a65f665e451a06d65bf32750d&quot;, [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
+========== Task 20  (workload key: [&quot;6d012ba18a086c11ee2b85c7324e16f2&quot;, [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
 p0 = PLACEHOLDER [1, 112, 112, 64]
 pad_temp(ax0, ax1, ax2, ax3) = tir.if_then_else(((((ax1 &gt;= 1) &amp;&amp; (ax1 &lt; 113)) &amp;&amp; (ax2 &gt;= 1)) &amp;&amp; (ax2 &lt; 113)), p0[ax0, (ax1 - 1), (ax2 - 1), ax3], -3.40282e+38f)
-tensor(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
+pool_max(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
 p1 = PLACEHOLDER [1, 1, 1, 64]
-T_add(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
+T_add(ax0, ax1, ax2, ax3) = (pool_max[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
 T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
 ========== Task 21  (workload key: [&quot;0fad1b42d0d33418e0a8d15d3bbad3c9&quot;, [1, 14, 14, 256], [1, 1, 256, 512], [1, 7, 7, 512]]) ==========
@@ -915,7 +915,7 @@ 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)
-   8.2044       8.2046       8.2096       8.1990       0.0043
+   7.8675       7.8677       7.8690       7.8659       0.0013
 </pre></div>
 </div>
 </div>
@@ -937,7 +937,7 @@ 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  4.752 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  4.757 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autoscheduler-tune-network-cuda-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/eafe360d52540634c9eea0fa89e804bd/tune_network_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_network_cuda.py</span></code></a></p>
diff --git a/docs/how_to/tune_with_autoscheduler/tune_network_mali.html b/docs/how_to/tune_with_autoscheduler/tune_network_mali.html
index 95a85b9673..934c9285d3 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_mali.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_mali.html
@@ -573,10 +573,10 @@ p2 = PLACEHOLDER [1, 1, 1, 1024]
 T_add(ax0, ax1, ax2, ax3) = (conv2d_nhwc[ax0, ax1, ax2, ax3] + p2[ax0, 0, 0, ax3])
 T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
-========== Task 5  (workload key: [&quot;be3babb9a46e32f66b717a3e2a2d522c&quot;, [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
+========== Task 5  (workload key: [&quot;8c53ca2904398da2889aa7508082d7bb&quot;, [1, 7, 7, 1024], [1, 1, 1, 1024]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 1024]
-tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
 ========== Task 6  (workload key: [&quot;7d79c516e212fe1d73f5dbb90eaca2cf&quot;, [1, 1000], [1, 1000]]) ==========
 p0 = PLACEHOLDER [1, 1000]
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 c60a1acb15..ab11af36a4 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -638,10 +638,10 @@ T_softmax_exp(i0, i1) = tir.exp((p0[i0, i1] - T_softmax_maxelem[i0]))
 T_softmax_expsum(i0) += T_softmax_exp[i0, k]
 T_softmax_norm(i0, i1) = (T_softmax_exp[i0, i1]/T_softmax_expsum[i0])
 
-========== Task 11  (workload key: [&quot;be3babb9a46e32f66b717a3e2a2d522c&quot;, [1, 7, 7, 2048], [1, 1, 1, 2048]]) ==========
+========== Task 11  (workload key: [&quot;8c53ca2904398da2889aa7508082d7bb&quot;, [1, 7, 7, 2048], [1, 1, 1, 2048]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 2048]
-tensor(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
-tensor(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
+adaptive_pool_sum(ax0, ax1, ax2, ax3) += p0[ax0, ((ax1*7) + rv0), ((ax2*7) + rv1), ax3]
+adaptive_pool_avg(ax0, ax1, ax2, ax3) = (adaptive_pool_sum[ax0, ax1, ax2, ax3]/(float32((select((bool)1, ((ax1 + 1)*7), (((ax1 + 1)*7) + 1)) - (ax1*7)))*float32((select((bool)1, ((ax2 + 1)*7), (((ax2 + 1)*7) + 1)) - (ax2*7)))))
 
 ========== Task 12  (workload key: [&quot;3060808fc5c74e18b1276729071fbae0&quot;, [1, 7, 7, 512], [1, 1, 512, 2048], [1, 7, 7, 2048], [1, 7, 7, 2048]]) ==========
 p0 = PLACEHOLDER [1, 7, 7, 512]
@@ -777,12 +777,12 @@ T_matmul_NT(i, j) += (p0[i, k]*p1[j, k])
 p2 = PLACEHOLDER [1, 1000]
 T_add(ax0, ax1) = (T_matmul_NT[ax0, ax1] + p2[ax0, ax1])
 
-========== Task 27  (workload key: [&quot;affd3c4a65f665e451a06d65bf32750d&quot;, [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
+========== Task 27  (workload key: [&quot;6d012ba18a086c11ee2b85c7324e16f2&quot;, [1, 112, 112, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
 p0 = PLACEHOLDER [1, 112, 112, 64]
 pad_temp(ax0, ax1, ax2, ax3) = tir.if_then_else(((((ax1 &gt;= 1) &amp;&amp; (ax1 &lt; 113)) &amp;&amp; (ax2 &gt;= 1)) &amp;&amp; (ax2 &lt; 113)), p0[ax0, (ax1 - 1), (ax2 - 1), ax3], -3.40282e+38f)
-tensor(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
+pool_max(ax0, ax1, ax2, ax3) max= pad_temp[ax0, ((ax1*2) + rv0), ((ax2*2) + rv1), ax3]
 p1 = PLACEHOLDER [1, 1, 1, 64]
-T_add(ax0, ax1, ax2, ax3) = (tensor[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
+T_add(ax0, ax1, ax2, ax3) = (pool_max[ax0, ax1, ax2, ax3] + p1[ax0, 0, 0, ax3])
 T_relu(ax0, ax1, ax2, ax3) = max(T_add[ax0, ax1, ax2, ax3], 0f)
 
 ========== Task 28  (workload key: [&quot;6d628209072e3e3dd8f49359935acea6&quot;, [1, 56, 56, 64], [1, 1, 64, 64], [1, 1, 1, 64], [1, 56, 56, 64]]) ==========
@@ -934,7 +934,7 @@ 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)
-  763.1550     763.9276     764.1947     761.3427      1.2861
+  757.8686     757.7418     758.5651     757.2990      0.5246
 </pre></div>
 </div>
 </div>
@@ -956,7 +956,7 @@ 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  34.813 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  35.031 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autoscheduler-tune-network-x86-py">
 <div class="sphx-glr-download sphx-glr-download-python 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 df4119901d..6e53a3e78b 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -632,77 +632,77 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
              placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [128, 512], []),
              compute: Buffer(compute_2: Pointer(float32), float32, [128, 512], [])}
   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.i1.outer.fused: int32, 0, 256) &quot;parallel&quot; {
-    allocate(compute_3: Pointer(global float32), float32, [256]), storage_scope = global {
-      for (nb_j.inner: int32, 0, 2) {
-        for (i.inner.init: int32, 0, 8) {
-          let cse_var_1: int32 = ((i.inner.init*32) + (nb_j.inner*16))
-           {
-            compute_4: Buffer(compute_3, float32, [256], [])[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
-          }
-        }
-        for (elem_idx: int32, 0, let cse_var_2: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner) in (placeholder_15: Buffer(placeholder_13, int32, [33], [])[(cse_var_2 + 1)] - placeholder_15[cse_var_2])) {
-          for (i.inner: int32, 0, 8) {
-            let cse_var_21: int32 = (elem_idx*16)
-            let cse_var_20: int32 = ((i.inner*32) + (nb_j.inner*16))
-            let cse_var_19: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner)
-            let cse_var_18: int32 = ((floordiv(i0.outer.i1.outer.fused, 16)*2048) + (i.inner*256))
-            let cse_var_17: int32 = (cse_var_20 + 9)
-            let cse_var_16: int32 = (cse_var_20 + 8)
-            let cse_var_15: int32 = (cse_var_20 + 7)
-            let cse_var_14: int32 = (cse_var_20 + 6)
-            let cse_var_13: int32 = (cse_var_20 + 5)
-            let cse_var_12: int32 = (cse_var_20 + 4)
-            let cse_var_11: int32 = (cse_var_20 + 3)
-            let cse_var_10: int32 = (cse_var_20 + 2)
-            let cse_var_9: int32 = (cse_var_20 + 15)
-            let cse_var_8: int32 = (cse_var_20 + 14)
-            let cse_var_7: int32 = (cse_var_20 + 13)
-            let cse_var_6: int32 = (cse_var_20 + 12)
-            let cse_var_5: int32 = (cse_var_20 + 11)
-            let cse_var_4: int32 = (cse_var_20 + 10)
-            let cse_var_3: int32 = (cse_var_20 + 1)
+  for (i0.outer.i1.outer.fused: int32, 0, 32) &quot;parallel&quot; {
+    allocate(compute_3: Pointer(global float32), float32, [2048]), storage_scope = global {
+      for (i.outer.inner: int32, 0, 4) {
+        for (nb_j.inner: int32, 0, 2) {
+          for (i.inner.init: int32, 0, 16) {
+            let cse_var_1: int32 = (((i.outer.inner*512) + (i.inner.init*32)) + (nb_j.inner*16))
              {
-              compute_4[cse_var_20] = (compute_4[cse_var_20] + (placeholder_16: Buffer(placeholder_11, float32, [78656], [])[((placeholder_15[cse_var_19]*16) + cse_var_21)]*max(placeholder_17: Buffer(placeholder_10, float32, [32768], [])[(cse_var_18 + placeholder_18: Buffer(placeholder_12, int32, [4916], [])[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 1)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 2)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 3)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 4)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 5)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 6)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 7)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 8)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 9)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 10)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 11)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 12)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 13)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 14)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
-              compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_16[(((placeholder_15[cse_var_19]*16) + cse_var_21) + 15)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_19] + elem_idx)])], 0f32)))
+              compute_4: Buffer(compute_3, float32, [2048], [])[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
+            }
+          }
+          for (elem_idx: int32, 0, let cse_var_2: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner) in (placeholder_15: Buffer(placeholder_13, int32, [33], [])[(cse_var_2 + 1)] - placeholder_15[cse_var_2])) {
+            for (i.inner: int32, 0, 16) {
+              let cse_var_21: int32 = (elem_idx*16)
+              let cse_var_20: int32 = ((floormod(i0.outer.i1.outer.fused, 16)*2) + nb_j.inner)
+              let cse_var_19: int32 = (((i.outer.inner*512) + (i.inner*32)) + (nb_j.inner*16))
+              let cse_var_18: int32 = (((floordiv(i0.outer.i1.outer.fused, 16)*16384) + (i.outer.inner*4096)) + (i.inner*256))
+              let cse_var_17: int32 = (cse_var_19 + 9)
+              let cse_var_16: int32 = (cse_var_19 + 8)
+              let cse_var_15: int32 = (cse_var_19 + 7)
+              let cse_var_14: int32 = (cse_var_19 + 6)
+              let cse_var_13: int32 = (cse_var_19 + 5)
+              let cse_var_12: int32 = (cse_var_19 + 4)
+              let cse_var_11: int32 = (cse_var_19 + 3)
+              let cse_var_10: int32 = (cse_var_19 + 2)
+              let cse_var_9: int32 = (cse_var_19 + 15)
+              let cse_var_8: int32 = (cse_var_19 + 14)
+              let cse_var_7: int32 = (cse_var_19 + 13)
+              let cse_var_6: int32 = (cse_var_19 + 12)
+              let cse_var_5: int32 = (cse_var_19 + 11)
+              let cse_var_4: int32 = (cse_var_19 + 10)
+              let cse_var_3: int32 = (cse_var_19 + 1)
+               {
+                compute_4[cse_var_19] = (compute_4[cse_var_19] + (placeholder_16: Buffer(placeholder_11, float32, [78656], [])[((placeholder_15[cse_var_20]*16) + cse_var_21)]*max(placeholder_17: Buffer(placeholder_10, float32, [32768], [])[(cse_var_18 + placeholder_18: Buffer(placeholder_12, int32, [4916], [])[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_3] = (compute_4[cse_var_3] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 1)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_10] = (compute_4[cse_var_10] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 2)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_11] = (compute_4[cse_var_11] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 3)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_12] = (compute_4[cse_var_12] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 4)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_13] = (compute_4[cse_var_13] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 5)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_14] = (compute_4[cse_var_14] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 6)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_15] = (compute_4[cse_var_15] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 7)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_16] = (compute_4[cse_var_16] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 8)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_17] = (compute_4[cse_var_17] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 9)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_4] = (compute_4[cse_var_4] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 10)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_5] = (compute_4[cse_var_5] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 11)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_6] = (compute_4[cse_var_6] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 12)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_7] = (compute_4[cse_var_7] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 13)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_8] = (compute_4[cse_var_8] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 14)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+                compute_4[cse_var_9] = (compute_4[cse_var_9] + (placeholder_16[(((placeholder_15[cse_var_20]*16) + cse_var_21) + 15)]*max(placeholder_17[(cse_var_18 + placeholder_18[(placeholder_15[cse_var_20] + elem_idx)])], 0f32)))
+              }
             }
           }
         }
       }
-      for (i0.inner: int32, 0, 8) {
-        for (i1.inner: int32, 0, 32) {
-          let cse_var_22: int32 = ((((floordiv(i0.outer.i1.outer.fused, 16)*4096) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 16)*32)) + i1.inner)
-          compute_5: Buffer(compute_2, float32, [65536], [])[cse_var_22] = max((compute_4[((i0.inner*32) + i1.inner)] + placeholder_19: Buffer(placeholder_14, float32, [65536], [])[cse_var_22]), 0f32)
-        }
+      for (i0.inner: int32, 0, 64) {
+        let cse_var_22: int32 = (((floordiv(i0.outer.i1.outer.fused, 16)*32768) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 16)*32))
+        compute_5: Buffer(compute_2, float32, [65536], [])[ramp(cse_var_22, 1, 32)] = max((compute_4[ramp((i0.inner*32), 1, 32)] + placeholder_19: Buffer(placeholder_14, float32, [65536], [])[ramp(cse_var_22, 1, 32)]), broadcast(0f32, 32))
       }
     }
   }
@@ -740,7 +740,7 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
 <span class="p">)</span>
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.906 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.724 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 25273eb709..c648dc35fe 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>00:35.755</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:46.368</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -349,11 +349,11 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:35.720</p></td>
+<td><p>00:46.333</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:00.019</p></td>
+<td><p>00:00.020</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
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 8fe20e60e2..156758689e 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -567,8 +567,7 @@ for this template</p>
 waiting for device...
 device available
 Get devices for measurement successfully!
-No: 1   GFLOPS: 2.32/2.32       result: MeasureResult(costs=(0.099728897,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.013670921325684, timestamp=1668620461.9634821) [(&#39;tile_f&#39;, [-1, 8, 8, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,1781024
-No: 2   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+No: 1   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -690,9 +689,26 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 16, 1, 8]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 2, 128]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,10440020
-No: 3   GFLOPS: 1.98/2.32       result: MeasureResult(costs=(0.11707182299999999,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.8137314319610596, timestamp=1668620464.7798214)        [(&#39;tile_f&#39;, [-1, 128, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 4, 2]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2561687
-No: 4   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+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, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 64, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2927506
+No: 2   GFLOPS: 0.00/0.00       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
+    return self.__get_result()
+  File &quot;/usr/lib/python3.7/concurrent/futures/_base.py&quot;, line 384, in __get_result
+    raise self._exception
+  File &quot;/usr/lib/python3.7/concurrent/futures/thread.py&quot;, line 57, in run
+    result = self.fn(*self.args, **self.kwargs)
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 432, in &lt;lambda&gt;
+    worker = lambda *args: self._worker_run(*args)
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 401, in _worker_run
+    return proc.recv()
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 309, in recv
+    raise TimeoutError()
+TimeoutError
+
+        [(&#39;tile_f&#39;, [-1, 1, 1, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 16, 2]), (&#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;, 0)],None,4310425
+No: 3   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -814,8 +830,9 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 256, 2]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,3355501
-No: 5   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 64, 8]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 256]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5607743
+No: 4   GFLOPS: 397.48/397.48   result: MeasureResult(costs=(0.0005824169812206572,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2404870986938477, timestamp=1668621352.333588)       [(&#39;tile_f&#39;, [-1, 4, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 8, 1]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,8337662
+No: 5   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -937,8 +954,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 2, 32]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 8, 16]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,3036631
-No: 6   GFLOPS: 0.00/2.32       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 256, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 64, 1]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7765172
+No: 6   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1060,9 +1077,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 512, 1]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7778477
-No: 7   GFLOPS: 10.83/10.83     result: MeasureResult(costs=(0.021368804833333335,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.6941542625427246, timestamp=1668620468.6158936)       [(&#39;tile_f&#39;, [-1, 32, 8, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 2]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,3326432
-No: 8   GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 256, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 8]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7259219
+No: 7   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1184,8 +1200,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 256, 2, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 16, 8]), (&#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,6304998
-No: 9   GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 64, 4]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 2, 64]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5392113
+No: 8   GFLOPS: 0.00/397.48     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1307,8 +1323,9 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 4, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 256, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,609645
-No: 10  GFLOPS: 0.00/10.83      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 16, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 32, 2]), (&#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;, 0)],None,1216947
+No: 9   GFLOPS: 851.04/851.04   result: MeasureResult(costs=(0.0002720207644110276,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.3329086303710938, timestamp=1668621356.4669878)      [(&#39;tile_f&#39;, [-1, 1, 8, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 8, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,3497859
+No: 10  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1430,10 +1447,9 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 4]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 16, 4]), (&#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,9569677
-No: 11  GFLOPS: 160.31/160.31   result: MeasureResult(costs=(0.0014441192545454544,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.183053970336914, timestamp=1668620471.017138)        [(&#39;tile_f&#39;, [-1, 2, 4, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 16]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9412773
-No: 12  GFLOPS: 15.38/160.31    result: MeasureResult(costs=(0.015050043285714285,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.1918892860412598, timestamp=1668620471.784751)        [(&#39;tile_f&#39;, [-1, 8, 16, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 1]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,2520449
-No: 13  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 512, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#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;, 0), (&#39;unroll_explicit&#39;, 1)],None,6432369
+No: 11  GFLOPS: 40.76/851.04    result: MeasureResult(costs=(0.005679272050000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4551396369934082, timestamp=1668621357.2073452)       [(&#39;tile_f&#39;, [-1, 4, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 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,6203782
+No: 12  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1555,8 +1571,9 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 2, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 16, 1]), (&#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;, 0)],None,4276066
-No: 14  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 16, 2, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 16, 4]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,6857028
+No: 13  GFLOPS: 2.45/851.04     result: MeasureResult(costs=(0.09452290325,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.229509353637695, timestamp=1668621369.0406008)       [(&#39;tile_f&#39;, [-1, 32, 2, 1]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 32, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7025275
+No: 14  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1678,8 +1695,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 4, 8]), (&#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, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 0)],None,3255932
-No: 15  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 4, 64, 2]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#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, 3, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4547496
+No: 15  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1801,8 +1818,26 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 32, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 16]), (&#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,3803233
-No: 16  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 4, 32]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 16]), (&#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;, 0)],None,1284555
+No: 16  GFLOPS: 0.00/851.04     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
+    return self.__get_result()
+  File &quot;/usr/lib/python3.7/concurrent/futures/_base.py&quot;, line 384, in __get_result
+    raise self._exception
+  File &quot;/usr/lib/python3.7/concurrent/futures/thread.py&quot;, line 57, in run
+    result = self.fn(*self.args, **self.kwargs)
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 432, in &lt;lambda&gt;
+    worker = lambda *args: self._worker_run(*args)
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 401, in _worker_run
+    return proc.recv()
+  File &quot;/workspace/python/tvm/contrib/popen_pool.py&quot;, line 309, in recv
+    raise TimeoutError()
+TimeoutError
+
+        [(&#39;tile_f&#39;, [-1, 8, 1, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 16]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7284038
+No: 17  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -1924,9 +1959,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 1, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 16, 32]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5578288
-No: 17  GFLOPS: 36.23/160.31    result: MeasureResult(costs=(0.006389735090909091,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.769231081008911, timestamp=1668620476.9480226)        [(&#39;tile_f&#39;, [-1, 1, 4, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 16]), (&#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,4576115
-No: 18  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 4, 32, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 64, 8]), (&#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,9604632
+No: 18  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -2048,8 +2082,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 1, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 64]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7713998
-No: 19  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 16, 8]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 32, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5827079
+No: 19  GFLOPS: 0.00/851.04     result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
@@ -2171,130 +2205,8 @@ Traceback (most recent call last):
   File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 2, 32]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 32]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,8664231
-No: 20  GFLOPS: 0.00/160.31     result: Traceback (most recent call last):
-  File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 588, 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 540, in _build_func_common
-    func = build(s, args, target_host=task.target_host, runtime=runtime)
-  File &quot;/workspace/python/tvm/driver/build_module.py&quot;, line 227, in build
-    input_mod = lower(inputs, args, name=name, binds=binds)
-  File &quot;/workspace/python/tvm/driver/build_module.py&quot;, line 134, in lower
-    return ffi.lower_schedule(inp, args, name, binds, simple_mode)
-  File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
-  File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 276, in tvm._ffi._cy3.core.FuncCall
-  File &quot;tvm/_ffi/_cython/./base.pxi&quot;, line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
-  24: TVMFuncCall
-        at ../src/runtime/c_runtime_api.cc:477
-  23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-        at ../include/tvm/runtime/packed_func.h:1217
-  22: Call
-        at ../include/tvm/runtime/packed_func.h:1213
-  21: operator()
-        at ../include/tvm/runtime/packed_func.h:1731
-  20: unpack_call&lt;tvm::IRModule, 5, tvm::&lt;lambda(tvm::te::Schedule, const tvm::runtime::Array&lt;tvm::runtime::ObjectRef&gt;&amp;, const tvm::runtime::String&amp;, const tvm::runtime::Map&lt;tvm::te::Tensor, tvm::tir::Buffer&gt;&amp;, bool)&gt; &gt;
-        at ../include/tvm/runtime/packed_func.h:1671
-  19: run&lt;&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  18: run&lt;tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  17: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  16: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  15: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  14: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1646
-  13: operator()
-        at ../src/driver/driver_api.cc:388
-  12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array&lt;tvm::runtime::ObjectRef, void&gt; const&amp;, std::__cxx11::basic_string&lt;char, std::char_traits&lt;char&gt;, std::allocator&lt;char&gt; &gt; const&amp;, std::unordered_map&lt;tvm::te::Tensor, tvm::tir::Buffer, std::hash&lt;tvm::te::Tensor&gt;, std::equal_to&lt;tvm::te::Tensor&gt;, std::allocator&lt;std::pair&lt;tvm::te::Tensor const, tvm::tir::Buffer&gt; &gt; &gt; const&amp;, tvm::GlobalVarSupply, bool)
-        at ../src/driver/driver_api.cc:374
-  11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array&lt;tvm::transform::Pass, void&gt;)
-        at ../src/driver/driver_api.cc:269
-  10: tvm::transform::Pass::operator()(tvm::IRModule) const
-        at ../src/ir/transform.cc:258
-  9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:274
-  8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:453
-  7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:274
-  6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/tir/ir/transform.cc:100
-  5: tvm::runtime::TypedPackedFunc&lt;tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)&gt;::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
-        at ../include/tvm/runtime/packed_func.h:1750
-  4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher&lt;tvm::tir::PrimFunc&gt;::run&lt;tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext&gt;(tvm::runtime::PackedFunc const&amp;, tvm::tir::PrimFunc&amp;&amp;, tvm::IRModule&amp;&amp;, tvm::transform::PassContext&amp;&amp;)
-        at ../include/tvm/runtime/packed_func.h:1694
-  3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()&lt;tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext&gt;(tvm::tir::PrimFunc&amp;&amp;, tvm::IRModule&amp;&amp;, tvm::transform::PassContext&amp;&amp;) const
-        at ../include/tvm/runtime/packed_func.h:1618
-  2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-        at ../include/tvm/runtime/packed_func.h:1217
-  1: Call
-        at ../include/tvm/runtime/packed_func.h:1213
-  0: operator()
-        at ../src/runtime/c_runtime_api.cc:534
-  File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
-  File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, in verify_pass
-    raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
-  24: TVMFuncCall
-        at ../src/runtime/c_runtime_api.cc:477
-  23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-        at ../include/tvm/runtime/packed_func.h:1217
-  22: Call
-        at ../include/tvm/runtime/packed_func.h:1213
-  21: operator()
-        at ../include/tvm/runtime/packed_func.h:1731
-  20: unpack_call&lt;tvm::IRModule, 5, tvm::&lt;lambda(tvm::te::Schedule, const tvm::runtime::Array&lt;tvm::runtime::ObjectRef&gt;&amp;, const tvm::runtime::String&amp;, const tvm::runtime::Map&lt;tvm::te::Tensor, tvm::tir::Buffer&gt;&amp;, bool)&gt; &gt;
-        at ../include/tvm/runtime/packed_func.h:1671
-  19: run&lt;&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  18: run&lt;tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  17: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  16: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  15: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1631
-  14: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1646
-  13: operator()
-        at ../src/driver/driver_api.cc:388
-  12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array&lt;tvm::runtime::ObjectRef, void&gt; const&amp;, std::__cxx11::basic_string&lt;char, std::char_traits&lt;char&gt;, std::allocator&lt;char&gt; &gt; const&amp;, std::unordered_map&lt;tvm::te::Tensor, tvm::tir::Buffer, std::hash&lt;tvm::te::Tensor&gt;, std::equal_to&lt;tvm::te::Tensor&gt;, std::allocator&lt;std::pair&lt;tvm::te::Tensor const, tvm::tir::Buffer&gt; &gt; &gt; const&amp;, tvm::GlobalVarSupply, bool)
-        at ../src/driver/driver_api.cc:374
-  11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array&lt;tvm::transform::Pass, void&gt;)
-        at ../src/driver/driver_api.cc:269
-  10: tvm::transform::Pass::operator()(tvm::IRModule) const
-        at ../src/ir/transform.cc:258
-  9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:274
-  8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:453
-  7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/ir/transform.cc:274
-  6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&amp;) const
-        at ../src/tir/ir/transform.cc:100
-  5: tvm::runtime::TypedPackedFunc&lt;tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)&gt;::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
-        at ../include/tvm/runtime/packed_func.h:1750
-  4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher&lt;tvm::tir::PrimFunc&gt;::run&lt;tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext&gt;(tvm::runtime::PackedFunc const&amp;, tvm::tir::PrimFunc&amp;&amp;, tvm::IRModule&amp;&amp;, tvm::transform::PassContext&amp;&amp;)
-        at ../include/tvm/runtime/packed_func.h:1694
-  3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()&lt;tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext&gt;(tvm::tir::PrimFunc&amp;&amp;, tvm::IRModule&amp;&amp;, tvm::transform::PassContext&amp;&amp;) const
-        at ../include/tvm/runtime/packed_func.h:1618
-  2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
-        at ../include/tvm/runtime/packed_func.h:1217
-  1: Call
-        at ../include/tvm/runtime/packed_func.h:1213
-  0: operator()
-        at ../src/runtime/c_runtime_api.cc:534
-  File &quot;tvm/_ffi/_cython/./packed_func.pxi&quot;, line 56, in tvm._ffi._cy3.core.tvm_callback
-  File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 871, 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, 256, 1]), (&#39;tile_y&#39;, [-1, 7, 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,943632
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 64, 1]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 32]), (&#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,7500508
+No: 20  GFLOPS: 28.37/851.04    result: MeasureResult(costs=(0.008161008923076923,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.737098217010498, timestamp=1668621372.0138779)        [(&#39;tile_f&#39;, [-1, 1, 4, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 4]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,6845812
 </pre></div>
 </div>
 <p>Finally we can inspect the best config from log file, check correctness,
@@ -2333,9 +2245,9 @@ and measure running time.</p>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Finish loading 20 records
 
 Best config:
-[(&#39;tile_f&#39;, [-1, 2, 4, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 16]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9412773
+[(&#39;tile_f&#39;, [-1, 1, 8, 2]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 8, 1]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,3497859
 Finish loading 20 records
-Time cost of this operator: 0.001880
+Time cost of this operator: 0.000728
 </pre></div>
 </div>
 <div class="sphx-glr-footer 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 bef57d13b5..bf4e94fd6e 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -596,10 +596,10 @@ 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  Measurements(us)
 ---------                                     ---                                           --------  -------  -----              ------  -------  ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  309.4     98.728   (1, 2, 10, 10, 3)  2       1        [309.4]
-tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.023     0.965    (1, 6, 10, 10)     1       1        [3.023]
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.962     0.307    (1, 1, 10, 10, 3)  1       1        [0.962]
-Total_time                                    -                                             313.386   -        -                  -       -        -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  311.1     98.715   (1, 2, 10, 10, 3)  2       1        [311.1]
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.082     0.978    (1, 6, 10, 10)     1       1        [3.082]
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.969     0.307    (1, 1, 10, 10, 3)  1       1        [0.969]
+Total_time                                    -                                             315.151   -        -                  -       -        -
 </pre></div>
 </div>
 </div>
@@ -650,10 +650,10 @@ 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  Measurements(us)
 ---------                                     ---                                           --------  -------  -----              ------  -------  ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  101.8     97.407   (1, 6, 10, 10, 1)  2       1        [101.8]
-tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.756     1.68     (1, 6, 10, 10)     1       1        [1.756]
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.955     0.913    (1, 1, 10, 10, 3)  1       1        [0.955]
-Total_time                                    -                                             104.51    -        -                  -       -        -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  102.9     97.488   (1, 6, 10, 10, 1)  2       1        [102.9]
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.808     1.713    (1, 6, 10, 10)     1       1        [1.808]
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.843     0.799    (1, 3, 10, 10, 1)  1       1        [0.843]
+Total_time                                    -                                             105.552   -        -                  -       -        -
 </pre></div>
 </div>
 <div class="sphx-glr-footer 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/micro_pytorch.html b/docs/how_to/work_with_microtvm/micro_pytorch.html
index 245a0bfd4f..6a17d94cbb 100644
--- a/docs/how_to/work_with_microtvm/micro_pytorch.html
+++ b/docs/how_to/work_with_microtvm/micro_pytorch.html
@@ -440,8 +440,7 @@ download a cat image and preprocess it to use as the model input.</p>
 Downloading: &quot;https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth&quot; to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2_qnnpack_37f702c5.pth
 
   0%|          | 0.00/3.42M [00:00&lt;?, ?B/s]
- 76%|#######6  | 2.61M/3.42M [00:00&lt;00:00, 27.3MB/s]
-100%|##########| 3.42M/3.42M [00:00&lt;00:00, 27.9MB/s]
+100%|##########| 3.42M/3.42M [00:00&lt;00:00, 65.2MB/s]
 /workspace/python/tvm/relay/frontend/pytorch_utils.py:47: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
   return LooseVersion(torch_ver) &gt; ver
 /venv/apache-tvm-py3.7/lib/python3.7/site-packages/setuptools/_distutils/version.py:346: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
@@ -565,7 +564,7 @@ via the host <cite>main.cc`</cite> or if a Zephyr emulated board is selected as
 Torch top-1 id: 282, class name: tiger cat
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  5.341 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  5.355 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-pytorch-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/12b9ecc04c41abaa12022061771821d1/micro_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">micro_pytorch.py</span></code></a></p>
diff --git a/docs/how_to/work_with_microtvm/micro_train.html b/docs/how_to/work_with_microtvm/micro_train.html
index d63332177b..96aa9500d0 100644
--- a/docs/how_to/work_with_microtvm/micro_train.html
+++ b/docs/how_to/work_with_microtvm/micro_train.html
@@ -530,7 +530,7 @@ take about <strong>2 minutes</strong> to download the Stanford Cars, while COCO
 <a href="https://docs.python.org/3/library/shutil.html#shutil.move" title="shutil.move" class="sphx-glr-backref-module-shutil sphx-glr-backref-type-py-function"><span class="n">shutil</span><span class="o">.</span><span class="n">move</span></a><span class="p">(</span><span class="sa">f</span><span class="s2">&quot;</span><span class="si">{</span><a href="https://docs.python.org/3/library/stdtypes.html#str" title="builtins.str" class="sphx-glr-backref-module-builtins sphx-glr-backref-typ [...]
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&#39;/tmp/tmpe0nb0235/images/random&#39;
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&#39;/tmp/tmpa9mo0yuu/images/random&#39;
 </pre></div>
 </div>
 </div>
@@ -590,8 +590,8 @@ objects to other stuff? We can display some examples from our datasets using <co
     <span class="n">plt</span><span class="o">.</span><span class="n">axis</span><span class="p">(</span><span class="s2">&quot;off&quot;</span><span class="p">)</span>
 </pre></div>
 </div>
-<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmpe0nb0235/images/target contains 8144 images
-/tmp/tmpe0nb0235/images/random contains 5000 images
+<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmpa9mo0yuu/images/target contains 8144 images
+/tmp/tmpa9mo0yuu/images/random contains 5000 images
 </pre></div>
 </div>
 </div>
@@ -703,13 +703,13 @@ the time on our validation set).</p>
 </pre></div>
 </div>
 <div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Epoch 1/3
-328/328 - 47s - loss: 0.2109 - accuracy: 0.9260 - val_loss: 0.1198 - val_accuracy: 0.9569 - 47s/epoch - 144ms/step
+328/328 - 47s - loss: 0.2387 - accuracy: 0.9199 - val_loss: 0.1355 - val_accuracy: 0.9494 - 47s/epoch - 144ms/step
 Epoch 2/3
-328/328 - 44s - loss: 0.1042 - accuracy: 0.9624 - val_loss: 0.1374 - val_accuracy: 0.9498 - 44s/epoch - 133ms/step
+328/328 - 44s - loss: 0.1044 - accuracy: 0.9632 - val_loss: 0.1209 - val_accuracy: 0.9581 - 44s/epoch - 133ms/step
 Epoch 3/3
-328/328 - 44s - loss: 0.0650 - accuracy: 0.9738 - val_loss: 0.1059 - val_accuracy: 0.9660 - 44s/epoch - 133ms/step
+328/328 - 44s - loss: 0.0669 - accuracy: 0.9757 - val_loss: 0.1102 - val_accuracy: 0.9619 - 44s/epoch - 133ms/step
 
-&lt;keras.callbacks.History object at 0x7f34dd15a3d0&gt;
+&lt;keras.callbacks.History object at 0x7f1545f86750&gt;
 </pre></div>
 </div>
 </div>
@@ -971,7 +971,7 @@ as intended.</p>
 <p>From here, we could modify the model to read live images from the camera - we have another
 Arduino tutorial for how to do that <a class="reference external" href="https://github.com/guberti/tvm-arduino-demos/tree/master/examples/person_detection">on GitHub</a>. Alternatively, we could also
 <a class="reference external" href="https://tvm.apache.org/docs/how_to/work_with_microtvm/micro_autotune.html">use TVM’s autotuning capabilities</a> to dramatically improve the model’s performance.</p>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes  42.129 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes  50.868 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-train-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/b52cec46baf4f78d6bcd94cbe269c8a6/micro_train.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">micro_train.py</span></code></a></p>
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 8880d52b3f..b7ef8a40f0 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>06:51.708</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>07:00.828</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 83%" />
@@ -349,27 +349,27 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><a class="reference internal" href="micro_train.html#sphx-glr-how-to-work-with-microtvm-micro-train-py"><span class="std std-ref">Training Vision Models for microTVM on Arduino</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_train.py</span></code>)</p></td>
-<td><p>04:42.129</p></td>
+<td><p>04:50.868</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="micro_pytorch.html#sphx-glr-how-to-work-with-microtvm-micro-pytorch-py"><span class="std std-ref">microTVM PyTorch Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_pytorch.py</span></code>)</p></td>
-<td><p>01:05.341</p></td>
+<td><p>01:05.355</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:51.638</p></td>
+<td><p>00:51.852</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="micro_aot.html#sphx-glr-how-to-work-with-microtvm-micro-aot-py"><span class="std std-ref">microTVM Host-Driven AoT</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_aot.py</span></code>)</p></td>
-<td><p>00:08.628</p></td>
+<td><p>00:08.831</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:03.969</p></td>
+<td><p>00:03.920</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:00.001</p></td>
+<td><p>00:00.002</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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 with CMSIS-NN</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></td>
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 ab4df7f4d4..0165412ae1 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>00:44.784</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:44.968</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -349,15 +349,15 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><a class="reference internal" href="using_pipeline_executor.html#sphx-glr-how-to-work-with-relay-using-pipeline-executor-py"><span class="std std-ref">Using Pipeline Executor in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_pipeline_executor.py</span></code>)</p></td>
-<td><p>00:32.784</p></td>
+<td><p>00:33.233</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:10.272</p></td>
+<td><p>00:10.284</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:01.721</p></td>
+<td><p>00:01.444</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
diff --git a/docs/how_to/work_with_schedules/intrin_math.html b/docs/how_to/work_with_schedules/intrin_math.html
index 4d5b751c11..60a59b8785 100644
--- a/docs/how_to/work_with_schedules/intrin_math.html
+++ b/docs/how_to/work_with_schedules/intrin_math.html
@@ -535,7 +535,7 @@ The following example customizes CUDA lowering rule for <code class="code docuti
 <a href="../../reference/api/python/ir.html#tvm.ir.register_intrin_lowering" title="tvm.ir.register_intrin_lowering" class="sphx-glr-backref-module-tvm-ir sphx-glr-backref-type-py-function"><span class="n">register_intrin_lowering</span></a><span class="p">(</span><span class="s2">&quot;tir.exp&quot;</span><span class="p">,</span> <span class="n">target</span><span class="o">=</span><span class="s2">&quot;cuda&quot;</span><span class="p">,</span> <span class="n">f</span><span class="o">= [...]
 </pre></div>
 </div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&lt;function my_cuda_math_rule at 0x7f330c705e60&gt;
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&lt;function my_cuda_math_rule at 0x7f15942b48c0&gt;
 </pre></div>
 </div>
 <p>Register the rule to TVM with override option to override existing rule.
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 04c12e1015..9d393ef996 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -340,7 +340,7 @@
             
   <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>00:06.448</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:06.163</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 83%" />
@@ -349,19 +349,19 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:03.974</p></td>
+<td><p>00:03.763</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:01.086</p></td>
+<td><p>00:01.058</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
-<td><p>00:00.594</p></td>
+<td><p>00:00.571</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:00.577</p></td>
+<td><p>00:00.553</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
@@ -369,7 +369,7 @@
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><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></td>
-<td><p>00:00.050</p></td>
+<td><p>00:00.051</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><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></td>
diff --git a/docs/how_to/work_with_schedules/tensorize.html b/docs/how_to/work_with_schedules/tensorize.html
index f28b0c0f34..b940eab8ca 100644
--- a/docs/how_to/work_with_schedules/tensorize.html
+++ b/docs/how_to/work_with_schedules/tensorize.html
@@ -586,7 +586,7 @@ The importing needs to happen before the tensorized GEMV being executed.</p>
              B: Buffer(B_2: Pointer(float32), float32, [512, 64], []),
              C: Buffer(C_2: Pointer(float32), float32, [1024, 512], [])}
   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/tmp8j_p51zl/input0.cc&#39;\nsource_filename = \&quot;/tmp/tmp8j_p51zl/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/tmpkzqa97qs/input0.cc&#39;\nsource_filename = \&quot;/tmp/tmpkzqa97qs/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/nn_2pooling_8h_source.html b/docs/reference/api/doxygen/nn_2pooling_8h_source.html
index 45ccd960e9..2c1afc1fe6 100644
--- a/docs/reference/api/doxygen/nn_2pooling_8h_source.html
+++ b/docs/reference/api/doxygen/nn_2pooling_8h_source.html
@@ -66,9 +66,9 @@ $(function() {
 <div class="title">pooling.h</div>  </div>
 </div><!--header-->
 <div class="contents">
-<a href="nn_2pooling_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 [...]
+<a href="nn_2pooling_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 [...]
 <div class="ttc" id="namespacetvm_1_1topi_html_a938350880b154670bea57cd9ce69d490"><div class="ttname"><a href="namespacetvm_1_1topi.html#a938350880b154670bea57cd9ce69d490">tvm::topi::kCommReduceIdx</a></div><div class="ttdeci">constexpr auto kCommReduceIdx</div><div class="ttdef"><b>Definition:</b> tags.h:35</div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_aca7c280684bfa7f8eb16a4a2ae0891f4"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#aca7c280684bfa7f8eb16a4a2ae0891f4">tvm::topi::nn::pool1d</a></div><div class="ttdeci">Tensor pool1d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_aca7c280684bfa7f8eb16a4a2ae0891f4"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#aca7c280684bfa7f8eb16a4a2ae0891f4">tvm::topi::nn::pool1d</a></div><div class="ttdeci">Tensor pool1d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
 <div class="ttc" id="namespacetvm_html_a3b37fa55ea93d6868751a2441996b072"><div class="ttname"><a href="namespacetvm.html#a3b37fa55ea93d6868751a2441996b072">tvm::min_value</a></div><div class="ttdeci">PrimExpr min_value(const DataType &amp;dtype, Span span=Span())</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_html_a4b434e701bc9835e2a7de8f0fadebea5"><div class="ttname"><a href="namespacetvm_1_1topi.html#a4b434e701bc9835e2a7de8f0fadebea5">tvm::topi::MakeArgmaxReducer</a></div><div class="ttdeci">FCommReduce MakeArgmaxReducer(bool select_last_index=false)</div><div class="ttdef"><b>Definition:</b> reduction.h:495</div></div>
 <div class="ttc" id="namespacetvm_html_ada5ad8338d3144221d8f16380e6c4855"><div class="ttname"><a href="namespacetvm.html#ada5ad8338d3144221d8f16380e6c4855">tvm::indexmod</a></div><div class="ttdeci">PrimExpr indexmod(PrimExpr a, PrimExpr b, Span span=Span())</div><div class="ttdoc">compute the remainder floor(a / b) where a and b are non-negative. </div></div>
@@ -80,7 +80,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1tir_1_1Var_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></div><div class="ttdoc">a named variable in TIR </div><div class="ttdef"><b>Definition:</b> var.h:88</div></div>
 <div class="ttc" id="namespacetvm_html_a353217978feabae3575560bf1586885f"><div class="ttname"><a href="namespacetvm.html#a353217978feabae3575560bf1586885f">tvm::if_then_else</a></div><div class="ttdeci">PrimExpr if_then_else(PrimExpr cond, PrimExpr true_value, PrimExpr false_value, Span span=Span())</div><div class="ttdoc">Conditional expression. </div></div>
 <div class="ttc" id="analyzer_8h_html"><div class="ttname"><a href="analyzer_8h.html">analyzer.h</a></div><div class="ttdoc">Algebra expression simplifications. </div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_ae5bacf7e4d7aa499e4c8420d39d7ad7e"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#ae5bacf7e4d7aa499e4c8420d39d7ad7e">tvm::topi::nn::pool3d</a></div><div class="ttdeci">Tensor pool3d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_ae5bacf7e4d7aa499e4c8420d39d7ad7e"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#ae5bacf7e4d7aa499e4c8420d39d7ad7e">tvm::topi::nn::pool3d</a></div><div class="ttdeci">Tensor pool3d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
 <div class="ttc" id="reduction_8h_html"><div class="ttname"><a href="reduction_8h.html">reduction.h</a></div><div class="ttdoc">Reduction op constructors. </div></div>
 <div class="ttc" id="classtvm_1_1arith_1_1Analyzer_html_a9b440f852f12ad0a4d8ed5ed97054425"><div class="ttname"><a href="classtvm_1_1arith_1_1Analyzer.html#a9b440f852f12ad0a4d8ed5ed97054425">tvm::arith::Analyzer::Simplify</a></div><div class="ttdeci">PrimExpr Simplify(const PrimExpr &amp;expr, int steps=2)</div><div class="ttdoc">Simplify expr. </div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a3ffa0974d8cdcd5b8ca7afb3cfbaf53c"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a3ffa0974d8cdcd5b8ca7afb3cfbaf53c">tvm::topi::nn::PoolType</a></div><div class="ttdeci">PoolType</div><div class="ttdoc">Pooling type. </div><div class="ttdef"><b>Definition:</b> pooling.h:44</div></div>
@@ -94,12 +94,13 @@ $(function() {
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a3ffa0974d8cdcd5b8ca7afb3cfbaf53ca9780bd19bf6706355258ca09ddeab335"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a3ffa0974d8cdcd5b8ca7afb3cfbaf53ca9780bd19bf6706355258ca09ddeab335">tvm::topi::nn::kAvgPool</a></div><div class="ttdef"><b>Definition:</b> pooling.h:45</div></div>
 <div class="ttc" id="namespacetvm_html_afdad0c0329bd39949ba8d296cfb85d76"><div class="ttname"><a href="namespacetvm.html#afdad0c0329bd39949ba8d296cfb85d76">tvm::sum</a></div><div class="ttdeci">PrimExpr sum(PrimExpr source, Array&lt; tir::IterVar &gt; axis, Array&lt; PrimExpr &gt; init={}, Span span=Span())</div><div class="ttdoc">sum of source expression over axis </div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Array_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html">tvm::runtime::Array</a></div><div class="ttdoc">Array, container representing a contiguous sequence of ObjectRefs. </div><div class="ttdef"><b>Definition:</b> array.h:289</div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a27571804c2096b32ab05e7b3e32c5af6"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a27571804c2096b32ab05e7b3e32c5af6">tvm::topi::nn::pool_impl_nd</a></div><div class="ttdeci">Tensor pool_impl_nd(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_t [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a27571804c2096b32ab05e7b3e32c5af6"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a27571804c2096b32ab05e7b3e32c5af6">tvm::topi::nn::pool_impl_nd</a></div><div class="ttdeci">Tensor pool_impl_nd(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_t [...]
 <div class="ttc" id="pad__utils_8h_html"><div class="ttname"><a href="pad__utils_8h.html">pad_utils.h</a></div><div class="ttdoc">Padding helpers. </div></div>
 <div class="ttc" id="namespacetvm_1_1topi_html_a0250c4095f19ae8a22ed85bc4ce5a40d"><div class="ttname"><a href="namespacetvm_1_1topi.html#a0250c4095f19ae8a22ed85bc4ce5a40d">tvm::topi::kElementWise</a></div><div class="ttdeci">constexpr auto kElementWise</div><div class="ttdef"><b>Definition:</b> tags.h:32</div></div>
 <div class="ttc" id="namespacetvm_html_a8f30aa0685ca52f846843e76a1ad1dc7"><div class="ttname"><a href="namespacetvm.html#a8f30aa0685ca52f846843e76a1ad1dc7">tvm::indexdiv</a></div><div class="ttdeci">PrimExpr indexdiv(PrimExpr a, PrimExpr b, Span span=Span())</div><div class="ttdoc">compute floor(a / b) where a and b are non-negative. </div></div>
 <div class="ttc" id="namespacetvm_html_a0df5ca82d2c566f628ebb2f1e84a3fcb"><div class="ttname"><a href="namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb">tvm::max</a></div><div class="ttdeci">PrimExpr max(PrimExpr a, PrimExpr b, Span span=Span())</div><div class="ttdoc">take maximum of two values </div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_af2bcf6bad18e028ce14ef854974629c0"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#af2bcf6bad18e028ce14ef854974629c0">tvm::topi::nn::global_pool</a></div><div class="ttdeci">Tensor global_pool(const Tensor &amp;x, PoolType pool_type, const std::string &amp;layout=&quot;NCHW&quot;)</div><div class="ttdoc">Perform global pooling on height and width dimension of data. It decides the height and width dimensi...</div><div class [...]
+<div class="ttc" id="classtvm_1_1runtime_1_1String_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1String.html">tvm::runtime::String</a></div><div class="ttdoc">Reference to string objects. </div><div class="ttdef"><b>Definition:</b> string.h:97</div></div>
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_af2bcf6bad18e028ce14ef854974629c0"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#af2bcf6bad18e028ce14ef854974629c0">tvm::topi::nn::global_pool</a></div><div class="ttdeci">Tensor global_pool(const Tensor &amp;x, PoolType pool_type, const std::string &amp;layout=&quot;NCHW&quot;)</div><div class="ttdoc">Perform global pooling on height and width dimension of data. It decides the height and width dimensi...</div><div class [...]
 <div class="ttc" id="namespacetvm_1_1te_html_aae384e9b73c2271905486e4a74b69265"><div class="ttname"><a href="namespacetvm_1_1te.html#aae384e9b73c2271905486e4a74b69265">tvm::te::reduce_axis</a></div><div class="ttdeci">IterVar reduce_axis(Range dom, std::string name=&quot;rv&quot;)</div><div class="ttdoc">Create a new IterVar for reduction operations. </div></div>
 <div class="ttc" id="namespacetvm_1_1topi_html_a3305d377f96cd20c23032eeada2756d5"><div class="ttname"><a href="namespacetvm_1_1topi.html#a3305d377f96cd20c23032eeada2756d5">tvm::topi::pad</a></div><div class="ttdeci">tvm::te::Tensor 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=kElem [...]
 <div class="ttc" id="namespacetvm_1_1topi_html_a59e9137c0e02da820906a44c6c9616b7"><div class="ttname"><a href="namespacetvm_1_1topi.html#a59e9137c0e02da820906a44c6c9616b7">tvm::topi::argmax</a></div><div class="ttdeci">Tensor argmax(const Tensor &amp;data, const Array&lt; Integer &gt; &amp;axis, bool keepdims=false, bool atleast1d=false, bool select_last_index=false)</div><div class="ttdoc">Creates an operation that finds the indices of the maximum values over a given axis. </div><div cl [...]
@@ -108,19 +109,21 @@ $(function() {
 <div class="ttc" id="classtvm_1_1te_1_1Tensor_html"><div class="ttname"><a href="classtvm_1_1te_1_1Tensor.html">tvm::te::Tensor</a></div><div class="ttdoc">Tensor structure representing a possible input, or intermediate computation result. </div><div class="ttdef"><b>Definition:</b> tensor.h:102</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Array_html_ae9734c8e6324f9be27fce8d45d0aa6f4"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#ae9734c8e6324f9be27fce8d45d0aa6f4">tvm::runtime::Array::begin</a></div><div class="ttdeci">iterator begin() const</div><div class="ttdef"><b>Definition:</b> array.h:387</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a8d244c196a5a481640cfc610fad3c7db"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a8d244c196a5a481640cfc610fad3c7db">tvm::topi::nn::end_index</a></div><div class="ttdeci">PrimExpr end_index(const Var &amp;out_index, const PrimExpr &amp;odim, const PrimExpr &amp;idim)</div><div class="ttdef"><b>Definition:</b> pooling.h:307</div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a9db26746e71db4065a913ff8046fd2fa"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a9db26746e71db4065a913ff8046fd2fa">tvm::topi::nn::pool2d</a></div><div class="ttdeci">Tensor pool2d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a9db26746e71db4065a913ff8046fd2fa"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a9db26746e71db4065a913ff8046fd2fa">tvm::topi::nn::pool2d</a></div><div class="ttdeci">Tensor pool2d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;dilation_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ce [...]
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a752c4130dac73fd2de0390c5f6b24b15"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a752c4130dac73fd2de0390c5f6b24b15">tvm::topi::nn::start_index</a></div><div class="ttdeci">PrimExpr start_index(const Var &amp;out_index, const PrimExpr &amp;odim, const PrimExpr &amp;idim)</div><div class="ttdef"><b>Definition:</b> pooling.h:303</div></div>
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a66fdbc05d6f8f5fade4e14ca1dbc686a"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a66fdbc05d6f8f5fade4e14ca1dbc686a">tvm::topi::nn::adaptive_pool3d</a></div><div class="ttdeci">Tensor adaptive_pool3d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCDHW&quot;)</div><div class="ttdoc">Adaptively perform pooling on three dimensional data. See the t [...]
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_ac4178bea4e2910c3ddefb2ef1ebe8575"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#ac4178bea4e2910c3ddefb2ef1ebe8575">tvm::topi::nn::adaptive_pool1d</a></div><div class="ttdeci">Tensor adaptive_pool1d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCW&quot;)</div><div class="ttdoc">Adaptively perform pooling on one dimensional data. See the two d [...]
-<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a150f883cae332ed85d56522da218ad73"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a150f883cae332ed85d56522da218ad73">tvm::topi::nn::adaptive_pool</a></div><div class="ttdeci">Tensor adaptive_pool(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCHW&quot;)</div><div class="ttdoc">Adaptively perform pooling on height and width dimension of data. Th [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a66fdbc05d6f8f5fade4e14ca1dbc686a"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a66fdbc05d6f8f5fade4e14ca1dbc686a">tvm::topi::nn::adaptive_pool3d</a></div><div class="ttdeci">Tensor adaptive_pool3d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCDHW&quot;)</div><div class="ttdoc">Adaptively perform pooling on three dimensional data. See the t [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_ac4178bea4e2910c3ddefb2ef1ebe8575"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#ac4178bea4e2910c3ddefb2ef1ebe8575">tvm::topi::nn::adaptive_pool1d</a></div><div class="ttdeci">Tensor adaptive_pool1d(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCW&quot;)</div><div class="ttdoc">Adaptively perform pooling on one dimensional data. See the two d [...]
+<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a150f883cae332ed85d56522da218ad73"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a150f883cae332ed85d56522da218ad73">tvm::topi::nn::adaptive_pool</a></div><div class="ttdeci">Tensor adaptive_pool(const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;output_size, PoolType pool_type, const std::string &amp;layout=&quot;NCHW&quot;)</div><div class="ttdoc">Adaptively perform pooling on height and width dimension of data. Th [...]
 <div class="ttc" id="classtvm_1_1tir_1_1Select_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Select.html">tvm::tir::Select</a></div><div class="ttdoc">Managed reference to SelectNode. </div><div class="ttdef"><b>Definition:</b> expr.h:589</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a7bb635a3241be1593258b374e651e344"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a7bb635a3241be1593258b374e651e344">tvm::topi::nn::pool_grad</a></div><div class="ttdeci">Tensor pool_grad(const Tensor &amp;out_grad, const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ceil_mode, const s [...]
+<div class="ttc" id="classtvm_1_1runtime_1_1Map_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Map.html">tvm::runtime::Map</a></div><div class="ttdoc">Map container of NodeRef-&gt;NodeRef in DSL graph. Map implements copy on write semantics, which means map is mutable but copy will happen when array is referenced in more than two places. </div><div class="ttdef"><b>Definition:</b> map.h:1271</div></div>
 <div class="ttc" id="classtvm_1_1tir_1_1And_html"><div class="ttname"><a href="classtvm_1_1tir_1_1And.html">tvm::tir::And</a></div><div class="ttdoc">Managed reference to AndNode. </div><div class="ttdef"><b>Definition:</b> expr.h:465</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a2806be7f95eec10be2b3555a9094cee2"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a2806be7f95eec10be2b3555a9094cee2">tvm::topi::nn::pool_grad_impl</a></div><div class="ttdeci">Tensor pool_grad_impl(const Tensor &amp;out_grad, const Tensor &amp;x, const Array&lt; PrimExpr &gt; &amp;kernel_size, const Array&lt; PrimExpr &gt; &amp;stride_size, const Array&lt; PrimExpr &gt; &amp;padding_size, PoolType pool_type, bool ceil_mod [...]
 <div class="ttc" id="tags_8h_html"><div class="ttname"><a href="tags_8h.html">tags.h</a></div><div class="ttdoc">External function interface to rocBLAS libraries. </div></div>
 <div class="ttc" id="namespacetvm_1_1te_html_afe4f57aeb3dd5ae9c0b58135e14d67ca"><div class="ttname"><a href="namespacetvm_1_1te.html#afe4f57aeb3dd5ae9c0b58135e14d67ca">tvm::te::compute</a></div><div class="ttdeci">Tensor compute(Array&lt; PrimExpr &gt; shape, FCompute fcompute, std::string name=&quot;tensor&quot;, std::string tag=&quot;&quot;, Map&lt; String, ObjectRef &gt; attrs={})</div><div class="ttdoc">Construct a new tensor by computing over shape, using the computation rule: resul [...]
 <div class="ttc" id="namespacetvm_1_1topi_html_a7da4c96db87c1459a2b097b87afd811f"><div class="ttname"><a href="namespacetvm_1_1topi.html#a7da4c96db87c1459a2b097b87afd811f">tvm::topi::cast</a></div><div class="ttdeci">Tensor cast(const Tensor &amp;x, DataType type, std::string name=&quot;T_cast&quot;, std::string tag=kElementWise)</div><div class="ttdoc">Cast each element of x to the given type. If expr is scalar and type is a corresponding vector type...</div><div class="ttdef"><b>Defini [...]
 <div class="ttc" id="classtvm_1_1PrimExpr_html"><div class="ttname"><a href="classtvm_1_1PrimExpr.html">tvm::PrimExpr</a></div><div class="ttdoc">Reference to PrimExprNode. </div><div class="ttdef"><b>Definition:</b> expr.h:112</div></div>
+<div class="ttc" id="classtvm_1_1runtime_1_1Map_html_ad3a78d88e3a9292d11ce04ff2dfe0702"><div class="ttname"><a href="classtvm_1_1runtime_1_1Map.html#ad3a78d88e3a9292d11ce04ff2dfe0702">tvm::runtime::Map::Set</a></div><div class="ttdeci">void Set(const K &amp;key, const V &amp;value)</div><div class="ttdoc">set the Map. </div><div class="ttdef"><b>Definition:</b> map.h:1374</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_a3ffa0974d8cdcd5b8ca7afb3cfbaf53cae5769af06e46579729c6b5e2d0c70563"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#a3ffa0974d8cdcd5b8ca7afb3cfbaf53cae5769af06e46579729c6b5e2d0c70563">tvm::topi::nn::kMaxPool</a></div><div class="ttdef"><b>Definition:</b> pooling.h:46</div></div>
 <div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_ab03974ee6b2d02a4619de1a4d0b42891"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#ab03974ee6b2d02a4619de1a4d0b42891">tvm::topi::nn::find_depth_height_width</a></div><div class="ttdeci">bool find_depth_height_width(const std::string &amp;layout, int *depth_axis, int *height_axis, int *width_axis)</div><div class="ttdef"><b>Definition:</b> pooling.h:217</div></div>
 <div class="ttc" id="classtvm_1_1arith_1_1Analyzer_html"><div class="ttname"><a href="classtvm_1_1arith_1_1Analyzer.html">tvm::arith::Analyzer</a></div><div class="ttdoc">Analyzer that contains bunch of sub-analyzers. </div><div class="ttdef"><b>Definition:</b> analyzer.h:563</div></div>
diff --git a/docs/reference/api/python/auto_scheduler.html b/docs/reference/api/python/auto_scheduler.html
index 99337088f8..345a40f6c7 100644
--- a/docs/reference/api/python/auto_scheduler.html
+++ b/docs/reference/api/python/auto_scheduler.html
@@ -1615,7 +1615,7 @@ history states as starting point to perform Evolutionary Search).</p></li>
 
 <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>
@@ -1899,7 +1899,7 @@ Candidates:
 
 <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">
diff --git a/docs/reference/api/typedoc/classes/bytestreamreader.html b/docs/reference/api/typedoc/classes/bytestreamreader.html
index 6f2e1db736..5347023fa1 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/52739ef8c/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L42">rpc_server.ts:42</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L63">rpc_server.ts:63</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L49">rpc_server.ts:49</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L57">rpc_server.ts:57</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 932a65b88f..fe8fd7d77f 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/52739ef8c/web/src/memory.ts#L223">memory.ts:223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L208">memory.ts:208</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L312">memory.ts:312</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L284">memory.ts:284</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L388">memory.ts:388</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L376">memory.ts:376</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L267">memory.ts:267</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L243">memory.ts:243</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L321">memory.ts:321</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L252">memory.ts:252</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L359">memory.ts:359</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L342">memory.ts:342</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L350">memory.ts:350</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L326">memory.ts:326</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L363">memory.ts:363</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L346">memory.ts:346</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L334">memory.ts:334</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 70053c7b20..88d6b36384 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/52739ef8c/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L260">runtime.ts:260</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L258">runtime.ts:258</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L279">runtime.ts:279</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L270">runtime.ts:270</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 0228a9dc09..089fae6ef6 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/52739ef8c/web/src/runtime.ts#L202">runtime.ts:202</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L200">runtime.ts:200</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L198">runtime.ts:198</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L223">runtime.ts:223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L230">runtime.ts:230</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 4efbf7782f..41cfed1270 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/52739ef8c/web/src/environment.ts#L86">environment.ts:86</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/environment.ts#L70">environment.ts:70</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/environment.ts#L69">environment.ts:69</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/environment.ts#L78">environment.ts:78</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/environment.ts#L84">environment.ts:84</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/environment.ts#L105">environment.ts:105</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 8917a8decb..9982d2424c 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/52739ef8c/web/src/runtime.ts#L49">runtime.ts:49</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L46">runtime.ts:46</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L45">runtime.ts:45</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L44">runtime.ts:44</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L47">runtime.ts:47</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L76">runtime.ts:76</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L66">runtime.ts:66</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L84">runtime.ts:84</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L95">runtime.ts:95</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L72">runtime.ts:72</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 f3d2808f09..b9149e1ff9 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/52739ef8c/web/src/runtime.ts#L583">runtime.ts:583</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L579">runtime.ts:579</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L654">runtime.ts:654</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L597">runtime.ts:597</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L631">runtime.ts:631</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L644">runtime.ts:644</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L621">runtime.ts:621</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L609">runtime.ts:609</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 57589bd1f0..84cd0d15b6 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/52739ef8c/web/src/runtime.ts#L692">runtime.ts:692</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L684">runtime.ts:684</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L683">runtime.ts:683</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L932">runtime.ts:932</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L994">runtime.ts:994</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L924">runtime.ts:924</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L732">runtime.ts:732</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L952">runtime.ts:952</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L816">runtime.ts:816</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L1033">runtime.ts:1033</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L846">runtime.ts:846</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L750">runtime.ts:750</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L1013">runtime.ts:1013</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L789">runtime.ts:789</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L914">runtime.ts:914</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L1145">runtime.ts:1145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/runtime.ts#L1145">runtime.ts:1145</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/52739ef8c/web/src/runtime.ts#L740">runtime.ts:740</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L868">runtime.ts:868</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L857">runtime.ts:857</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L940">runtime.ts:940</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 ae08e57e27..d78d281462 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/52739ef8c/web/src/memory.ts#L40">memory.ts:40</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L32">memory.ts:32</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L33">memory.ts:33</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L154">memory.ts:154</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L90">memory.ts:90</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L97">memory.ts:97</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L74">memory.ts:74</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L81">memory.ts:81</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L104">memory.ts:104</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L132">memory.ts:132</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L145">memory.ts:145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L60">memory.ts:60</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L67">memory.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L53">memory.ts:53</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L114">memory.ts:114</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L124">memory.ts:124</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/memory.ts#L175">memory.ts:175</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 8a1ac06b6e..557cd58759 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/52739ef8c/web/src/runtime.ts#L504">runtime.ts:504</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L502">runtime.ts:502</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L516">runtime.ts:516</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L530">runtime.ts:530</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L561">runtime.ts:561</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 da03326591..0db79d7ba0 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/52739ef8c/web/src/runtime.ts#L304">runtime.ts:304</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L297">runtime.ts:297</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L293">runtime.ts:293</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L289">runtime.ts:289</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L291">runtime.ts:291</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L295">runtime.ts:295</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L370">runtime.ts:370</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L414">runtime.ts:414</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L355">runtime.ts:355</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L474">runtime.ts:474</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L443">runtime.ts:443</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 fd601e8faf..4af15b89da 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/52739ef8c/web/src/runtime.ts#L158">runtime.ts:158</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L157">runtime.ts:157</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L165">runtime.ts:165</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 46269058be..0c298a21bf 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/52739ef8c/web/src/rpc_server.ts#L92">rpc_server.ts:92</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L82">rpc_server.ts:82</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L78">rpc_server.ts:78</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L81">rpc_server.ts:81</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L79">rpc_server.ts:79</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L80">rpc_server.ts:80</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L77">rpc_server.ts:77</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 7b05c6b17c..b7266e6acb 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/52739ef8c/web/src/runtime.ts#L145">runtime.ts:145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L145">runtime.ts:145</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L143">runtime.ts:143</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 e8712071db..98c9e1db2c 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/52739ef8c/web/src/webgpu.ts#L57">webgpu.ts:57</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/webgpu.ts#L50">webgpu.ts:50</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/webgpu.ts#L51">webgpu.ts:51</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/webgpu.ts#L84">webgpu.ts:84</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/webgpu.ts#L170">webgpu.ts:170</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/webgpu.ts#L67">webgpu.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 7eb4380646..77bc55817a 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/52739ef8c/web/src/ctypes.ts#L220">ctypes.ts:220</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L216">ctypes.ts:216</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L214">ctypes.ts:214</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L218">ctypes.ts:218</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L226">ctypes.ts:226</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L221">ctypes.ts:221</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L219">ctypes.ts:219</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L223">ctypes.ts:223</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L227">ctypes.ts:227</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L222">ctypes.ts:222</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L228">ctypes.ts:228</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L217">ctypes.ts:217</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L224">ctypes.ts:224</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L225">ctypes.ts:225</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L215">ctypes.ts:215</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 088afaba00..5b2329b995 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/52739ef8c/web/src/runtime.ts#L676">runtime.ts:676</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L675">runtime.ts:675</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 4694291e2f..256745db8b 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/52739ef8c/web/src/runtime.ts#L242">runtime.ts:242</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L240">runtime.ts:240</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L243">runtime.ts:243</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/runtime.ts#L241">runtime.ts:241</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 3c9533760a..743b1f91e3 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/52739ef8c/web/src/rpc_server.ts#L27">rpc_server.ts:27</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L28">rpc_server.ts:28</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L29">rpc_server.ts:29</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L32">rpc_server.ts:32</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L31">rpc_server.ts:31</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/rpc_server.ts#L30">rpc_server.ts:30</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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 2dd5a43908..52263669cf 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/52739ef8c/web/src/ctypes.ts#L206">ctypes.ts:206</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L207">ctypes.ts:207</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/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/52739ef8c/web/src/ctypes.ts#L203">ctypes.ts:203</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L203">ctypes.ts:203</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -130,7 +130,7 @@
 					<div class="tsd-signature tsd-kind-icon">F64<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/52739ef8c/web/src/ctypes.ts#L204">ctypes.ts:204</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L204">ctypes.ts:204</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -140,7 +140,7 @@
 					<div class="tsd-signature tsd-kind-icon">I32<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/52739ef8c/web/src/ctypes.ts#L201">ctypes.ts:201</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L201">ctypes.ts:201</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -150,7 +150,7 @@
 					<div class="tsd-signature tsd-kind-icon">I64<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/52739ef8c/web/src/ctypes.ts#L202">ctypes.ts:202</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L202">ctypes.ts:202</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -160,7 +160,7 @@
 					<div class="tsd-signature tsd-kind-icon">TVMValue<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/52739ef8c/web/src/ctypes.ts#L205">ctypes.ts:205</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L205">ctypes.ts:205</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -170,7 +170,7 @@
 					<div class="tsd-signature tsd-kind-icon">U16<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/52739ef8c/web/src/ctypes.ts#L200">ctypes.ts:200</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L200">ctypes.ts:200</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -180,7 +180,7 @@
 					<div class="tsd-signature tsd-kind-icon">U8<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/52739ef8c/web/src/ctypes.ts#L199">ctypes.ts:199</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L199">ctypes.ts:199</a></li>
 						</ul>
 					</aside>
 				</section>
diff --git a/docs/reference/api/typedoc/index.html b/docs/reference/api/typedoc/index.html
index e23b580cb9..9fa01b8881 100644
--- a/docs/reference/api/typedoc/index.html
+++ b/docs/reference/api/typedoc/index.html
@@ -174,7 +174,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMArray<wbr>Alloc<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>shape<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, ndim<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">number</span>, dtypeCode<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">number</span>, dtypeBits<span class="tsd [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L112">ctypes.ts:112</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L112">ctypes.ts:112</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -238,7 +238,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMArray<wbr>Copy<wbr>From<wbr>Bytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>handle<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, data<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, nbytes<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">num [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L128">ctypes.ts:128</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L128">ctypes.ts:128</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -282,7 +282,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMArray<wbr>Copy<wbr>From<wbr>To<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>from<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, to<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, stream<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-sig [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L144">ctypes.ts:144</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L144">ctypes.ts:144</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -326,7 +326,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMArray<wbr>Copy<wbr>ToBytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>handle<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, data<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, nbytes<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">number</sp [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L136">ctypes.ts:136</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L136">ctypes.ts:136</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -370,7 +370,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMArray<wbr>Free<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>handle<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> =&gt; </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/52739ef8c/web/src/ctypes.ts#L121">ctypes.ts:121</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L121">ctypes.ts:121</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -406,7 +406,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMBackend<wbr>PackedCFunc<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>argValues<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, argCodes<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, nargs<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">number< [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L160">ctypes.ts:160</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L160">ctypes.ts:160</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -458,7 +458,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMCFunc<wbr>Set<wbr>Return<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>ret<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, value<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, typeCode<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signa [...]
 					<aside class="tsd-sources">
 						<ul>
-							<li>Defined in <a href="https://github.com/apache/tvm/blob/52739ef8c/web/src/ctypes.ts#L77">ctypes.ts:77</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/271ad4302/web/src/ctypes.ts#L77">ctypes.ts:77</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -506,7 +506,7 @@
 					<div class="tsd-signature tsd-kind-icon">FTVMCb<wbr>Arg<wbr>ToReturn<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>value<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a>, code<span class="tsd-signature-symbol">: </span><a href="index.html#pointer" class="tsd-signature-type">Pointer</a><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> =&gt; </span><span c [...]
 					<aside class="tsd-sources">
 						<ul>
... 1293 lines suppressed ...