You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by tq...@apache.org on 2023/03/08 08:38:25 UTC

[tvm-site] branch asf-site updated: deploying docs (apache/tvm@9d732d0faa170ba254b6decb8e69d5148d111c03)

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 4603b1407f deploying docs (apache/tvm@9d732d0faa170ba254b6decb8e69d5148d111c03)
4603b1407f is described below

commit 4603b1407f8b81a13775bf10f6dd032ff332ed00
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Wed Mar 8 08:38:17 2023 +0000

    deploying docs (apache/tvm@9d732d0faa170ba254b6decb8e69d5148d111c03)
---
 docs/_images/sphx_glr_micro_train_001.png          |  Bin 309479 -> 364844 bytes
 docs/_images/sphx_glr_micro_train_thumb.png        |  Bin 23348 -> 24068 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_adreno.rst.txt   |    7 +-
 .../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       |   20 +-
 .../extend_tvm/bring_your_own_datatypes.rst.txt    |    2 +-
 .../how_to/extend_tvm/sg_execution_times.rst.txt   |   10 +-
 .../how_to/extend_tvm/use_pass_instrument.rst.txt  |   16 +-
 .../optimize_operators/opt_conv_cuda.rst.txt       |    2 +-
 .../optimize_operators/opt_conv_tensorcore.rst.txt |    2 +-
 .../how_to/optimize_operators/opt_gemm.rst.txt     |   16 +-
 .../optimize_operators/sg_execution_times.rst.txt  |    8 +-
 .../sg_execution_times.rst.txt                     |   14 +-
 .../tune_conv2d_layer_cuda.rst.txt                 | 2006 +++++++++-----------
 .../tune_network_cuda.rst.txt                      |    4 +-
 .../tune_network_x86.rst.txt                       |    4 +-
 .../tune_sparse_x86.rst.txt                        |  139 +-
 .../tune_with_autotvm/sg_execution_times.rst.txt   |    8 +-
 .../tune_with_autotvm/tune_conv2d_cuda.rst.txt     |  198 +-
 .../work_with_microtvm/micro_autotune.rst.txt      |   18 +-
 .../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  |   12 +-
 .../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 |   16 +-
 .../tutorials/autotvm/sg_execution_times.rst.txt   |    4 +-
 .../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     |    6 +-
 docs/_sources/tutorial/autotvm_matmul_x86.rst.txt  |   20 +-
 docs/_sources/tutorial/autotvm_relay_x86.rst.txt   |   60 +-
 .../tutorial/cross_compilation_and_rpc.rst.txt     |    2 +-
 docs/_sources/tutorial/intro_topi.rst.txt          |    2 +-
 docs/_sources/tutorial/sg_execution_times.rst.txt  |   22 +-
 .../tutorial/tensor_expr_get_started.rst.txt       |   44 +-
 docs/commit_hash                                   |    2 +-
 docs/genindex.html                                 |    8 +-
 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       |   12 +-
 docs/how_to/compile_models/from_pytorch.html       |    8 +-
 docs/how_to/compile_models/from_tensorflow.html    |    2 +-
 docs/how_to/compile_models/sg_execution_times.html |   22 +-
 .../deploy_models/deploy_model_on_adreno.html      |    3 +-
 .../deploy_models/deploy_model_on_android.html     |    2 +-
 .../deploy_object_detection_pytorch.html           |   38 +-
 docs/how_to/deploy_models/deploy_prequantized.html |    8 +-
 .../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  |   39 +-
 docs/how_to/deploy_models/sg_execution_times.html  |   20 +-
 .../extend_tvm/bring_your_own_datatypes.html       |    2 +-
 docs/how_to/extend_tvm/sg_execution_times.html     |   10 +-
 docs/how_to/extend_tvm/use_pass_instrument.html    |   16 +-
 docs/how_to/optimize_operators/opt_conv_cuda.html  |    2 +-
 .../optimize_operators/opt_conv_tensorcore.html    |    2 +-
 docs/how_to/optimize_operators/opt_gemm.html       |   16 +-
 .../optimize_operators/sg_execution_times.html     |    8 +-
 .../sg_execution_times.html                        |   14 +-
 .../tune_conv2d_layer_cuda.html                    | 2006 +++++++++-----------
 .../tune_with_autoscheduler/tune_network_cuda.html |    4 +-
 .../tune_with_autoscheduler/tune_network_x86.html  |    4 +-
 .../tune_with_autoscheduler/tune_sparse_x86.html   |  139 +-
 .../tune_with_autotvm/sg_execution_times.html      |    8 +-
 .../how_to/tune_with_autotvm/tune_conv2d_cuda.html |  198 +-
 docs/how_to/work_with_microtvm/micro_autotune.html |   18 +-
 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     |   12 +-
 .../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    |   16 +-
 docs/objects.inv                                   |  Bin 24299 -> 24309 bytes
 .../classtvm_1_1tir_1_1ScheduleNode-members.html   |    2 +
 .../doxygen/classtvm_1_1tir_1_1ScheduleNode.html   |  126 ++
 ...lasstvm_1_1tir_1_1ScheduleNode__coll__graph.svg |    2 +-
 ...stvm_1_1tir_1_1ScheduleNode__inherit__graph.svg |    2 +-
 docs/reference/api/doxygen/database_8h_source.html |    2 +-
 docs/reference/api/doxygen/functions_func_r.html   |    8 +-
 docs/reference/api/doxygen/functions_func_u.html   |    2 +-
 docs/reference/api/doxygen/functions_r.html        |    8 +-
 docs/reference/api/doxygen/functions_s.html        |    2 +-
 docs/reference/api/doxygen/functions_t.html        |    4 +-
 docs/reference/api/doxygen/functions_v.html        |    2 +-
 .../api/doxygen/measure__candidate_8h_source.html  |    2 +-
 docs/reference/api/doxygen/postproc_8h_source.html |    2 +-
 .../api/doxygen/schedule__rule_8h_source.html      |    2 +-
 docs/reference/api/doxygen/search/all_11.js        |    2 +-
 docs/reference/api/doxygen/search/all_13.js        |    4 +-
 docs/reference/api/doxygen/search/all_14.js        |   14 +-
 docs/reference/api/doxygen/search/all_15.js        |    4 +-
 docs/reference/api/doxygen/search/all_16.js        |    4 +-
 docs/reference/api/doxygen/search/all_17.js        |    2 +-
 docs/reference/api/doxygen/search/all_7.js         |    2 +-
 docs/reference/api/doxygen/search/all_e.js         |    6 +-
 docs/reference/api/doxygen/search/all_f.js         |    2 +-
 docs/reference/api/doxygen/search/functions_12.js  |    4 +-
 docs/reference/api/doxygen/search/functions_14.js  |    2 +-
 docs/reference/api/doxygen/search/functions_15.js  |    4 +-
 docs/reference/api/doxygen/search/functions_16.js  |    2 +-
 docs/reference/api/doxygen/search/functions_d.js   |    6 +-
 docs/reference/api/doxygen/search/functions_e.js   |    2 +-
 .../doxygen/tir_2schedule_2schedule_8h_source.html |    4 +-
 docs/reference/api/doxygen/trace_8h_source.html    |    2 +-
 docs/reference/api/python/auto_scheduler.html      |    4 +-
 docs/reference/api/python/tir.html                 |  148 +-
 .../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 +-
 docs/reference/api/typedoc/classes/instance.html   |   58 +-
 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 +-
 .../api/typedoc/classes/runtimecontext.html        |   22 +-
 docs/reference/api/typedoc/classes/scalar.html     |    6 +-
 docs/reference/api/typedoc/classes/tvmarray.html   |   16 +-
 docs/reference/api/typedoc/classes/tvmobject.html  |   12 +-
 .../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              |  124 +-
 .../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  |    4 +-
 .../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       |    5 +-
 docs/tutorial/autotvm_matmul_x86.html              |   20 +-
 docs/tutorial/autotvm_relay_x86.html               |  275 ++-
 docs/tutorial/cross_compilation_and_rpc.html       |    2 +-
 docs/tutorial/intro_topi.html                      |    2 +-
 docs/tutorial/sg_execution_times.html              |   22 +-
 docs/tutorial/tensor_expr_get_started.html         |   44 +-
 163 files changed, 3191 insertions(+), 3565 deletions(-)

diff --git a/docs/_images/sphx_glr_micro_train_001.png b/docs/_images/sphx_glr_micro_train_001.png
index e5bb8b4d46..2b438e4497 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 3e59e21850..d68a0ab65a 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 ef942659c4..57b6445031 100644
--- a/docs/_sources/how_to/compile_models/from_darknet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_darknet.rst.txt
@@ -318,7 +318,7 @@ The process is no different from other examples.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  20.865 seconds)
+   **Total running time of the script:** ( 1 minutes  19.259 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 f37d684f25..2cc7bdddf3 100644
--- a/docs/_sources/how_to/compile_models/from_keras.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_keras.rst.txt
@@ -232,7 +232,7 @@ 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 959ms/step
+
    1/1 [==============================] - ETA: 0s
    1/1 [==============================] - 1s 928ms/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 8ba7365b99..4bbdf88983 100644
--- a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
@@ -116,7 +116,7 @@ In this section, we download a pretrained imagenet model and classify an image.
 
  .. code-block:: none
 
-    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip0198aed7-cf6c-4692-b757-fb9d0313d96c from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+    Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip778ab4bb-0685-469e-bc2a-98cdd2d04766 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 f64e8222f3..47d9a44eba 100644
--- a/docs/_sources/how_to/compile_models/from_oneflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_oneflow.rst.txt
@@ -121,7 +121,7 @@ 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]
     23%|##3       | 9.66M/41.5M [00:00<00:00, 101MB/s]
     47%|####6     | 19.3M/41.5M [00:00<00:00, 75.2MB/s]
     65%|######4   | 26.9M/41.5M [00:00<00:00, 50.8MB/s]
     78%|#######8  | 32.5M/41.5M [00:00<00:00, 52.7MB/s]
     92%|#########2| 38.3M/41.5M [00:00<00:00, 45.6MB/s]
    100%|##########| 41.5M/41.5M [00:00<00:00, 52.6MB/s]
+
      0%|          | 0.00/41.5M [00:00<?, ?B/s]
     24%|##3       | 9.80M/41.5M [00:00<00:00, 103MB/s]
     47%|####7     | 19.6M/41.5M [00:00<00:00, 47.3MB/s]
     62%|######1   | 25.5M/41.5M [00:00<00:00, 50.8MB/s]
     77%|#######7  | 32.0M/41.5M [00:00<00:00, 46.9MB/s]
     89%|########9 | 37.1M/41.5M [00:00<00:00, 46.9MB/s]
    100%|##########| 41.5M/41.5M [00:00<00:00, 47.6MB/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 2369c9b18c..6531fe2df0 100644
--- a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
@@ -101,7 +101,7 @@ 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]
     30%|##9       | 13.4M/44.7M [00:00<00:00, 140MB/s]
     60%|#####9    | 26.8M/44.7M [00:00<00:00, 112MB/s]
     85%|########4 | 37.8M/44.7M [00:00<00:00, 108MB/s]
    100%|##########| 44.7M/44.7M [00:00<00:00, 101MB/s]
+
      0%|          | 0.00/44.7M [00:00<?, ?B/s]
     18%|#7        | 7.99M/44.7M [00:00<00:00, 70.8MB/s]
     54%|#####3    | 24.0M/44.7M [00:00<00:00, 98.9MB/s]
     81%|########  | 36.0M/44.7M [00:00<00:00, 109MB/s] 
    100%|##########| 44.7M/44.7M [00:00<00:00, 99.0MB/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 f05e525452..2b8e4ccbb5 100644
--- a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
@@ -424,7 +424,7 @@ Run the corresponding model on tensorflow
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  24.904 seconds)
+   **Total running time of the script:** ( 1 minutes  23.504 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 237cac9cac..3ede709bd5 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
 =================
-**06:42.702** total execution time for **how_to_compile_models** files:
+**06:34.745** total execution time for **how_to_compile_models** files:
 
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:24.904 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:23.504 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)       | 01:20.865 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)       | 01:19.259 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)         | 00:56.107 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)         | 00:54.967 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``)       | 00:38.249 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``)       | 00:37.468 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)         | 00:32.175 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)         | 00:31.655 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)           | 00:31.502 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)           | 00:30.554 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)         | 00:27.377 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)         | 00:27.004 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)       | 00:25.672 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)       | 00:25.776 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)           | 00:23.137 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)           | 00:21.869 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)             | 00:02.714 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)             | 00:02.689 | 0.0 MB |
 +-----------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt b/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt
index 5ee8d54055..611a7794df 100644
--- a/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt
@@ -727,13 +727,18 @@ well as provides information about the model's performance
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-     2545.8674    2543.4880    2563.2436    2540.1589      6.9047   
+     3336.0923    3336.1272    3339.9720    3332.7477      1.8435   
                
 
 
 
 
 
+.. rst-class:: sphx-glr-timing
+
+   **Total running time of the script:** ( 1 minutes  3.975 seconds)
+
+
 .. _sphx_glr_download_how_to_deploy_models_deploy_model_on_adreno.py:
 
 .. only:: html
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 d01e7d8111..653b61022d 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
@@ -437,7 +437,7 @@ Execute on TVM
     Evaluate inference time cost...
     Execution time summary:
      mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)  
-      16.1492      16.1049      16.6712      15.8424       0.2481   
+      16.4196      16.4974      16.9973      15.5352       0.3802   
                
 
 
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 5487fe72d2..896ab1355f 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
@@ -130,7 +130,7 @@ 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%|5         | 8.54M/170M [00:00<00:01, 89.5MB/s]
     13%|#2        | 21.6M/170M [00:00<00:01, 117MB/s] 
     19%|#9        | 32.8M/170M [00:00<00:01, 101MB/s]
     28%|##7       | 47.1M/170M [00:00<00:01, 118MB/s]
     35%|###4      | 58.6M/170M [00:00<00:01, 105MB/s]
     41%|####1     | 70.0M/170M [00:00<00:00, 109MB/s]
     48%|####7     | 80.7M/170M [00:00<00:00, 96.4MB/s]
     53%|#####3    | 90.3M/170M [00:00<00:00, 89.4MB/s]
     61%|######1   | 104M/170M [00:01<00:00, 101MB/s]  
     69%|######8   | 117M/170M [00:01<00:00, 110MB/s]
     75%|#######5  | 128M/170M [00:01<00:00, 107MB/s]
     81%|########1 | 138M/170M [00:01<00:00, 105MB/s]
     87%|########7 | 148M/170M [00:01<00:00, 103MB/s]
     93%|#########3| 158M/170M [00:01<00:00, 102MB/s]
     99%|#########8| 168M/170M [00:01<00:00, 76.6MB/s]
    100%|##########| 170M/170M [00:01<00:00, 97.3MB/s]
+
      0%|          | 0.00/170M [00:00<?, ?B/s]
      5%|4         | 7.99M/170M [00:00<00:03, 53.5MB/s]
      9%|9         | 16.1M/170M [00:00<00:02, 68.4MB/s]
     14%|#3        | 23.3M/170M [00:00<00:02, 71.2MB/s]
     19%|#8        | 32.0M/170M [00:00<00:01, 73.3MB/s]
     27%|##6       | 45.3M/170M [00:00<00:01, 95.1MB/s]
     33%|###2      | 56.0M/170M [00:00<00:01, 89.9MB/s]
     38%|###8      | 64.8M/170M [00:00<00:01, 69.9MB/s]
     46%|####6     | 78.1M/170M [00:01<00:01, 86.7MB/s]
     51%|#####1    | 87.4M/170M [00:01<00:01, 84.6MB/s]
     57%|#####6    | 96.1M/170M [00:01<00:01, 69.9MB/s]
     61%|######1   | 104M/170M [00:01<00:00, 70.9MB/s] 
     66%|######5   | 112M/170M [00:01<00:00, 74.1MB/s]
     70%|#######   | 120M/170M [00:01<00:00, 74.7MB/s]
     75%|#######4  | 127M/170M [00:01<00:00, 61.1MB/s]
     79%|#######8  | 133M/170M [00:01<00:00, 61.9MB/s]
     83%|########2 | 140M/170M [00:02<00:00, 57.1MB/s]
     87%|########6 | 148M/170M [00:02<00:00, 61.9MB/s]
   
   92%|#########2| 156M/170M [00:02<00:00, 69.4MB/s]
     96%|#########6| 163M/170M [00:02<00:00, 63.8MB/s]
    100%|##########| 170M/170M [00:02<00:00, 70.8MB/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').
@@ -299,7 +299,7 @@ Get boxes with score larger than 0.9
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 3 minutes  37.492 seconds)
+   **Total running time of the script:** ( 3 minutes  33.293 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 47b32fff06..b6ec026199 100644
--- a/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
@@ -227,7 +227,7 @@ 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]
     72%|#######2  | 9.79M/13.6M [00:00<00:00, 103MB/s]
    100%|##########| 13.6M/13.6M [00:00<00:00, 118MB/s]
+
      0%|          | 0.00/13.6M [00:00<?, ?B/s]
     59%|#####8    | 7.99M/13.6M [00:00<00:00, 63.8MB/s]
    100%|##########| 13.6M/13.6M [00:00<00:00, 90.4MB/s]
 
 
 
@@ -409,7 +409,7 @@ 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)  
-      90.5422      90.5458      91.3528      90.1322       0.2162   
+      90.4924      90.1736      105.9132     89.9654       1.6432   
                
 
 
@@ -458,7 +458,7 @@ TODO
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  17.475 seconds)
+   **Total running time of the script:** ( 1 minutes  16.438 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 d66f109654..6a60ab3a3d 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
@@ -423,7 +423,7 @@ 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)  
-      120.6696     120.5876     128.2718     119.6104      0.9786   
+      121.1309     121.1470     121.9582     120.3322      0.3616   
                
 
 
@@ -460,7 +460,7 @@ 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  31.722 seconds)
+   **Total running time of the script:** ( 2 minutes  36.366 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 bfbb146040..78819dabce 100644
--- a/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
@@ -257,7 +257,7 @@ We create a Relay VM to build and execute the model.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  34.853 seconds)
+   **Total running time of the script:** ( 1 minutes  31.299 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 8f9f93124f..34c98074ea 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
@@ -170,7 +170,7 @@ 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]
      4%|3         | 5000/132723 [00:00<00:02, 49996.95KB/s]
      9%|9         | 12265/132723 [00:00<00:01, 63320.20KB/s]
     15%|#4        | 19793/132723 [00:00<00:01, 68777.13KB/s]
     20%|##        | 27158/132723 [00:00<00:01, 70698.74KB/s]
     26%|##5       | 34410/132723 [00:00<00:01, 71321.28KB/s]
     31%|###1      | 41754/132723 [00:00<00:01, 72038.61KB/s]
     38%|###7      | 49792/132723 [00:00<00:01, 74762.70KB/s]
     44%|####3     | 57759/132723 [00:00<00:00, 76321.19KB/s]
     49%|####9     | 65688/132723 [00:00<00:00, 77247.43KB/s]
     55%|#####5    | 73613/132723 [00:01<00:00, 77861.54KB/s]
     61%|######1   | 81568/132723 [00:01<00:00, 78376.65KB/s]
     67%|######7   | 89487/132723 [00:01<00:00, 78615.64KB/s]
     73%|#######3  | 97455/132723 [00:01<00:00, 78934.85KB/s]
     79%|#######9  | 105349/132723 [00:01<00:00, 78816.63KB/s]
     85%|########5 | 113231/132723 [00:01<00:00, 77704.92KB/s]
     91%|#########
 1| 121005/132723 [00:01<00:00, 76952.44KB/s]
     97%|#########6| 128704/132723 [00:01<00:00, 76767.77KB/s]
    100%|##########| 132723/132723 [00:01<00:00, 75349.98KB/s]
+
      0%|          | 0/132723 [00:00<?, ?KB/s]
      1%|          | 1287/132723 [00:00<00:11, 11746.58KB/s]
      5%|4         | 6075/132723 [00:00<00:03, 32195.06KB/s]
     10%|#         | 13313/132723 [00:00<00:02, 50151.41KB/s]
     16%|#5        | 21235/132723 [00:00<00:01, 61477.69KB/s]
     22%|##1       | 29162/132723 [00:00<00:01, 67838.64KB/s]
     28%|##7       | 37037/132723 [00:00<00:01, 71524.28KB/s]
     34%|###4      | 45141/132723 [00:00<00:01, 74619.84KB/s]
     40%|###9      | 52798/132723 [00:00<00:01, 75233.16KB/s]
     46%|####5     | 60637/132723 [00:00<00:00, 76215.30KB/s]
     52%|#####1    | 68393/132723 [00:01<00:00, 76624.85KB/s]
     57%|#####7    | 76148/132723 [00:01<00:00, 76905.73KB/s]
     63%|######3   | 83844/132723 [00:01<00:00, 76866.09KB/s]
     69%|######9   | 91671/132723 [00:01<00:00, 77289.32KB/s]
     75%|#######4  | 99403/132723 [00:01<00:00, 77274.47KB/s]
     81%|########  | 107265/132723 [00:01<00:00, 77674.88KB/s]
     87%|########6 |
  115129/132723 [00:01<00:00, 77964.58KB/s]
     93%|#########2| 122929/132723 [00:01<00:00, 77970.50KB/s]
     99%|#########8| 130795/132723 [00:01<00:00, 78176.64KB/s]
    100%|##########| 132723/132723 [00:01<00:00, 72366.44KB/s]
 
 
 
@@ -246,7 +246,7 @@ Display result
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 3 minutes  48.761 seconds)
+   **Total running time of the script:** ( 3 minutes  45.424 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 714cfd6f22..dbe0deeff7 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
@@ -5,26 +5,26 @@
 
 Computation times
 =================
-**15:24.431** total execution time for **how_to_deploy_models** files:
+**15:25.837** total execution time for **how_to_deploy_models** files:
 
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)                           | 03:48.761 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)                           | 03:45.424 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:37.492 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:33.293 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)           | 02:31.722 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)           | 02:36.366 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)                               | 01:34.853 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)                               | 01:31.299 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)                         | 01:17.475 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)                         | 01:16.438 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``)                   | 00:54.814 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``)                   | 01:03.975 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)                 | 00:43.000 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)                 | 00:42.309 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``)                       | 00:28.356 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``)                       | 00:28.488 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)                       | 00:27.951 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)                       | 00:28.239 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)                                     | 00:00.006 | 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 9b7ac496fa..e26508c805 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
@@ -463,7 +463,7 @@ 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.zip3e28c82f-ca1f-4e34-aaed-17f934dcf6d2 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+    Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip61858a39-05f3-4d4b-bcb1-53d26f89705d 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 05cb4db21e..67c39301ef 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
 =================
-**00:55.057** total execution time for **how_to_extend_tvm** files:
+**00:53.636** 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``) | 00:51.172 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:49.872 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)           | 00:02.774 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)           | 00:02.685 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)                     | 00:01.102 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)                     | 00:01.072 | 0.0 MB |
 +-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)       | 00:00.009 | 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 d9ee060d21..72b7269222 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
@@ -220,10 +220,10 @@ profile the execution time of each passes.
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 22302us [22302us] (48.68%; 48.68%)
-    FoldScaleAxis: 23510us [8us] (51.32%; 51.32%)
-            FoldConstant: 23501us [1725us] (51.30%; 99.96%)
-                    InferType: 21777us [21777us] (47.53%; 92.66%)
+    InferType: 22409us [22409us] (47.83%; 47.83%)
+    FoldScaleAxis: 24439us [7us] (52.17%; 52.17%)
+            FoldConstant: 24432us [1696us] (52.15%; 99.97%)
+                    InferType: 22736us [22736us] (48.53%; 93.06%)
 
 
 
@@ -262,10 +262,10 @@ Refer to following sections and :py:func:`tvm.instrument.pass_instrument` for th
  .. code-block:: none
 
     Printing results of timing profile...
-    InferType: 21885us [21885us] (48.30%; 48.30%)
-    FoldScaleAxis: 23427us [6us] (51.70%; 51.70%)
-            FoldConstant: 23421us [1749us] (51.69%; 99.97%)
-                    InferType: 21672us [21672us] (47.83%; 92.53%)
+    InferType: 21571us [21571us] (48.43%; 48.43%)
+    FoldScaleAxis: 22968us [5us] (51.57%; 51.57%)
+            FoldConstant: 22963us [1704us] (51.56%; 99.98%)
+                    InferType: 21259us [21259us] (47.73%; 92.58%)
 
 
 
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 b222ea2e90..c42f174833 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
@@ -331,7 +331,7 @@ latency of convolution.
 
  .. code-block:: none
 
-    Convolution: 54.161792 ms
+    Convolution: 54.239936 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 e13a39b8a9..16d7f0220b 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
@@ -598,7 +598,7 @@ be able to run on our build server
 
  .. code-block:: none
 
-    conv2d with tensor core: 7.498425 ms
+    conv2d with tensor core: 6.683037 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 8ebf498753..067727fe28 100644
--- a/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
@@ -134,8 +134,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
 
  .. code-block:: none
 
-    Numpy running time: 0.018414
-    Baseline: 3.427426
+    Numpy running time: 0.018453
+    Baseline: 3.409857
 
 
 
@@ -227,7 +227,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
 
  .. code-block:: none
 
-    Opt1: 0.303755
+    Opt1: 0.295121
 
 
 
@@ -318,7 +318,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
 
  .. code-block:: none
 
-    Opt2: 0.349998
+    Opt2: 0.332745
 
 
 
@@ -406,7 +406,7 @@ the access pattern for A matrix is more cache friendly.
 
  .. code-block:: none
 
-    Opt3: 0.115294
+    Opt3: 0.114450
 
 
 
@@ -523,7 +523,7 @@ flattening.
 
  .. code-block:: none
 
-    Opt4: 0.107893
+    Opt4: 0.109112
 
 
 
@@ -635,7 +635,7 @@ write to C when all the block results are ready.
 
  .. code-block:: none
 
-    Opt5: 0.110806
+    Opt5: 0.111139
 
 
 
@@ -748,7 +748,7 @@ Furthermore, we can also utilize multi-core processors to do the thread-level pa
 
  .. code-block:: none
 
-    Opt6: 0.146479
+    Opt6: 0.146767
 
 
 
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 a2d35bff31..7dda45d41c 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
 =================
-**00:35.107** total execution time for **how_to_optimize_operators** files:
+**00:34.737** total execution time for **how_to_optimize_operators** files:
 
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)                       | 00:32.496 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)                       | 00:32.165 | 0.0 MB |
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.499 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.500 | 0.0 MB |
 +-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)             | 00:01.112 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)             | 00:01.072 | 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 9118521517..c695b166b0 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
 =================
-**09:59.540** total execution time for **how_to_tune_with_autoscheduler** files:
+**09:57.787** 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``) | 06:06.168 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 06:08.646 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)             | 01:42.432 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)             | 01:41.629 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)           | 01:07.705 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)           | 01:07.266 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)               | 00:35.267 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)               | 00:32.690 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)             | 00:14.289 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)             | 00:14.002 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)           | 00:13.678 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)           | 00:13.554 | 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 43b7b76e63..cb7585f46d 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
@@ -244,11 +244,11 @@ cooperative fetching, unrolling and operator fusion.
         def main(data: T.Buffer((1, 512, 7, 7), "float32"), kernel: T.Buffer((512, 512, 3, 3), "float32"), bias: T.Buffer((1, 512, 1, 1), "float32"), compute: T.Buffer((1, 512, 7, 7), "float32")):
             T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
             blockIdx_x = T.launch_thread("blockIdx.x", 28)
-            conv2d_nchw = T.allocate([7], "float32", "local")
-            pad_temp_shared = T.allocate([216], "float32", "shared")
-            kernel_shared = T.allocate([9216], "float32", "shared")
-            threadIdx_x = T.launch_thread("threadIdx.x", 128)
-            conv2d_nchw_1 = T.Buffer((7,), data=conv2d_nchw, scope="local", align=16)
+            conv2d_nchw = T.allocate([14], "float32", "local")
+            pad_temp_shared = T.allocate([72], "float32", "shared")
+            kernel_shared = T.allocate([3072], "float32", "shared")
+            threadIdx_x = T.launch_thread("threadIdx.x", 64)
+            conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope="local", align=32)
             conv2d_nchw_1[0] = T.float32(0)
             conv2d_nchw_1[1] = T.float32(0)
             conv2d_nchw_1[2] = T.float32(0)
@@ -256,563 +256,466 @@ cooperative fetching, unrolling and operator fusion.
             conv2d_nchw_1[4] = T.float32(0)
             conv2d_nchw_1[5] = T.float32(0)
             conv2d_nchw_1[6] = T.float32(0)
-            for rc_outer_outer in range(64):
-                cse_var_1: T.int32 = rc_outer_outer * 72
+            conv2d_nchw_1[7] = T.float32(0)
+            conv2d_nchw_1[8] = T.float32(0)
+            conv2d_nchw_1[9] = T.float32(0)
+            conv2d_nchw_1[10] = T.float32(0)
+            conv2d_nchw_1[11] = T.float32(0)
+            conv2d_nchw_1[12] = T.float32(0)
+            conv2d_nchw_1[13] = T.float32(0)
+            for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
+                cse_var_2: T.int32 = rc_outer_outer * 72
+                cse_var_1: T.int32 = ry_outer_outer * 3
+                pad_temp_shared_1 = T.Buffer((72,), data=pad_temp_shared, scope="shared")
+                with T.launch_thread("threadIdx.x", 64) as threadIdx_x_1:
+                    data_1 = T.Buffer((25088,), data=data.data)
+                    if T.likely(threadIdx_x_1 < 18):
+                        pad_temp_shared_1[threadIdx_x_1 * 4] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= threadIdx_x_1 * 4 % 9 and threadIdx_x_1 * 4 % 9 < 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 * 4 // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + threadIdx_x_1 * 4 % 9 - 8], T.float32(0))
+                    if T.likely(threadIdx_x_1 < 18):
+                        pad_temp_shared_1[threadIdx_x_1 * 4 + 1] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 1) % 9 and (threadIdx_x_1 * 4 + 1) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 1) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 1) % 9 - 8], T.float32(0))
+                    if T.likely(threadIdx_x_1 < 18):
+                        pad_temp_shared_1[threadIdx_x_1 * 4 + 2] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 2) % 9 and (threadIdx_x_1 * 4 + 2) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 2) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 2) % 9 - 8], T.float32(0))
+                    if T.likely(threadIdx_x_1 < 18):
+                        pad_temp_shared_1[threadIdx_x_1 * 4 + 3] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 3) % 9 and (threadIdx_x_1 * 4 + 3) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 3) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 3) % 9 - 8], T.float32(0))
                 threadIdx_x_1 = T.env_thread("threadIdx.x")
-                pad_temp_shared_1 = T.Buffer((216,), data=pad_temp_shared, scope="shared")
-                data_1 = T.Buffer((25088,), data=data.data)
-                with T.launch_thread(threadIdx_x_1, 128):
-                    pad_temp_shared_1[threadIdx_x_1] = T.if_then_else(3 <= threadIdx_x_1 % 27 and threadIdx_x_1 % 27 < 24 and 1 <= blockIdx_x % 7 + threadIdx_x_1 % 3 and blockIdx_x % 7 + threadIdx_x_1 % 3 < 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 // 27 * 49 + threadIdx_x_1 % 27 // 3 * 7 + blockIdx_x % 7 + threadIdx_x_1 % 3 - 8], T.float32(0))
-                with T.launch_thread(threadIdx_x_1, 128):
-                    if T.likely(threadIdx_x_1 < 88):
-                        pad_temp_shared_1[threadIdx_x_1 + 128] = T.if_then_else(3 <= (threadIdx_x_1 + 20) % 27 and (threadIdx_x_1 + 20) % 27 < 24 and 1 <= blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 and blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 + 128) // 27 * 49 + (threadIdx_x_1 + 20) % 27 // 3 * 7 + blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 - 8], T.float32(0))
-                threadIdx_x_2 = T.env_thread("threadIdx.x")
-                kernel_shared_1 = T.Buffer((9216,), data=kernel_shared, scope="shared")
+                kernel_shared_1 = T.Buffer((3072,), data=kernel_shared, scope="shared")
                 kernel_1 = T.Buffer((2359296,), data=kernel.data)
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4:threadIdx_x_2 * 4 + 4] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4:blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 512:threadIdx_x_2 * 4 + 512 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 512) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 512) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 1024:threadIdx_x_2 * 4 + 1024 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1024) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1024) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 1536:threadIdx_x_2 * 4 + 1536 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1536) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1536) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 2048:threadIdx_x_2 * 4 + 2048 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2048) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2048) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 2560:threadIdx_x_2 * 4 + 2560 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2560) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2560) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 3072:threadIdx_x_2 * 4 + 3072 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3072) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3072) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 3584:threadIdx_x_2 * 4 + 3584 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3584) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3584) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 4096:threadIdx_x_2 * 4 + 4096 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 4096) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 4096) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 4608:threadIdx_x_2 * 4 + 4608 + 4] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 294912:blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 294912 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 5120:threadIdx_x_2 * 4 + 5120 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5120) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5120) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 5632:threadIdx_x_2 * 4 + 5632 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5632) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5632) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 6144:threadIdx_x_2 * 4 + 6144 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6144) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6144) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 6656:threadIdx_x_2 * 4 + 6656 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6656) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6656) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 7168:threadIdx_x_2 * 4 + 7168 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7168) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7168) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 7680:threadIdx_x_2 * 4 + 7680 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7680) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7680) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 8192:threadIdx_x_2 * 4 + 8192 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8192) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8192) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72 + 4]
-                with T.launch_thread(threadIdx_x_2, 128):
-                    kernel_shared_1[threadIdx_x_2 * 4 + 8704:threadIdx_x_2 * 4 + 8704 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8704) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8704) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72 + 4]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[81] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[82] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[83] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[84] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[85] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[86] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[84] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[85] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[86] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[75] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[76] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[77] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[102] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[103] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[104] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 1]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 2]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 72 + 3]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 72 + 4]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 72 + 5]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 72 + 6]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 72 + 7]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 72 + 8]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 9]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 10]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 11]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 72 + 12]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 72 + 13]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 72 + 14]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 72 + 15]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 72 + 16]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 72 + 17]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 18]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 19]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 20]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[75] * kernel_shared_1[threadIdx_x * 72 + 21]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[76] * kernel_shared_1[threadIdx_x * 72 + 22]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[77] * kernel_shared_1[threadIdx_x * 72 + 23]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[78] * kernel_shared_1[threadIdx_x * 72 + 24]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[79] * kernel_shared_1[threadIdx_x * 72 + 25]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[80] * kernel_shared_1[threadIdx_x * 72 + 26]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 27]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 28]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 29]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[102] * kernel_shared_1[threadIdx_x * 72 + 30]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[103] * kernel_shared_1[threadIdx_x * 72 + 31]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[104] * kernel_shared_1[threadIdx_x * 72 + 32]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[105] * kernel_shared_1[threadIdx_x * 72 + 33]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[106] * kernel_shared_1[threadIdx_x * 72 + 34]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[107] * kernel_shared_1[threadIdx_x * 72 + 35]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[108] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[109] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[110] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[111] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[112] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[113] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[135] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[136] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[137] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[138] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[139] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[140] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[162] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[163] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[164] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[165] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[166] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[167] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[189] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[190] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[191] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[192] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[193] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[194] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[111] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[112] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[113] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[138] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[139] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[140] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[165] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[166] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[167] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[192] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[193] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[194] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[129] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[130] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[131] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[156] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[157] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[158] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[183] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[184] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[185] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[210] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[211] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[212] * kernel_shared_1[threadIdx_x * 72 + 71]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 36]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 37]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 38]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[129] * kernel_shared_1[threadIdx_x * 72 + 39]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[130] * kernel_shared_1[threadIdx_x * 72 + 40]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[131] * kernel_shared_1[threadIdx_x * 72 + 41]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[132] * kernel_shared_1[threadIdx_x * 72 + 42]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[133] * kernel_shared_1[threadIdx_x * 72 + 43]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[134] * kernel_shared_1[threadIdx_x * 72 + 44]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 45]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 46]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 47]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[156] * kernel_shared_1[threadIdx_x * 72 + 48]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[157] * kernel_shared_1[threadIdx_x * 72 + 49]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[158] * kernel_shared_1[threadIdx_x * 72 + 50]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[159] * kernel_shared_1[threadIdx_x * 72 + 51]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[160] * kernel_shared_1[threadIdx_x * 72 + 52]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[161] * kernel_shared_1[threadIdx_x * 72 + 53]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 54]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 55]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 56]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[183] * kernel_shared_1[threadIdx_x * 72 + 57]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[184] * kernel_shared_1[threadIdx_x * 72 + 58]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[185] * kernel_shared_1[threadIdx_x * 72 + 59]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[186] * kernel_shared_1[threadIdx_x * 72 + 60]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[187] * kernel_shared_1[threadIdx_x * 72 + 61]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[188] * kernel_shared_1[threadIdx_x * 72 + 62]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 63]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 64]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 65]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[210] * kernel_shared_1[threadIdx_x * 72 + 66]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[211] * kernel_shared_1[threadIdx_x * 72 + 67]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[212] * kernel_shared_1[threadIdx_x * 72 + 68]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[213] * kernel_shared_1[threadIdx_x * 72 + 69]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[214] * kernel_shared_1[threadIdx_x * 72 + 70]
-                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[215] * kernel_shared_1[threadIdx_x * 72 + 71]
-            for i2_inner in range(7):
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+                with T.launch_thread(threadIdx_x_1, 64):
+                    kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+                conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+                conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
+            for i1_inner, i3_inner in T.grid(2, 7):
                 compute_1 = T.Buffer((25088,), data=compute.data)
                 bias_1 = T.Buffer((512,), data=bias.data)
-                compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 49 + i2_inner * 7 + blockIdx_x % 7] = T.max(conv2d_nchw_1[i2_inner] + bias_1[blockIdx_x // 7 * 128 + threadIdx_x], T.float32(0))
+                compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 98 + i1_inner * 49 + blockIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x // 7 * 128 + threadIdx_x * 2 + i1_inner], T.float32(0))
 
 
 
@@ -862,7 +765,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 0.198 ms
+    Execution time of this operator: 0.364 ms
 
 
 
@@ -911,34 +814,34 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
     conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
     conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
     conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
-    conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
-    conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=128)
+    conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=2)
+    conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=64)
     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=7)
+    conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
     conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
     conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
     conv2d_nchw_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_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
     conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=1)
     conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
-    conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
-    conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=2)
-    conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=3)
+    conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=2)
+    conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=4)
+    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_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=3)
-    conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+    conv2d_nchw_rx_o_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=3)
     s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2 [...]
     compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
     compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
     compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-    compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
-    compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=128)
+    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=64)
     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_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
     compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
     compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
-    compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
+    compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=7)
     compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=1)
     compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
     s[compute].reorder(compute_i0_o_o_o, compute_i1_o_o_o, compute_i2_o_o_o, compute_i3_o_o_o, compute_i0_o_o_i, compute_i1_o_o_i, compute_i2_o_o_i, compute_i3_o_o_i, compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i, compute_i0_i, compute_i1_i, compute_i2_i, compute_i3_i)
@@ -957,16 +860,16 @@ 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=128)
+    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=64)
     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=4)
     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=128)
+    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=64)
     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:
@@ -984,10 +887,10 @@ 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__(128) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
-      float conv2d_nchw[7];
-      __shared__ float pad_temp_shared[216];
-      __shared__ float kernel_shared[9216];
+    extern "C" __global__ void __launch_bounds__(64) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+      float conv2d_nchw[14];
+      __shared__ float pad_temp_shared[72];
+      __shared__ float kernel_shared[3072];
       conv2d_nchw[0] = 0.000000e+00f;
       conv2d_nchw[1] = 0.000000e+00f;
       conv2d_nchw[2] = 0.000000e+00f;
@@ -995,538 +898,419 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
       conv2d_nchw[4] = 0.000000e+00f;
       conv2d_nchw[5] = 0.000000e+00f;
       conv2d_nchw[6] = 0.000000e+00f;
+      conv2d_nchw[7] = 0.000000e+00f;
+      conv2d_nchw[8] = 0.000000e+00f;
+      conv2d_nchw[9] = 0.000000e+00f;
+      conv2d_nchw[10] = 0.000000e+00f;
+      conv2d_nchw[11] = 0.000000e+00f;
+      conv2d_nchw[12] = 0.000000e+00f;
+      conv2d_nchw[13] = 0.000000e+00f;
       for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
-        __syncthreads();
-        pad_temp_shared[((int)threadIdx.x)] = (((((3 <= (((int)threadIdx.x) % 27)) && ((((int)threadIdx.x) % 27) < 24)) && (1 <= ((((int)blockIdx.x) % 7) + (((int)threadIdx.x) % 3)))) && (((((int)blockIdx.x) % 7) + (((int)threadIdx.x) % 3)) < 8)) ? data[((((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 27) * 49)) + (((((int)threadIdx.x) % 27) / 3) * 7)) + (((int)blockIdx.x) % 7)) + (((int)threadIdx.x) % 3)) - 8)] : 0.000000e+00f);
-        if (((int)threadIdx.x) < 88) {
-          pad_temp_shared[(((int)threadIdx.x) + 128)] = (((((3 <= ((((int)threadIdx.x) + 20) % 27)) && (((((int)threadIdx.x) + 20) % 27) < 24)) && (1 <= ((((int)blockIdx.x) % 7) + ((((int)threadIdx.x) + 2) % 3)))) && (((((int)blockIdx.x) % 7) + ((((int)threadIdx.x) + 2) % 3)) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 128) / 27) * 49)) + ((((((int)threadIdx.x) + 20) % 27) / 3) * 7)) + (((int)blockIdx.x) % 7)) + ((((int)threadIdx.x) + 2) % 3)) - 8)] : 0.000000e+00f);
+        for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
+          __syncthreads();
+          if (((int)threadIdx.x) < 18) {
+            pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) * 4) % 9))) && (((((int)threadIdx.x) * 4) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
+          }
+          if (((int)threadIdx.x) < 18) {
+            pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 1) % 9))) && ((((((int)threadIdx.x) * 4) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
+          }
+          if (((int)threadIdx.x) < 18) {
+            pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 2) % 9))) && ((((((int)threadIdx.x) * 4) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
+          }
+          if (((int)threadIdx.x) < 18) {
+            pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 3) % 9))) && ((((((int)threadIdx.x) * 4) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
+          }
+          kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 64) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 128) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
+          kernel_shared[(((((((int)threadIdx.x) + 256) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 320) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
+          kernel_shared[(((((((int)threadIdx.x) + 448) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 512) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
+          kernel_shared[(((((((int)threadIdx.x) + 640) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 704) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
+          kernel_shared[(((((((int)threadIdx.x) + 832) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 896) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
+          kernel_shared[(((((((int)threadIdx.x) + 1024) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 1088) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
+          kernel_shared[(((((((int)threadIdx.x) + 1216) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 1280) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
+          kernel_shared[(((((((int)threadIdx.x) + 1408) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 1472) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
+          kernel_shared[(((((((int)threadIdx.x) + 1600) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 1664) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
+          kernel_shared[(((((((int)threadIdx.x) + 1792) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 1856) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
+          kernel_shared[(((((((int)threadIdx.x) + 1984) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 2048) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
+          kernel_shared[(((((((int)threadIdx.x) + 2176) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 2240) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
+          kernel_shared[(((((((int)threadIdx.x) + 2368) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 2432) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
+          kernel_shared[(((((((int)threadIdx.x) + 2560) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 2624) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
+          kernel_shared[(((((((int)threadIdx.x) + 2752) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 2816) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
+          kernel_shared[(((((((int)threadIdx.x) + 2944) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+          kernel_shared[(((((((int)threadIdx.x) + 3008) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+          __syncthreads();
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[1] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[2] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[4] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[5] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 48)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[0] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+          conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+          conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
         }
-        *(float4*)(kernel_shared + (((int)threadIdx.x) * 4)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 18) * 4608)) + (rc_outer_outer * 72)) + ((((int)threadIdx.x) % 18) * 4)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 512)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 512) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 8) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 1024)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 1024) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 16) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 1536)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 1536) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 24) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 2048)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 2048) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 32) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 2560)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 2560) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 40) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 3072)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 3072) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 48) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 3584)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 3584) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 56) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 4096)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 4096) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 64) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 4608)) = *(float4*)(kernel + ((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 18) * 4608)) + (rc_outer_outer * 72)) + ((((int)threadIdx.x) % 18) * 4)) + 294912));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 5120)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 5120) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 8) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 5632)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 5632) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 16) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 6144)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 6144) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 24) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 6656)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 6656) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 32) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 7168)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 7168) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 40) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 7680)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 7680) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 48) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 8192)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 8192) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 56) % 72)));
-        *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 8704)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 8704) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 64) % 72)));
-        __syncthreads();
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[81] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[82] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[83] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[84] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[85] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[86] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[84] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[85] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[86] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[9] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[12] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[75] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[76] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[77] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[102] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[103] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[104] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[18] * kernel_shared[(((int)threadIdx.x) * 72)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[75] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[76] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[77] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[78] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[79] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[80] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[102] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[103] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[104] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[105] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[106] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[107] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[108] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[109] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[110] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[111] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[112] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[113] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[135] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[136] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[137] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[138] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[139] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[140] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[162] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[163] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[164] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[165] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[166] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[167] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[189] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[190] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[191] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[192] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[193] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[194] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[111] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[112] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[113] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[138] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[139] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[140] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[165] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[166] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[167] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[192] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[193] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[194] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[129] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[130] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[131] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[156] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[157] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[158] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[183] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[184] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[185] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[210] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[211] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[212] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[129] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[130] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[131] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[132] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[133] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[134] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[156] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[157] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[158] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[159] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[160] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[161] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[183] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[184] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[185] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[186] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[187] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[188] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[210] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[211] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[212] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[213] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[214] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-        conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[215] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
       }
-      for (int i2_inner = 0; i2_inner < 7; ++i2_inner) {
-        compute[(((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 49)) + (i2_inner * 7)) + (((int)blockIdx.x) % 7))] = max((conv2d_nchw[i2_inner] + bias[(((((int)blockIdx.x) / 7) * 128) + ((int)threadIdx.x))]), 0.000000e+00f);
+      for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
+        for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
+          compute[((((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 98)) + (i1_inner * 49)) + ((((int)blockIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[((((((int)blockIdx.x) / 7) * 128) + (((int)threadIdx.x) * 2)) + i1_inner)]), 0.000000e+00f);
+        }
       }
     }
 
@@ -1586,7 +1370,7 @@ In the example below we resume the status and do more 5 trials.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 6 minutes  6.168 seconds)
+   **Total running time of the script:** ( 6 minutes  8.646 seconds)
 
 
 .. _sphx_glr_download_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py:
diff --git a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
index e10686bad7..ec436ce888 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
@@ -647,7 +647,7 @@ 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)  
-       7.9169       7.9170       7.9266       7.9072       0.0079   
+       7.9036       7.9046       7.9074       7.8987       0.0036   
                
 
 
@@ -675,7 +675,7 @@ Other Tips
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  7.705 seconds)
+   **Total running time of the script:** ( 1 minutes  7.266 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_x86.rst.txt b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_x86.rst.txt
index 732edc85f5..dbc33f44b7 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
@@ -666,7 +666,7 @@ 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)  
-      749.0074     749.4644     749.6283     747.9293      0.7652   
+      756.5266     754.0870     762.1017     753.3910      3.9525   
                
 
 
@@ -694,7 +694,7 @@ Other Tips
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  42.432 seconds)
+   **Total running time of the script:** ( 1 minutes  41.629 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 2a1892e417..1284d60ff5 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
@@ -389,74 +389,75 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
         @T.prim_func
         def main(placeholder: T.Buffer((128, 256), "float32"), placeholder_1: T.Buffer((4916, 16, 1), "float32"), placeholder_2: T.Buffer((4916,), "int32"), placeholder_3: T.Buffer((33,), "int32"), placeholder_4: T.Buffer((128, 512), "float32"), compute: T.Buffer((128, 512), "float32")):
             T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
-            for i0_outer_i1_outer_fused in T.parallel(16):
-                compute_1 = T.allocate([4096], "float32", "global")
-                compute_2 = T.Buffer((4096,), data=compute_1)
-                for i_outer_inner, nb_j_inner in T.grid(4, 2):
-                    for i_inner_init in range(32):
-                        cse_var_1: T.int32 = i_outer_inner * 1024 + i_inner_init * 32 + nb_j_inner * 16
-                        compute_2[cse_var_1] = T.float32(0)
-                        compute_2[cse_var_1 + 1] = T.float32(0)
-                        compute_2[cse_var_1 + 2] = T.float32(0)
-                        compute_2[cse_var_1 + 3] = T.float32(0)
-                        compute_2[cse_var_1 + 4] = T.float32(0)
-                        compute_2[cse_var_1 + 5] = T.float32(0)
-                        compute_2[cse_var_1 + 6] = T.float32(0)
-                        compute_2[cse_var_1 + 7] = T.float32(0)
-                        compute_2[cse_var_1 + 8] = T.float32(0)
-                        compute_2[cse_var_1 + 9] = T.float32(0)
-                        compute_2[cse_var_1 + 10] = T.float32(0)
-                        compute_2[cse_var_1 + 11] = T.float32(0)
-                        compute_2[cse_var_1 + 12] = T.float32(0)
-                        compute_2[cse_var_1 + 13] = T.float32(0)
-                        compute_2[cse_var_1 + 14] = T.float32(0)
-                        compute_2[cse_var_1 + 15] = T.float32(0)
-                    for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i0_outer_i1_outer_fused * 2 + nb_j_inner}), 32):
-                        cse_var_2 = T.int32()
-                        placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
-                        cse_var_21: T.int32 = elem_idx * 16
-                        cse_var_20: T.int32 = i0_outer_i1_outer_fused * 2 + nb_j_inner
-                        cse_var_19: T.int32 = i_outer_inner * 8192 + i_inner * 256
-                        cse_var_18: T.int32 = i_outer_inner * 1024 + i_inner * 32 + nb_j_inner * 16
-                        cse_var_17: T.int32 = cse_var_18 + 9
-                        cse_var_16: T.int32 = cse_var_18 + 8
-                        cse_var_15: T.int32 = cse_var_18 + 7
-                        cse_var_14: T.int32 = cse_var_18 + 6
-                        cse_var_13: T.int32 = cse_var_18 + 5
-                        cse_var_12: T.int32 = cse_var_18 + 4
-                        cse_var_11: T.int32 = cse_var_18 + 3
-                        cse_var_10: T.int32 = cse_var_18 + 2
-                        cse_var_9: T.int32 = cse_var_18 + 15
-                        cse_var_8: T.int32 = cse_var_18 + 14
-                        cse_var_7: T.int32 = cse_var_18 + 13
-                        cse_var_6: T.int32 = cse_var_18 + 12
-                        cse_var_5: T.int32 = cse_var_18 + 11
-                        cse_var_4: T.int32 = cse_var_18 + 10
-                        cse_var_3: T.int32 = cse_var_18 + 1
-                        placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
-                        placeholder_7 = T.Buffer((32768,), data=placeholder.data)
-                        placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
-                        compute_2[cse_var_18] = compute_2[cse_var_18] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                        compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                for i0_inner in range(128):
-                    cse_var_22: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 32
-                    compute_3 = T.Buffer((65536,), data=compute.data)
-                    placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
-                    compute_3[cse_var_22:cse_var_22 + 32] = T.max(compute_2[i0_inner * 32:i0_inner * 32 + 32] + placeholder_5[cse_var_22:cse_var_22 + 32], T.Broadcast(T.float32(0), 32))
+            for i0_outer in T.parallel(8):
+                compute_1 = T.allocate([512], "float32", "global")
+                for i1_outer in range(16):
+                    compute_2 = T.Buffer((512,), data=compute_1)
+                    for nb_j_inner in range(2):
+                        for i_inner_init in range(16):
+                            cse_var_1: T.int32 = i_inner_init * 32 + nb_j_inner * 16
+                            compute_2[cse_var_1] = T.float32(0)
+                            compute_2[cse_var_1 + 1] = T.float32(0)
+                            compute_2[cse_var_1 + 2] = T.float32(0)
+                            compute_2[cse_var_1 + 3] = T.float32(0)
+                            compute_2[cse_var_1 + 4] = T.float32(0)
+                            compute_2[cse_var_1 + 5] = T.float32(0)
+                            compute_2[cse_var_1 + 6] = T.float32(0)
+                            compute_2[cse_var_1 + 7] = T.float32(0)
+                            compute_2[cse_var_1 + 8] = T.float32(0)
+                            compute_2[cse_var_1 + 9] = T.float32(0)
+                            compute_2[cse_var_1 + 10] = T.float32(0)
+                            compute_2[cse_var_1 + 11] = T.float32(0)
+                            compute_2[cse_var_1 + 12] = T.float32(0)
+                            compute_2[cse_var_1 + 13] = T.float32(0)
+                            compute_2[cse_var_1 + 14] = T.float32(0)
+                            compute_2[cse_var_1 + 15] = T.float32(0)
+                        for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i1_outer * 2 + nb_j_inner}), 16):
+                            cse_var_2 = T.int32()
+                            placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
+                            cse_var_21: T.int32 = elem_idx * 16
+                            cse_var_20: T.int32 = i1_outer * 2 + nb_j_inner
+                            cse_var_19: T.int32 = i0_outer * 4096 + i_inner * 256
+                            cse_var_18: T.int32 = i_inner * 32 + nb_j_inner * 16
+                            cse_var_17: T.int32 = cse_var_18 + 9
+                            cse_var_16: T.int32 = cse_var_18 + 8
+                            cse_var_15: T.int32 = cse_var_18 + 7
+                            cse_var_14: T.int32 = cse_var_18 + 6
+                            cse_var_13: T.int32 = cse_var_18 + 5
+                            cse_var_12: T.int32 = cse_var_18 + 4
+                            cse_var_11: T.int32 = cse_var_18 + 3
+                            cse_var_10: T.int32 = cse_var_18 + 2
+                            cse_var_9: T.int32 = cse_var_18 + 15
+                            cse_var_8: T.int32 = cse_var_18 + 14
+                            cse_var_7: T.int32 = cse_var_18 + 13
+                            cse_var_6: T.int32 = cse_var_18 + 12
+                            cse_var_5: T.int32 = cse_var_18 + 11
+                            cse_var_4: T.int32 = cse_var_18 + 10
+                            cse_var_3: T.int32 = cse_var_18 + 1
+                            placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
+                            placeholder_7 = T.Buffer((32768,), data=placeholder.data)
+                            placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
+                            compute_2[cse_var_18] = compute_2[cse_var_18] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                            compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                    for i0_inner in range(16):
+                        cse_var_22: T.int32 = i0_outer * 8192 + i0_inner * 512 + i1_outer * 32
+                        compute_3 = T.Buffer((65536,), data=compute.data)
+                        placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
+                        compute_3[cse_var_22:cse_var_22 + 32] = T.max(compute_2[i0_inner * 32:i0_inner * 32 + 32] + placeholder_5[cse_var_22:cse_var_22 + 32], T.Broadcast(T.float32(0), 32))
 
 
 
@@ -506,7 +507,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 1.740 ms
+    Execution time of this operator: 1.667 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 4c17c88d6a..03313280b2 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
@@ -5,16 +5,16 @@
 
 Computation times
 =================
-**00:32.201** total execution time for **how_to_tune_with_autotvm** files:
+**00:34.245** 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``)           | 00:32.167 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)           | 00:34.212 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)               | 00:00.021 | 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_cuda.py` (``tune_relay_cuda.py``)             | 00:00.005 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)               | 00:00.004 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)               | 00:00.005 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``) | 00:00.004 | 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 6d100ffd45..4cfa348f99 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
@@ -268,7 +268,8 @@ for this template
     waiting for device...
     device available
     Get devices for measurement successfully!
-    No: 1   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    No: 1   GFLOPS: 67.10/67.10     result: MeasureResult(costs=(0.0034500863142857144,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.1896886825561523, timestamp=1678262153.3454974)      [('tile_f', [-1, 4, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1283341
+    No: 2   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -390,8 +391,8 @@ 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 875, 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, 4, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7440299
-    No: 2   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 1, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4252327
+    No: 3   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -513,8 +514,9 @@ 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 875, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
-    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 32, 1, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9937101
-    No: 3   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 4, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 64, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9575209
+    No: 4   GFLOPS: 30.59/67.10     result: MeasureResult(costs=(0.007568822571428571,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.9978384971618652, timestamp=1678262157.216696)        [('tile_f', [-1, 8, 1, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8170023
+    No: 5   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -636,8 +638,8 @@ 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 875, 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, 16, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3479826
-    No: 4   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 2, 128]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3128394
+    No: 6   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -759,8 +761,8 @@ 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 875, 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, 16, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8692675
-    No: 5   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 2, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4985634
+    No: 7   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -882,8 +884,8 @@ 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 875, 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, 128]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4706012
-    No: 6   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 4, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2080495
+    No: 8   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1005,8 +1007,8 @@ 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 875, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
-    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 32, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5777709
-    No: 7   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 1, 512]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4864419
+    No: 9   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1128,8 +1130,9 @@ 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 875, 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, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,724858
-    No: 8   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 1, 64]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3660343
+    No: 10  GFLOPS: 25.07/67.10     result: MeasureResult(costs=(0.009234245909090909,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.9820995330810547, timestamp=1678262163.0424588)       [('tile_f', [-1, 4, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4703482
+    No: 11  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1251,10 +1254,8 @@ 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 875, 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, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1534274
-    No: 9   GFLOPS: 81.89/81.89     result: MeasureResult(costs=(0.0028270755277777777,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.0235955715179443, timestamp=1678247077.446814)       [('tile_f', [-1, 4, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4080186
-    No: 10  GFLOPS: 7.31/81.89      result: MeasureResult(costs=(0.03166355125,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.513221263885498, timestamp=1678247078.2887018)       [('tile_f', [-1, 1, 1, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4718000
-    No: 11  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 2, 128]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 32, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6335773
+    No: 12  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1376,8 +1377,8 @@ 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 875, 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, 1, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2678508
-    No: 12  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 256, 1, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,732828
+    No: 13  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1499,8 +1500,8 @@ 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 875, 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, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2695065
-    No: 13  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 32, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 256]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7348130
+    No: 14  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1622,8 +1623,8 @@ 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 875, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
-    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 32, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9570404
-    No: 14  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 16, 4, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4563799
+    No: 15  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1745,9 +1746,8 @@ 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 875, 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, 64, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7951594
-    No: 15  GFLOPS: 2.34/81.89      result: MeasureResult(costs=(0.09909812150000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.931825876235962, timestamp=1678247082.4376202) [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,8714765
-    No: 16  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 2, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 512]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5804673
+    No: 16  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1869,8 +1869,9 @@ 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 875, 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, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1878382
-    No: 17  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 1, 4, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,446089
+    No: 17  GFLOPS: 4.26/67.10      result: MeasureResult(costs=(0.05434708775,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.2146730422973633, timestamp=1678262166.633485)       [('tile_f', [-1, 32, 1, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7128545
+    No: 18  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1992,8 +1993,8 @@ 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 875, 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, 128, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9950955
-    No: 18  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 8, 1, 64]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 512]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,384543
+    No: 19  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -2115,8 +2116,8 @@ 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 875, 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, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7193512
-    No: 19  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 2, 4, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 512, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,420053
+    No: 20  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
         func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
       File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -2238,130 +2239,7 @@ 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 875, in verify_pass
         raise InstantiationError("Skipped because of invalid gpu kernel")
-    tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7522353
-    No: 20  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
-        func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
-      File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
-        func = build(s, args, target=target, 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:1734
-      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:1674
-      19: run<>
-            at ../include/tvm/runtime/packed_func.h:1634
-      18: run<tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1649
-      13: operator()
-            at ../src/driver/driver_api.cc:402
-      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:388
-      11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
-            at ../src/driver/driver_api.cc:283
-      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:451
-      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:1753
-      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:1697
-      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:1621
-      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 875, 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:1734
-      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:1674
-      19: run<>
-            at ../include/tvm/runtime/packed_func.h:1634
-      18: run<tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1634
-      14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
-            at ../include/tvm/runtime/packed_func.h:1649
-      13: operator()
-            at ../src/driver/driver_api.cc:402
-      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:388
-      11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
-            at ../src/driver/driver_api.cc:283
-      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:451
-      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:1753
-      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:1697
-      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:1621
-      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 875, 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, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7027804
+    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, 1, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10032685
 
 
 
@@ -2416,9 +2294,9 @@ and measure running time.
     Finish loading 20 records
 
     Best config:
-    [('tile_f', [-1, 4, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4080186
+    [('tile_f', [-1, 4, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1283341
     Finish loading 20 records
-    Time cost of this operator: 0.001595
+    Time cost of this operator: 0.003360
 
 
 
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 d8ab90f0fe..83c8102270 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
@@ -360,10 +360,10 @@ 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  317.2     98.752   (1, 2, 10, 10, 3)  2       1        [317.2]           
-    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.055     0.951    (1, 6, 10, 10)     1       1        [3.055]           
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.954     0.297    (1, 1, 10, 10, 3)  1       1        [0.954]           
-    Total_time                                    -                                             321.209   -        -                  -       -        -                 
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  316.0     98.734   (1, 2, 10, 10, 3)  2       1        [316.0]           
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.099     0.968    (1, 6, 10, 10)     1       1        [3.099]           
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.954     0.298    (1, 1, 10, 10, 3)  1       1        [0.954]           
+    Total_time                                    -                                             320.053   -        -                  -       -        -                 
 
 
 
@@ -428,10 +428,10 @@ 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  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       1.796     1.702    (1, 6, 10, 10)     1       1        [1.796]           
-    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.855     0.81     (1, 3, 10, 10, 1)  1       1        [0.855]           
-    Total_time                                    -                                             105.552   -        -                  -       -        -                 
+    tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  114.1     97.661   (1, 6, 10, 10, 1)  2       1        [114.1]           
+    tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.769     1.514    (1, 6, 10, 10)     1       1        [1.769]           
+    tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.965     0.826    (1, 1, 10, 10, 3)  1       1        [0.965]           
+    Total_time                                    -                                             116.833   -        -                  -       -        -                 
 
 
 
@@ -439,7 +439,7 @@ Timing the tuned program
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  21.564 seconds)
+   **Total running time of the script:** ( 1 minutes  20.055 seconds)
 
 
 .. _sphx_glr_download_how_to_work_with_microtvm_micro_autotune.py:
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 4f8969a24e..50ed556765 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
@@ -118,7 +118,7 @@ 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]
    100%|##########| 3.42M/3.42M [00:00<00:00, 114MB/s]
+
      0%|          | 0.00/3.42M [00:00<?, ?B/s]
     61%|######    | 2.09M/3.42M [00:00<00:00, 20.2MB/s]
    100%|##########| 3.42M/3.42M [00:00<00:00, 31.3MB/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.
@@ -324,7 +324,7 @@ Look up prediction top 1 index in 1000 class synset.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  20.710 seconds)
+   **Total running time of the script:** ( 1 minutes  18.750 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 9ff8dc10d1..93f29f6556 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
@@ -218,7 +218,7 @@ take about **2 minutes** to download the Stanford Cars, while COCO 2017 validati
  .. code-block:: none
 
 
-    '/tmp/tmprd4ob6f2/images/random'
+    '/tmp/tmpsk87vt3y/images/random'
 
 
 
@@ -309,7 +309,7 @@ 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], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0]
+   :alt: [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [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
 
@@ -318,8 +318,8 @@ objects to other stuff? We can display some examples from our datasets using ``m
 
  .. code-block:: none
 
-    /tmp/tmprd4ob6f2/images/target contains 8144 images
-    /tmp/tmprd4ob6f2/images/random contains 5000 images
+    /tmp/tmpsk87vt3y/images/target contains 8144 images
+    /tmp/tmpsk87vt3y/images/random contains 5000 images
 
 
 
@@ -494,13 +494,13 @@ the time on our validation set).
  .. code-block:: none
 
     Epoch 1/3
-    328/328 - 47s - loss: 0.2333 - accuracy: 0.9228 - val_loss: 0.1396 - val_accuracy: 0.9490 - 47s/epoch - 145ms/step
+    328/328 - 47s - loss: 0.2111 - accuracy: 0.9258 - val_loss: 0.1170 - val_accuracy: 0.9603 - 47s/epoch - 143ms/step
     Epoch 2/3
-    328/328 - 43s - loss: 0.1045 - accuracy: 0.9619 - val_loss: 0.0988 - val_accuracy: 0.9679 - 43s/epoch - 132ms/step
+    328/328 - 43s - loss: 0.1019 - accuracy: 0.9631 - val_loss: 0.1391 - val_accuracy: 0.9547 - 43s/epoch - 132ms/step
     Epoch 3/3
-    328/328 - 43s - loss: 0.0714 - accuracy: 0.9730 - val_loss: 0.0917 - val_accuracy: 0.9675 - 43s/epoch - 132ms/step
+    328/328 - 43s - loss: 0.0651 - accuracy: 0.9761 - val_loss: 0.1452 - val_accuracy: 0.9528 - 43s/epoch - 132ms/step
 
-    <keras.callbacks.History object at 0x7f2c4652b910>
+    <keras.callbacks.History object at 0x7f131c467750>
 
 
 
@@ -861,7 +861,7 @@ 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  30.774 seconds)
+   **Total running time of the script:** ( 4 minutes  32.244 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 1d30b00246..3d7bd27149 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
@@ -5,18 +5,18 @@
 
 Computation times
 =================
-**07:30.654** total execution time for **how_to_work_with_microtvm** files:
+**07:28.310** total execution time for **how_to_work_with_microtvm** files:
 
 +-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``)           | 04:30.774 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``)           | 04:32.244 | 0.0 MB |
 +-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)     | 01:21.564 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)     | 01:20.055 | 0.0 MB |
 +-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``)       | 01:20.710 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``)       | 01:18.750 | 0.0 MB |
 +-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``)               | 00:10.263 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``)               | 00:10.165 | 0.0 MB |
 +-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)         | 00:07.343 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)         | 00:07.096 | 0.0 MB |
 +-----------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)         | 00:00.000 | 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 1ee918436f..a219f6892e 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
 =================
-**00:46.404** total execution time for **how_to_work_with_relay** files:
+**00:45.343** 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``) | 00:33.930 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:33.413 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)           | 00:10.589 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)           | 00:10.146 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)                             | 00:01.879 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)                             | 00:01.779 | 0.0 MB |
 +----------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)                 | 00:00.006 | 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 b79374fef8..32ddf88719 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
@@ -264,7 +264,7 @@ The following example customizes CUDA lowering rule for :code:`exp`.
  .. code-block:: none
 
 
-    <function my_cuda_math_rule at 0x7f2aea840440>
+    <function my_cuda_math_rule at 0x7f11cb1e0d40>
 
 
 
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 cc96bb0e84..797e50c09a 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
@@ -5,22 +5,22 @@
 
 Computation times
 =================
-**00:08.284** total execution time for **how_to_work_with_schedules** files:
+**00:07.845** total execution time for **how_to_work_with_schedules** files:
 
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)                 | 00:05.786 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)                 | 00:05.463 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)                     | 00:01.135 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)                     | 00:01.031 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)                     | 00:00.573 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)                     | 00:00.572 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)                               | 00:00.556 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)                               | 00:00.552 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)                     | 00:00.121 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)                     | 00:00.116 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.052 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)                               | 00:00.033 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)                               | 00:00.032 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)               | 00:00.027 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)               | 00:00.026 | 0.0 MB |
 +------------------------------------------------------------------------------------------------+-----------+--------+
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 61564f9db4..c7d4446af0 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
 =================
-**00:31.640** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:30.964** total execution time for **topic_vta_tutorials_autotvm** files:
 
 +---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.633 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:30.958 | 0.0 MB |
 +---------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)     | 00:00.007 | 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 16470378e2..71f9a36ce2 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
@@ -293,7 +293,7 @@ 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 33.78s!
+    resnet18_v1 inference graph built in 32.84s!
 
 
 
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 da9ccea1a9..796392db67 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
@@ -337,7 +337,7 @@ 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 22.92s!
+    yolov3-tiny inference graph built in 22.49s!
 
 
 
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 d5197bad9f..9580efdb6a 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
 =================
-**01:40.447** total execution time for **topic_vta_tutorials_frontend** files:
+**01:38.795** total execution time for **topic_vta_tutorials_frontend** files:
 
 +------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.531 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:49.495 | 0.0 MB |
 +------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)           | 00:49.916 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)           | 00:49.300 | 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 2dc477af40..57739e9299 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
 =================
-**00:03.189** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.127** total execution time for **topic_vta_tutorials_optimize** files:
 
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)         | 00:02.738 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)         | 00:02.680 | 0.0 MB |
 +--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.451 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.448 | 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 773c29b8e2..0aba8900b5 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
 =================
-**00:00.761** total execution time for **topic_vta_tutorials** files:
+**00:00.755** total execution time for **topic_vta_tutorials** files:
 
 +---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.393 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.390 | 0.0 MB |
 +---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.368 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.365 | 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 87110c1e8a..1ba2c45b11 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -318,7 +318,7 @@ We build the binary and check its correctness and performance.
 
  .. code-block:: none
 
-    Execution time of this operator: 94.116 ms
+    Execution time of this operator: 93.616 ms
 
 
 
@@ -416,7 +416,7 @@ resume the status and do more 5 trials.
  .. code-block:: none
 
     Resume search:
-    *E
+
 
 
 
@@ -434,7 +434,7 @@ operations.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  30.120 seconds)
+   **Total running time of the script:** ( 1 minutes  36.225 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 95fc385a5c..2eca09551f 100644
--- a/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_matmul_x86.rst.txt
@@ -454,16 +454,16 @@ reduce variance, we take 5 measurements and average them.
     waiting for device...
     device available
     Get devices for measurement successfully!
-    No: 1   GFLOPS: 2.23/2.23       result: MeasureResult(costs=(0.1201213538,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.186267852783203, timestamp=1678245488.5508614)        [('tile_y', [-1, 2]), ('tile_x', [-1, 16])],None,41
-    No: 2   GFLOPS: 2.07/2.23       result: MeasureResult(costs=(0.12976741460000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.3248813152313232, timestamp=1678245490.882508) [('tile_y', [-1, 256]), ('tile_x', [-1, 4])],None,28
-    No: 3   GFLOPS: 13.68/13.68     result: MeasureResult(costs=(0.0196172804,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5833511352539062, timestamp=1678245492.7074795)       [('tile_y', [-1, 128]), ('tile_x', [-1, 64])],None,67
-    No: 4   GFLOPS: 2.75/13.68      result: MeasureResult(costs=(0.097473613,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.806100606918335, timestamp=1678245495.7838056) [('tile_y', [-1, 2]), ('tile_x', [-1, 4])],None,21
-    No: 5   GFLOPS: 3.68/13.68      result: MeasureResult(costs=(0.0730037976,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4232501983642578, timestamp=1678245497.471335)        [('tile_y', [-1, 128]), ('tile_x', [-1, 16])],None,47
-    No: 6   GFLOPS: 10.31/13.68     result: MeasureResult(costs=(0.0260273598,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6705162525177002, timestamp=1678245498.1440794)       [('tile_y', [-1, 4]), ('tile_x', [-1, 32])],None,52
-    No: 7   GFLOPS: 11.47/13.68     result: MeasureResult(costs=(0.023395091,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6263086795806885, timestamp=1678245500.0269573)        [('tile_y', [-1, 4]), ('tile_x', [-1, 512])],None,92
-    No: 8   GFLOPS: 8.46/13.68      result: MeasureResult(costs=(0.031740723600000006,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.9154953956604004, timestamp=1678245500.7952504)       [('tile_y', [-1, 16]), ('tile_x', [-1, 64])],None,64
-    No: 9   GFLOPS: 9.73/13.68      result: MeasureResult(costs=(0.0275947852,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6466634273529053, timestamp=1678245501.6098535)       [('tile_y', [-1, 1]), ('tile_x', [-1, 512])],None,90
-    No: 10  GFLOPS: 10.60/13.68     result: MeasureResult(costs=(0.0253268806,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6569938659667969, timestamp=1678245502.2707105)       [('tile_y', [-1, 512]), ('tile_x', [-1, 64])],None,69
+    No: 1   GFLOPS: 10.47/10.47     result: MeasureResult(costs=(0.025644233199999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6875050067901611, timestamp=1678260571.8454494)       [('tile_y', [-1, 8]), ('tile_x', [-1, 32])],None,53
+    No: 2   GFLOPS: 11.50/11.50     result: MeasureResult(costs=(0.0233500936,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.650078296661377, timestamp=1678260572.4850514)        [('tile_y', [-1, 4]), ('tile_x', [-1, 256])],None,82
+    No: 3   GFLOPS: 11.34/11.50     result: MeasureResult(costs=(0.0236804312,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6256048679351807, timestamp=1678260574.3679078)       [('tile_y', [-1, 256]), ('tile_x', [-1, 512])],None,98
+    No: 4   GFLOPS: 1.36/11.50      result: MeasureResult(costs=(0.19770630220000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.396682024002075, timestamp=1678260579.0334325) [('tile_y', [-1, 1]), ('tile_x', [-1, 2])],None,10
+    No: 5   GFLOPS: 1.97/11.50      result: MeasureResult(costs=(0.136173345,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.417804479598999, timestamp=1678260581.5845983) [('tile_y', [-1, 8]), ('tile_x', [-1, 1])],None,3
+    No: 6   GFLOPS: 11.58/11.58     result: MeasureResult(costs=(0.0231856474,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6284253597259521, timestamp=1678260583.4592807)       [('tile_y', [-1, 8]), ('tile_x', [-1, 512])],None,93
+    No: 7   GFLOPS: 7.99/11.58      result: MeasureResult(costs=(0.0336107182,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.7713046073913574, timestamp=1678260584.255196)        [('tile_y', [-1, 1]), ('tile_x', [-1, 32])],None,50
+    No: 8   GFLOPS: 2.92/11.58      result: MeasureResult(costs=(0.0920347352,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7195570468902588, timestamp=1678260585.9844246)       [('tile_y', [-1, 512]), ('tile_x', [-1, 16])],None,49
+    No: 9   GFLOPS: 10.42/11.58     result: MeasureResult(costs=(0.025769255399999996,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6670513153076172, timestamp=1678260586.7896328)       [('tile_y', [-1, 8]), ('tile_x', [-1, 64])],None,63
+    No: 10  GFLOPS: 9.86/11.58      result: MeasureResult(costs=(0.027220091800000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.7201242446899414, timestamp=1678260587.4845228)       [('tile_y', [-1, 1]), ('tile_x', [-1, 128])],None,70
 
 
 
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index 78f5e8a0ac..8a72f52581 100644
--- a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
@@ -311,7 +311,7 @@ standard deviation.
 
  .. code-block:: none
 
-    {'mean': 511.48442123999985, 'median': 511.4933623000013, 'std': 1.0384856733831964}
+    {'mean': 509.04423982000253, 'median': 508.817199650008, 'std': 1.8627309514183343}
 
 
 
@@ -545,32 +545,30 @@ 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:    8.55/  13.55 GFLOPS | Progress: (4/20) | 12.03 s
    [Task  1/25]  Current/Best:   19.24/  19.24 GFLOPS | Progress: (8/20) | 15.68 s
    [Task  1/25]  Current/Best:   21.72/  21.72 GFLOPS | Progress: (12/20) | 19.17 s
    [Task  1/25]  Current/Best:   19.20/  21.72 GFLOPS | Progress: (16/20) | 21.49 s
    [Task  1/25]  Current/Best:   14.16/  23.25 GFLOPS | Progress: (20/20) | 23.49 s Done.
-
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  2/25]  Current/Best:   11.90/  16.32 GFLOPS | Progress: (4/20) | 4.59 s
    [Task  2/25]  Current/Best:   20.02/  20.02 GFLOPS | Progress: (8/20) | 6.62 s
    [Task  2/25]  Current/Best:   12.76/  20.02 GFLOPS | Progress: (12/20) | 8.26 s
    [Task  2/25]  Current/Best:    7.76/  21.50 GFLOPS | Progress: (16/20) | 9.83 s
    [Task  2/25]  Current/Best:    4.11/  21.50 GFLOPS | Progress: (20/20) | 11.69 s Done.
-
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  3/25]  Current/Best:   12.50/  17.23 GFLOPS | Progress: (4/20) | 5.02 s
    [Task  3/25]  Current/Best:   10.19/  17.23 GFLOPS | Progress: (8/20) | 8.66 s
    [Task  3/25]  Current/Best:   20.58/  20.58 GFLOPS | Progress: (12/20) | 12.42 s
    [Task  3/25]  Current/Best:    8.75/  23.61 GFLOPS | Progress: (16/20) | 14.59 s
    [Task  3/25]  Current/Best:   13.52/  23.61 GFLOPS | Progress: (20/20) | 17.09 s Done.
-
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  4/25]  Current/Best:   18.61/  18.61 GFLOPS | Progress: (4/20) | 7.47 s
    [Task  4/25]  Current/Best:    3.74/  18.61 GFLOPS | Progress: (8/20) | 9.82 s
    [Task  4/25]  Current/Best:    3.38/  18.61 GFLOPS | Progress: (12/20) | 14.52 s
    [Task  4/25]  Current/Best:   16.58/  18.61 GFLOPS | Progress: (16/20) | 20.97 s
    [Task  4/25]  Current/Best:    7.70/  18.61 GFLOPS | Progress: (20/20) | 23.04 s Done.
-
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  5/25]  Current/Best:   21.04/  21.04 GFLOPS | Progress: (4/20) | 4.84 s
    [Task  5/25]  Current/Best:   14.38/  21.04 GFLOPS | Progress: (8/20) | 6.81 s
    [Task  5/25]  Current/Best:   10.67/  21.55 GFLOPS | Progress: (12/20) | 8.69 s
    [Task  5/25]  Current/Best:   10.64/  21.55 GFLOPS | Progress: (16/20) | 10.57 s
    [Task  5/25]  Current/Best:    3.41/  21.55 GFLOPS | Progress: (20/20) | 13.05 s Done.
-
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  6/25]  Current/Best:   11.74/  13.57 GFLOPS | Progress: (4/20) | 6.11 s
    [Task  6/25]  Current/Best:    9.92/  13.57 GFLOPS | Progress: (8/20) | 10.13 s
    [Task  6/25]  Current/Best:   17.24/  18.65 GFLOPS | Progress: (12/20) | 12.63 s
    [Task  6/25]  Current/Best:    5.29/  18.65 GFLOPS | Progress: (16/20) | 15.63 s
    [Task  6/25]  Current/Best:   22.51/  22.51 GFLOPS | Progress: (20/20) | 18.31 s Done.
-
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  7/25]  Current/Best:   15.68/  19.08 GFLOPS | Progress: (4/20) | 5.08 s
    [Task  7/25]  Current/Best:   18.76/  22.52 GFLOPS | Progress: (8/20) | 7.13 s
    [Task  7/25]  Current/Best:   15.29/  22.52 GFLOPS | Progress: (12/20) | 9.33 s
    [Task  7/25]  Current/Best:    6.71/  22.52 GFLOPS | Progress: (16/20) | 11.95 s
    [Task  7/25]  Current/Best:   13.54/  22.52 GFLOPS | Progress: (20/20) | 14.35 s Done.
-
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  8/25]  Current/Best:   16.35/  21.60 GFLOPS | Progress: (4/20) | 4.94 s
    [Task  8/25]  Current/Best:   11.66/  21.60 GFLOPS | Progress: (8/20) | 9.98 s
    [Task  8/25]  Current/Best:   10.99/  21.60 GFLOPS | Progress: (12/20) | 14.98 s
    [Task  8/25]  Current/Best:   18.94/  21.60 GFLOPS | Progress: (16/20) | 17.26 s
    [Task  8/25]  Current/Best:    3.08/  21.60 GFLOPS | Progress: (20/20) | 20.36 s Done.
-
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  9/25]  Current/Best:   11.54/  20.49 GFLOPS | Progress: (4/20) | 5.52 s
    [Task  9/25]  Current/Best:    8.24/  20.49 GFLOPS | Progress: (8/20) | 7.31 s
    [Task  9/25]  Current/Best:    4.79/  20.49 GFLOPS | Progress: (12/20) | 14.63 s
    [Task  9/25]  Current/Best:    6.32/  20.49 GFLOPS | Progress: (16/20) | 20.16 s
    [Task  9/25]  Current/Best:   14.87/  20.49 GFLOPS | Progress: (20/20) | 23.32 s Done.
-
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 10/25]  Current/Best:   17.76/  17.76 GFLOPS | Progress: (4/20) | 6.12 s
    [Task 10/25]  Current/Best:   15.50/  17.76 GFLOPS | Progress: (8/20) | 8.28 s
    [Task 10/25]  Current/Best:   17.76/  17.76 GFLOPS | Progress: (12/20) | 10.47 s
    [Task 10/25]  Current/Best:    6.56/  18.28 GFLOPS | Progress: (16/20) | 12.40 s
    [Task 10/25]  Current/Best:   16.26/  18.28 GFLOPS | Progress: (20/20) | 14.20 s Done.
-
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 11/25]  Current/Best:   11.16/  16.43 GFLOPS | Progress: (4/20) | 6.05 s
    [Task 11/25]  Current/Best:   15.30/  16.43 GFLOPS | Progress: (8/20) | 8.77 s
    [Task 11/25]  Current/Best:   12.06/  16.46 GFLOPS | Progress: (12/20) | 11.42 s
    [Task 11/25]  Current/Best:   12.29/  19.36 GFLOPS | Progress: (16/20) | 13.90 s
    [Task 11/25]  Current/Best:   15.28/  19.36 GFLOPS | Progress: (20/20) | 16.18 s Done.
-
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 12/25]  Current/Best:   10.55/  14.02 GFLOPS | Progress: (4/20) | 7.25 s
    [Task 12/25]  Current/Best:   15.88/  15.88 GFLOPS | Progress: (8/20) | 11.37 s
    [Task 12/25]  Current/Best:   20.77/  20.77 GFLOPS | Progress: (12/20) | 13.57 s
    [Task 12/25]  Current/Best:    6.19/  20.77 GFLOPS | Progress: (16/20) | 17.60 s
    [Task 12/25]  Current/Best:    3.05/  20.77 GFLOPS | Progress: (20/20) | 21.36 s Done.
-
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 13/25]  Current/Best:   12.80/  19.95 GFLOPS | Progress: (4/20) | 6.63 s
    [Task 13/25]  Current/Best:   18.17/  19.95 GFLOPS | Progress: (8/20) | 9.96 s
    [Task 13/25]  Current/Best:    9.15/  20.74 GFLOPS | Progress: (12/20) | 14.92 s
    [Task 13/25]  Current/Best:   15.19/  20.74 GFLOPS | Progress: (16/20) | 18.21 s
    [Task 13/25]  Current/Best:   16.38/  20.74 GFLOPS | Progress: (20/20) | 20.84 s Done.
-
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 14/25]  Current/Best:    4.19/  14.65 GFLOPS | Progress: (4/20) | 7.21 s
    [Task 14/25]  Current/Best:   20.96/  20.96 GFLOPS | Progress: (8/20) | 10.51 s
    [Task 14/25]  Current/Best:   16.67/  20.96 GFLOPS | Progress: (12/20) | 16.39 s
    [Task 14/25]  Current/Best:   16.32/  20.96 GFLOPS | Progress: (16/20) | 21.08 s
    [Task 14/25]  Current/Best:   11.90/  20.96 GFLOPS | Progress: (20/20) | 23.83 s
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
-
    [Task 15/25]  Current/Best:   15.46/  15.46 GFLOPS | Progress: (4/20) | 5.97 s
    [Task 15/25]  Current/Best:    3.13/  16.00 GFLOPS | Progress: (8/20) | 8.03 s
    [Task 15/25]  Current/Best:    1.70/  16.00 GFLOPS | Progress: (12/20) | 17.96 s
    [Task 15/25]  Current/Best:   12.34/  19.10 GFLOPS | Progress: (16/20) | 19.44 s
    [Task 15/25]  Current/Best:    9.30/  19.10 GFLOPS | Progress: (20/20) | 26.81 s
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 16/25]  Current/Best:    9.67/  19.53 GFLOPS | Progress: (4/20) | 5.70 s
    [Task 16/25]  Current/Best:   13.55/  19.53 GFLOPS | Progress: (8/20) | 7.37 s
    [Task 16/25]  Current/Best:    6.01/  20.45 GFLOPS | Progress: (12/20) | 9.01 s
    [Task 16/25]  Current/Best:   10.35/  20.45 GFLOPS | Progress: (16/20) | 12.26 s
    [Task 16/25]  Current/Best:   13.58/  20.45 GFLOPS | Progress: (20/20) | 14.55 s Done.
-
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 17/25]  Current/Best:   16.29/  16.29 GFLOPS | Progress: (4/20) | 5.24 s
    [Task 17/25]  Current/Best:    5.37/  16.57 GFLOPS | Progress: (8/20) | 8.21 s
    [Task 17/25]  Current/Best:   12.07/  20.83 GFLOPS | Progress: (12/20) | 11.88 s
    [Task 17/25]  Current/Best:   18.27/  20.83 GFLOPS | Progress: (16/20) | 14.40 s
    [Task 17/25]  Current/Best:   20.27/  20.83 GFLOPS | Progress: (20/20) | 18.21 s Done.
-
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 18/25]  Current/Best:   17.44/  19.18 GFLOPS | Progress: (4/20) | 4.86 s
    [Task 18/25]  Current/Best:   14.44/  19.18 GFLOPS | Progress: (8/20) | 7.10 s
    [Task 18/25]  Current/Best:   10.10/  19.18 GFLOPS | Progress: (12/20) | 10.16 s
    [Task 18/25]  Current/Best:   10.30/  20.56 GFLOPS | Progress: (16/20) | 14.00 s
    [Task 18/25]  Current/Best:   15.37/  20.56 GFLOPS | Progress: (20/20) | 16.49 s Done.
-
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 19/25]  Current/Best:    5.87/  10.31 GFLOPS | Progress: (4/20) | 7.28 s
    [Task 19/25]  Current/Best:    9.59/  17.76 GFLOPS | Progress: (8/20) | 10.66 s
    [Task 19/25]  Current/Best:   16.26/  18.83 GFLOPS | Progress: (12/20) | 14.13 s
    [Task 19/25]  Current/Best:   19.32/  21.58 GFLOPS | Progress: (16/20) | 16.90 s
    [Task 19/25]  Current/Best:   20.59/  21.66 GFLOPS | Progress: (20/20) | 19.76 s Done.
-
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 20/25]  Current/Best:   20.26/  20.26 GFLOPS | Progress: (4/20) | 5.24 s
    [Task 20/25]  Current/Best:   18.06/  20.26 GFLOPS | Progress: (8/20) | 8.94 s
    [Task 20/25]  Current/Best:   20.46/  20.46 GFLOPS | Progress: (12/20) | 10.83 s
    [Task 20/25]  Current/Best:   12.55/  20.46 GFLOPS | Progress: (16/20) | 14.04 s
    [Task 20/25]  Current/Best:    7.28/  20.46 GFLOPS | Progress: (20/20) | 17.45 s
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
-     Done.
-
    [Task 21/25]  Current/Best:    4.42/  12.64 GFLOPS | Progress: (4/20) | 6.46 s
    [Task 21/25]  Current/Best:   18.45/  18.45 GFLOPS | Progress: (8/20) | 9.26 s
    [Task 21/25]  Current/Best:    2.64/  18.45 GFLOPS | Progress: (12/20) | 12.70 s
    [Task 21/25]  Current/Best:   22.14/  22.14 GFLOPS | Progress: (16/20) | 14.74 s
    [Task 21/25]  Current/Best:   16.27/  22.14 GFLOPS | Progress: (20/20) | 17.28 s
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 22/25]  Current/Best:    4.92/  22.29 GFLOPS | Progress: (4/20) | 6.85 s
    [Task 22/25]  Current/Best:    9.34/  22.29 GFLOPS | Progress: (8/20) | 9.08 s
    [Task 22/25]  Current/Best:    7.98/  22.29 GFLOPS | Progress: (12/20) | 11.41 s
    [Task 22/25]  Current/Best:   13.65/  22.29 GFLOPS | Progress: (16/20) | 13.13 s
    [Task 22/25]  Current/Best:    5.31/  22.29 GFLOPS | Progress: (20/20) | 16.92 s Done.
-
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 23/25]  Current/Best:   10.37/  15.28 GFLOPS | Progress: (4/20) | 5.61 s
    [Task 23/25]  Current/Best:   12.06/  19.49 GFLOPS | Progress: (8/20) | 8.53 s
    [Task 23/25]  Current/Best:   18.61/  19.49 GFLOPS | Progress: (12/20) | 11.83 s
    [Task 23/25]  Current/Best:   21.33/  21.33 GFLOPS | Progress: (16/20) | 14.23 s
    [Task 23/25]  Current/Best:    2.69/  22.98 GFLOPS | Progress: (20/20) | 18.42 s Done.
-
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 24/25]  Current/Best:    3.61/   4.83 GFLOPS | Progress: (4/20) | 13.77 s
    [Task 24/25]  Current/Best:    7.20/   7.20 GFLOPS | Progress: (8/20) | 24.15 s
    [Task 24/25]  Current/Best:    0.00/   7.20 GFLOPS | Progress: (12/20) | 34.21 s
    [Task 24/25]  Current/Best:    6.29/   7.20 GFLOPS | Progress: (16/20) | 46.86 s
    [Task 24/25]  Current/Best:    3.06/   7.20 GFLOPS | Progress: (20/20) | 57.81 s
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 25/25]  Current/Best:    4.16/   5.65 GFLOPS | Progress: (4/20) | 13.73 s
    [Task 25/25]  Current/Best:    3.00/   9.44 GFLOPS | Progress: (8/20) | 17.28 s
    [Task 25/25]  Current/Best:    7.97/   9.44 GFLOPS | Progress: (12/20) | 28.23 s Done.
-     Done.
-
    [Task 25/25]  Current/Best:    6.07/   9.44 GFLOPS | Progress: (16/20) | 34.71 s
    [Task 25/25]  Current/Best:    7.81/   9.44 GFLOPS | Progress: (20/20) | 37.19 s Done.
-
+
    [Task  1/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  1/25]  Current/Best:    6.96/  23.09 GFLOPS | Progress: (4/20) | 11.59 s
    [Task  1/25]  Current/Best:   23.00/  23.09 GFLOPS | Progress: (8/20) | 15.34 s
    [Task  1/25]  Current/Best:    9.65/  23.09 GFLOPS | Progress: (12/20) | 19.25 s
    [Task  1/25]  Current/Best:   12.67/  23.09 GFLOPS | Progress: (16/20) | 21.57 s
    [Task  1/25]  Current/Best:   10.00/  23.09 GFLOPS | Progress: (20/20) | 25.94 s Done.
+
    [Task  2/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  2/25]  Current/Best:   17.04/  17.73 GFLOPS | Progress: (4/20) | 4.31 s
    [Task  2/25]  Current/Best:   13.99/  17.73 GFLOPS | Progress: (8/20) | 5.97 s
    [Task  2/25]  Current/Best:   16.15/  20.77 GFLOPS | Progress: (12/20) | 7.75 s
    [Task  2/25]  Current/Best:    5.56/  20.77 GFLOPS | Progress: (16/20) | 9.59 s
    [Task  2/25]  Current/Best:   16.54/  20.77 GFLOPS | Progress: (20/20) | 11.17 s Done.
+
    [Task  3/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  3/25]  Current/Best:   19.50/  19.50 GFLOPS | Progress: (4/20) | 4.87 s
    [Task  3/25]  Current/Best:   14.68/  19.50 GFLOPS | Progress: (8/20) | 7.02 s
    [Task  3/25]  Current/Best:   11.52/  24.29 GFLOPS | Progress: (12/20) | 9.14 s
    [Task  3/25]  Current/Best:   13.92/  24.29 GFLOPS | Progress: (16/20) | 11.49 s
    [Task  3/25]  Current/Best:    2.98/  24.29 GFLOPS | Progress: (20/20) | 14.49 s Done.
+
    [Task  4/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  4/25]  Current/Best:    5.65/  20.14 GFLOPS | Progress: (4/20) | 6.41 s
    [Task  4/25]  Current/Best:    5.08/  20.19 GFLOPS | Progress: (8/20) | 9.51 s
    [Task  4/25]  Current/Best:   14.29/  20.19 GFLOPS | Progress: (12/20) | 11.95 s
    [Task  4/25]  Current/Best:   11.41/  20.19 GFLOPS | Progress: (16/20) | 17.20 s
    [Task  4/25]  Current/Best:   18.51/  20.19 GFLOPS | Progress: (20/20) | 19.19 s Done.
+
    [Task  5/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  5/25]  Current/Best:    8.06/  18.54 GFLOPS | Progress: (4/20) | 4.72 s
    [Task  5/25]  Current/Best:    5.58/  18.54 GFLOPS | Progress: (8/20) | 6.86 s
    [Task  5/25]  Current/Best:   20.01/  20.01 GFLOPS | Progress: (12/20) | 9.45 s
    [Task  5/25]  Current/Best:   10.76/  20.01 GFLOPS | Progress: (16/20) | 11.86 s
    [Task  5/25]  Current/Best:   16.25/  20.01 GFLOPS | Progress: (20/20) | 14.73 s Done.
+
    [Task  6/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  6/25]  Current/Best:    6.15/  14.05 GFLOPS | Progress: (4/20) | 5.77 s
    [Task  6/25]  Current/Best:   14.63/  20.43 GFLOPS | Progress: (8/20) | 8.91 s
    [Task  6/25]  Current/Best:   13.79/  20.43 GFLOPS | Progress: (12/20) | 12.39 s
    [Task  6/25]  Current/Best:   14.66/  20.43 GFLOPS | Progress: (16/20) | 15.69 s
    [Task  6/25]  Current/Best:   10.26/  20.43 GFLOPS | Progress: (20/20) | 23.52 s Done.
+
    [Task  7/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  7/25]  Current/Best:   12.41/  15.15 GFLOPS | Progress: (4/20) | 5.34 s
    [Task  7/25]  Current/Best:    8.40/  15.61 GFLOPS | Progress: (8/20) | 8.13 s
    [Task  7/25]  Current/Best:   21.93/  21.93 GFLOPS | Progress: (12/20) | 9.89 s
    [Task  7/25]  Current/Best:   16.37/  21.93 GFLOPS | Progress: (16/20) | 12.65 s
    [Task  7/25]  Current/Best:   14.03/  21.93 GFLOPS | Progress: (20/20) | 15.10 s Done.
+
    [Task  8/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  8/25]  Current/Best:   10.25/  19.51 GFLOPS | Progress: (4/20) | 5.21 s
    [Task  8/25]  Current/Best:    9.67/  19.51 GFLOPS | Progress: (8/20) | 7.49 s
    [Task  8/25]  Current/Best:    8.10/  19.51 GFLOPS | Progress: (12/20) | 10.29 s
    [Task  8/25]  Current/Best:    2.84/  19.51 GFLOPS | Progress: (16/20) | 14.14 s
    [Task  8/25]  Current/Best:   13.69/  19.51 GFLOPS | Progress: (20/20) | 16.82 s Done.
+
    [Task  9/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task  9/25]  Current/Best:   12.51/  15.69 GFLOPS | Progress: (4/20) | 5.88 s
    [Task  9/25]  Current/Best:   11.49/  19.67 GFLOPS | Progress: (8/20) | 10.61 s
    [Task  9/25]  Current/Best:    6.83/  19.67 GFLOPS | Progress: (12/20) | 14.90 s
    [Task  9/25]  Current/Best:   18.36/  19.67 GFLOPS | Progress: (16/20) | 17.65 s
    [Task  9/25]  Current/Best:   14.63/  19.67 GFLOPS | Progress: (20/20) | 19.42 s Done.
+
    [Task 10/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 10/25]  Current/Best:   16.90/  21.64 GFLOPS | Progress: (4/20) | 4.39 s
    [Task 10/25]  Current/Best:   10.97/  21.64 GFLOPS | Progress: (8/20) | 6.90 s
    [Task 10/25]  Current/Best:   16.01/  21.64 GFLOPS | Progress: (12/20) | 9.83 s
    [Task 10/25]  Current/Best:   18.44/  21.64 GFLOPS | Progress: (16/20) | 11.44 s
    [Task 10/25]  Current/Best:   13.13/  21.64 GFLOPS | Progress: (20/20) | 13.37 s Done.
+
    [Task 11/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 11/25]  Current/Best:    7.03/  16.07 GFLOPS | Progress: (4/20) | 5.76 s
    [Task 11/25]  Current/Best:   14.02/  20.12 GFLOPS | Progress: (8/20) | 8.61 s
    [Task 11/25]  Current/Best:   18.55/  20.12 GFLOPS | Progress: (12/20) | 11.01 s
    [Task 11/25]  Current/Best:   17.28/  20.12 GFLOPS | Progress: (16/20) | 14.77 s
    [Task 11/25]  Current/Best:   12.31/  20.12 GFLOPS | Progress: (20/20) | 17.34 s Done.
+
    [Task 12/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 12/25]  Current/Best:   14.96/  16.79 GFLOPS | Progress: (4/20) | 5.05 s
    [Task 12/25]  Current/Best:   13.94/  16.79 GFLOPS | Progress: (8/20) | 7.24 s
    [Task 12/25]  Current/Best:    3.23/  16.79 GFLOPS | Progress: (12/20) | 11.07 s
    [Task 12/25]  Current/Best:    5.03/  18.54 GFLOPS | Progress: (16/20) | 13.67 s
    [Task 12/25]  Current/Best:    3.29/  18.54 GFLOPS | Progress: (20/20) | 16.61 s Done.
+
    [Task 13/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 13/25]  Current/Best:   11.90/  20.22 GFLOPS | Progress: (4/20) | 6.80 s
    [Task 13/25]  Current/Best:   12.24/  20.22 GFLOPS | Progress: (8/20) | 10.34 s
    [Task 13/25]  Current/Best:   11.27/  20.22 GFLOPS | Progress: (12/20) | 14.27 s
    [Task 13/25]  Current/Best:   20.22/  20.22 GFLOPS | Progress: (16/20) | 17.32 s
    [Task 13/25]  Current/Best:    9.97/  20.22 GFLOPS | Progress: (20/20) | 20.24 s Done.
+
    [Task 14/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 14/25]  Current/Best:    9.63/  11.19 GFLOPS | Progress: (4/20) | 11.05 s
    [Task 14/25]  Current/Best:   12.94/  15.72 GFLOPS | Progress: (8/20) | 13.28 s
    [Task 14/25]  Current/Best:    9.39/  15.72 GFLOPS | Progress: (12/20) | 18.12 s
    [Task 14/25]  Current/Best:   14.15/  15.72 GFLOPS | Progress: (16/20) | 21.51 s
    [Task 14/25]  Current/Best:    5.15/  15.72 GFLOPS | Progress: (20/20) | 28.09 s Done.
+
    [Task 15/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 15/25]  Current/Best:   14.39/  20.02 GFLOPS | Progress: (4/20) | 4.88 s
    [Task 15/25]  Current/Best:   15.45/  20.02 GFLOPS | Progress: (8/20) | 6.81 s
    [Task 15/25]  Current/Best:   12.65/  20.02 GFLOPS | Progress: (12/20) | 10.10 s
    [Task 15/25]  Current/Best:   18.59/  20.02 GFLOPS | Progress: (16/20) | 12.20 s
    [Task 15/25]  Current/Best:   21.71/  21.71 GFLOPS | Progress: (20/20) | 15.91 s Done.
+
    [Task 16/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 16/25]  Current/Best:   12.37/  12.37 GFLOPS | Progress: (4/20) | 6.29 s
    [Task 16/25]  Current/Best:   18.60/  18.87 GFLOPS | Progress: (8/20) | 7.85 s
    [Task 16/25]  Current/Best:   18.91/  18.91 GFLOPS | Progress: (12/20) | 9.48 s
    [Task 16/25]  Current/Best:   18.26/  18.91 GFLOPS | Progress: (16/20) | 11.88 s
    [Task 16/25]  Current/Best:   10.94/  18.91 GFLOPS | Progress: (20/20) | 13.92 s Done.
+
    [Task 17/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 17/25]  Current/Best:   18.91/  18.91 GFLOPS | Progress: (4/20) | 5.94 s
    [Task 17/25]  Current/Best:   18.19/  18.91 GFLOPS | Progress: (8/20) | 8.78 s
    [Task 17/25]  Current/Best:   17.03/  20.54 GFLOPS | Progress: (12/20) | 10.94 s
    [Task 17/25]  Current/Best:    4.74/  20.54 GFLOPS | Progress: (16/20) | 14.12 s
    [Task 17/25]  Current/Best:   18.75/  20.54 GFLOPS | Progress: (20/20) | 17.09 s Done.
+
    [Task 18/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 18/25]  Current/Best:    5.34/   8.63 GFLOPS | Progress: (4/20) | 5.89 s
    [Task 18/25]  Current/Best:   11.23/  19.20 GFLOPS | Progress: (8/20) | 8.04 s
    [Task 18/25]  Current/Best:   15.08/  19.20 GFLOPS | Progress: (12/20) | 14.90 s
    [Task 18/25]  Current/Best:   21.37/  21.37 GFLOPS | Progress: (16/20) | 17.43 s
    [Task 18/25]  Current/Best:   11.05/  21.37 GFLOPS | Progress: (20/20) | 20.80 s Done.
+
    [Task 19/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 19/25]  Current/Best:    2.59/  12.16 GFLOPS | Progress: (4/20) | 7.25 s
    [Task 19/25]  Current/Best:   13.49/  18.23 GFLOPS | Progress: (8/20) | 13.38 s
    [Task 19/25]  Current/Best:   17.75/  21.19 GFLOPS | Progress: (12/20) | 16.24 s
    [Task 19/25]  Current/Best:   10.64/  21.19 GFLOPS | Progress: (16/20) | 20.93 s
    [Task 19/25]  Current/Best:   12.01/  21.19 GFLOPS | Progress: (20/20) | 23.88 s Done.
+
    [Task 20/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 20/25]  Current/Best:   10.34/  17.51 GFLOPS | Progress: (4/20) | 5.52 s
    [Task 20/25]  Current/Best:   15.42/  17.51 GFLOPS | Progress: (8/20) | 8.25 s
    [Task 20/25]  Current/Best:    6.07/  17.51 GFLOPS | Progress: (12/20) | 11.22 s
    [Task 20/25]  Current/Best:    8.19/  18.59 GFLOPS | Progress: (16/20) | 15.40 s
    [Task 20/25]  Current/Best:    8.23/  18.59 GFLOPS | Progress: (20/20) | 18.69 s
    [Task 21/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 21/25]  Current/Best:   14.15/  14.29 GFLOPS | Progress: (4/20) | 4.74 s
    [Task 21/25]  Current/Best:   17.13/  17.13 GFLOPS | Progress: (8/20) | 7.69 s
    [Task 21/25]  Current/Best:    8.88/  17.13 GFLOPS | Progress: (12/20) | 9.13 s Done.
+
    [Task 21/25]  Current/Best:   13.54/  17.13 GFLOPS | Progress: (16/20) | 12.58 s
    [Task 21/25]  Current/Best:    8.33/  17.13 GFLOPS | Progress: (20/20) | 15.56 s
    [Task 22/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 22/25]  Current/Best:    5.03/  19.86 GFLOPS | Progress: (4/20) | 5.65 s
    [Task 22/25]  Current/Best:    1.56/  19.86 GFLOPS | Progress: (8/20) | 8.07 s
    [Task 22/25]  Current/Best:   18.16/  20.43 GFLOPS | Progress: (12/20) | 9.96 s
    [Task 22/25]  Current/Best:   15.15/  20.43 GFLOPS | Progress: (16/20) | 11.62 s
    [Task 22/25]  Current/Best:   10.55/  20.43 GFLOPS | Progress: (20/20) | 13.58 s Done.
+
    [Task 23/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 23/25]  Current/Best:   19.21/  19.21 GFLOPS | Progress: (4/20) | 7.00 s
    [Task 23/25]  Current/Best:   23.01/  23.01 GFLOPS | Progress: (8/20) | 10.74 s
    [Task 23/25]  Current/Best:   18.91/  23.01 GFLOPS | Progress: (12/20) | 14.34 s
    [Task 23/25]  Current/Best:    2.69/  23.01 GFLOPS | Progress: (16/20) | 17.62 s
    [Task 23/25]  Current/Best:    9.50/  23.01 GFLOPS | Progress: (20/20) | 23.24 s Done.
+
    [Task 24/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s
    [Task 24/25]  Current/Best:   10.36/  10.36 GFLOPS | Progress: (4/20) | 13.71 s
    [Task 24/25]  Current/Best:    3.46/  10.36 GFLOPS | Progress: (8/20) | 16.39 s
    [Task 24/25]  Current/Best:    1.93/  10.36 GFLOPS | Progress: (12/20) | 19.97 s
    [Task 24/25]  Current/Best:    2.52/  10.36 GFLOPS | Progress: (16/20) | 30.63 s
    [Task 24/25]  Current/Best:    6.84/  10.36 GFLOPS | Progress: (20/20) | 42.60 s
    [Task 25/25]  Current/Best:    0.00/   0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
+
    [Task 25/25]  Current/Best:    4.15/   4.58 GFLOPS | Progress: (4/20) | 13.49 s
    [Task 25/25]  Current/Best:    3.02/   6.53 GFLOPS | Progress: (8/20) | 18.44 s
    [Task 25/25]  Current/Best:    3.03/   6.53 GFLOPS | Progress: (12/20) | 21.21 s
    [Task 25/25]  Current/Best:    2.98/   9.06 GFLOPS | Progress: (16/20) | 32.16 s
    [Task 25/25]  Current/Best:    7.85/   9.06 GFLOPS | Progress: (20/20) | 35.15 s
 
 
 
@@ -666,7 +664,7 @@ Verify that the optimized model runs and produces the same results:
 
  .. code-block:: none
 
-    class='n02123045 tabby, tabby cat' with probability=0.621105
+    class='n02123045 tabby, tabby cat' with probability=0.621104
     class='n02123159 tiger cat' with probability=0.356378
     class='n02124075 Egyptian cat' with probability=0.019712
     class='n02129604 tiger, Panthera tigris' with probability=0.001215
@@ -724,8 +722,8 @@ improvement in comparing the optimized model to the unoptimized model.
 
  .. code-block:: none
 
-    optimized: {'mean': 419.83642817000145, 'median': 418.6600732000102, 'std': 3.286108283628667}
-    unoptimized: {'mean': 511.48442123999985, 'median': 511.4933623000013, 'std': 1.0384856733831964}
+    optimized: {'mean': 415.37801539000156, 'median': 415.41821250001476, 'std': 0.5183408781290263}
+    unoptimized: {'mean': 509.04423982000253, 'median': 508.817199650008, 'std': 1.8627309514183343}
 
 
 
@@ -748,7 +746,7 @@ profiling/benchmarking.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 12 minutes  35.471 seconds)
+   **Total running time of the script:** ( 12 minutes  5.646 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 b0ff58808b..3d527a15b5 100644
--- a/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
+++ b/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
@@ -274,7 +274,7 @@ device and returns the measured cost. Network overhead is excluded.
 
  .. code-block:: none
 
-    1.267e-07 secs/op
+    1.278e-07 secs/op
 
 
 
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index 2ae62b9941..a43a557350 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -270,7 +270,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
 
  .. code-block:: none
 
-    [stage(a, placeholder(a, 0x7ba3f30)), stage(b, placeholder(b, 0xed06fb0)), stage(T_add, compute(T_add, body=[a[ax0, ax1, ax2] + b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.Range(0, 10), "DataPar", ""), T.iter_var(ax2, T.Range(0, 10), "DataPar", "")], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[a[ax0, ax1, ax2] * b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.R [...]
+    [stage(a, placeholder(a, 0x112dbdf0)), stage(b, placeholder(b, 0x11382ed0)), stage(T_add, compute(T_add, body=[a[ax0, ax1, ax2] + b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.Range(0, 10), "DataPar", ""), T.iter_var(ax2, T.Range(0, 10), "DataPar", "")], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[a[ax0, ax1, ax2] * b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T [...]
 
 
 
diff --git a/docs/_sources/tutorial/sg_execution_times.rst.txt b/docs/_sources/tutorial/sg_execution_times.rst.txt
index 1f92ccdf45..df1bfc7b58 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
 =================
-**16:08.585** total execution time for **tutorial** files:
+**15:44.771** total execution time for **tutorial** files:
 
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)                 | 12:35.471 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)                 | 12:05.646 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:30.120 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:36.225 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)     | 01:01.105 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)     | 01:01.318 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)                 | 00:36.886 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)                 | 00:36.397 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)               | 00:22.431 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)               | 00:22.722 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)       | 00:01.556 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)       | 00:01.444 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)                               | 00:00.850 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)                               | 00:00.851 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.166 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.168 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_uma.py` (``uma.py``)                                             | 00:00.000 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)                           | 00:00.000 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)                             | 00:00.000 | 0.0 MB |
-+------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)   | 00:00.000 | 0.0 MB |
 +------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)                             | 00:00.000 | 0.0 MB |
++------------------------------------------------------------------------------------------+-----------+--------+
 | :ref:`sphx_glr_tutorial_install.py` (``install.py``)                                     | 00:00.000 | 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 416a4b7aa7..fa390032e7 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -389,7 +389,7 @@ compile and run this new schedule with the parallel operation applied:
 
  .. code-block:: none
 
-    parallel: 0.000008
+    parallel: 0.000006
 
 
 
@@ -444,7 +444,7 @@ factor to be the number of threads on your CPU.
 
  .. code-block:: none
 
-    vector: 0.000030
+    vector: 0.000025
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -498,10 +498,10 @@ We can now compare the different schedules
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                   numpy    6.750370000645489e-06                    1.0
-                   naive              6.6944e-06       0.991708602544729
-                parallel              8.1529e-06      1.2077708331869803
-                  vector             2.96211e-05       4.388070579415284
+                   numpy    7.357099998444028e-06                    1.0
+                   naive    6.6532999999999995e-06    0.9043373070105236
+                parallel              6.0785e-06      0.8262086965360751
+                  vector             2.47011e-05       3.357450626636052
 
 
 
@@ -922,7 +922,7 @@ matrix multiplication.
 
  .. code-block:: none
 
-    Numpy running time: 0.018356
+    Numpy running time: 0.018659
 
 
 
@@ -980,7 +980,7 @@ optimizations.
 
  .. code-block:: none
 
-    none: 3.419552
+    none: 3.437093
 
 
 
@@ -1080,7 +1080,7 @@ schedule.
 
  .. code-block:: none
 
-    blocking: 0.299880
+    blocking: 0.296410
 
 
 
@@ -1164,7 +1164,7 @@ already cache friendly from our previous optimizations.
 
  .. code-block:: none
 
-    vectorization: 0.335259
+    vectorization: 0.337642
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -1230,7 +1230,7 @@ more cache friendly.
 
  .. code-block:: none
 
-    loop permutation: 0.116245
+    loop permutation: 0.115861
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -1321,7 +1321,7 @@ optimized schedule.
 
  .. code-block:: none
 
-    array packing: 0.108269
+    array packing: 0.108483
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -1404,7 +1404,7 @@ to `C` when all the block results are ready.
 
  .. code-block:: none
 
-    block caching: 0.110388
+    block caching: 0.110566
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -1478,7 +1478,7 @@ of thread-level parallelization.
 
  .. code-block:: none
 
-    parallelization: 0.146538
+    parallelization: 0.145977
     # from tvm.script import ir as I
     # from tvm.script import tir as T
 
@@ -1548,13 +1548,13 @@ working, we can compare the results.
  .. code-block:: none
 
                 Operator                  Timing             Performance
-                    none            3.4195517812                     1.0
-                blocking            0.2998804001     0.08769582076478018
-           vectorization            0.3352589483     0.09804178142386541
-        loop permutation     0.11624544920000002     0.03399435266314546
-           array packing            0.1082689081     0.03166172499426399
-           block caching     0.11038847160000001    0.032281561638251355
-         parallelization            0.1465378427      0.0428529386528478
+                    none            3.4370929784                     1.0
+                blocking            0.2964104448     0.08623870423720162
+           vectorization            0.3376419022     0.09823473043117256
+        loop permutation     0.11586110220000001    0.033709039274792754
+           array packing             0.108482769      0.0315623608909468
+           block caching     0.11056577540000001     0.03216839814774794
+         parallelization     0.14597743680000003      0.0424711922887678
 
 
 
@@ -1596,7 +1596,7 @@ the computation for specific platforms.
 
 .. rst-class:: sphx-glr-timing
 
-   **Total running time of the script:** ( 1 minutes  1.105 seconds)
+   **Total running time of the script:** ( 1 minutes  1.318 seconds)
 
 
 .. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
diff --git a/docs/commit_hash b/docs/commit_hash
index 8d3ef291f3..f622cf81d1 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-e59d1efc68a0e1b6f1936ab09f5635386b26616e
+9d732d0faa170ba254b6decb8e69d5148d111c03
diff --git a/docs/genindex.html b/docs/genindex.html
index d7a1bc1041..20176cf126 100644
--- a/docs/genindex.html
+++ b/docs/genindex.html
@@ -3394,6 +3394,10 @@
       <li><a href="reference/api/python/auto_scheduler.html#tvm.auto_scheduler.register_workload">register_workload() (in module tvm.auto_scheduler)</a>
 </li>
       <li><a href="reference/api/python/tir.html#tvm.tir.Schedule.reindex">reindex() (tvm.tir.Schedule method)</a>
+</li>
+      <li><a href="reference/api/python/tir.html#tvm.tir.Schedule.reindex_cache_read">reindex_cache_read() (tvm.tir.Schedule method)</a>
+</li>
+      <li><a href="reference/api/python/tir.html#tvm.tir.Schedule.reindex_cache_write">reindex_cache_write() (tvm.tir.Schedule method)</a>
 </li>
       <li><a href="reference/api/python/tir.html#tvm.tir.reinterpret">reinterpret() (in module tvm.tir)</a>
 </li>
@@ -3446,11 +3450,11 @@
 </li>
       </ul></li>
       <li><a href="reference/api/python/tir.html#tvm.tir.stmt_functor.renew_defs">renew_defs() (in module tvm.tir.stmt_functor)</a>
-</li>
-      <li><a href="reference/api/python/tir.html#tvm.tir.transform.RenormalizeSplitPattern">RenormalizeSplitPattern() (in module tvm.tir.transform)</a>
 </li>
   </ul></td>
   <td style="width: 33%; vertical-align: top;"><ul>
+      <li><a href="reference/api/python/tir.html#tvm.tir.transform.RenormalizeSplitPattern">RenormalizeSplitPattern() (in module tvm.tir.transform)</a>
+</li>
       <li><a href="reference/api/python/te.html#tvm.te.Stage.reorder">reorder() (tvm.te.Stage method)</a>
 
       <ul>
diff --git a/docs/how_to/compile_models/from_darknet.html b/docs/how_to/compile_models/from_darknet.html
index 14b3a8ad34..d919bfcfa0 100644
--- a/docs/how_to/compile_models/from_darknet.html
+++ b/docs/how_to/compile_models/from_darknet.html
@@ -590,7 +590,7 @@ 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  20.865 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  19.259 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 aa6e8c0f9d..04ce7dd60b 100644
--- a/docs/how_to/compile_models/from_keras.html
+++ b/docs/how_to/compile_models/from_keras.html
@@ -511,7 +511,7 @@ Tensorflow is also required since it’s used as the default backend of keras.</
 <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 959ms/step
+1/1 [==============================] - 1s 928ms/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 ee3d7fa07f..ef11d96828 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -444,7 +444,7 @@
 <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.zip0198aed7-cf6c-4692-b757-fb9d0313d96c 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.zip778ab4bb-0685-469e-bc2a-98cdd2d04766 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 84b805a267..a6dc9e2b89 100644
--- a/docs/how_to/compile_models/from_oneflow.html
+++ b/docs/how_to/compile_models/from_oneflow.html
@@ -454,12 +454,12 @@ 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]
- 23%|##3       | 9.66M/41.5M [00:00&lt;00:00, 101MB/s]
- 47%|####6     | 19.3M/41.5M [00:00&lt;00:00, 75.2MB/s]
- 65%|######4   | 26.9M/41.5M [00:00&lt;00:00, 50.8MB/s]
- 78%|#######8  | 32.5M/41.5M [00:00&lt;00:00, 52.7MB/s]
- 92%|#########2| 38.3M/41.5M [00:00&lt;00:00, 45.6MB/s]
-100%|##########| 41.5M/41.5M [00:00&lt;00:00, 52.6MB/s]
+ 24%|##3       | 9.80M/41.5M [00:00&lt;00:00, 103MB/s]
+ 47%|####7     | 19.6M/41.5M [00:00&lt;00:00, 47.3MB/s]
+ 62%|######1   | 25.5M/41.5M [00:00&lt;00:00, 50.8MB/s]
+ 77%|#######7  | 32.0M/41.5M [00:00&lt;00:00, 46.9MB/s]
+ 89%|########9 | 37.1M/41.5M [00:00&lt;00:00, 46.9MB/s]
+100%|##########| 41.5M/41.5M [00:00&lt;00:00, 47.6MB/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 87d163f691..710abdc8ff 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -437,10 +437,10 @@ 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]
- 30%|##9       | 13.4M/44.7M [00:00&lt;00:00, 140MB/s]
- 60%|#####9    | 26.8M/44.7M [00:00&lt;00:00, 112MB/s]
- 85%|########4 | 37.8M/44.7M [00:00&lt;00:00, 108MB/s]
-100%|##########| 44.7M/44.7M [00:00&lt;00:00, 101MB/s]
+ 18%|#7        | 7.99M/44.7M [00:00&lt;00:00, 70.8MB/s]
+ 54%|#####3    | 24.0M/44.7M [00:00&lt;00:00, 98.9MB/s]
+ 81%|########  | 36.0M/44.7M [00:00&lt;00:00, 109MB/s]
+100%|##########| 44.7M/44.7M [00:00&lt;00:00, 99.0MB/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 d3a535fb1e..12b0a06bf9 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -654,7 +654,7 @@ 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  24.904 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  23.504 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 20a71541e9..808094ace5 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>06:42.702</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>06:34.745</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 81%" />
@@ -354,43 +354,43 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><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>01:24.904</p></td>
+<td><p>01:23.504</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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>01:20.865</p></td>
+<td><p>01:19.259</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>00:56.107</p></td>
+<td><p>00:54.967</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>00:38.249</p></td>
+<td><p>00:37.468</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><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>00:32.175</p></td>
+<td><p>00:31.655</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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>00:31.502</p></td>
+<td><p>00:30.554</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>00:27.377</p></td>
+<td><p>00:27.004</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>00:25.672</p></td>
+<td><p>00:25.776</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>00:23.137</p></td>
+<td><p>00:21.869</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>00:02.714</p></td>
+<td><p>00:02.689</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
diff --git a/docs/how_to/deploy_models/deploy_model_on_adreno.html b/docs/how_to/deploy_models/deploy_model_on_adreno.html
index 255deb319b..e8e24b2b12 100644
--- a/docs/how_to/deploy_models/deploy_model_on_adreno.html
+++ b/docs/how_to/deploy_models/deploy_model_on_adreno.html
@@ -925,9 +925,10 @@ Top5 predictions:
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
- 2545.8674    2543.4880    2563.2436    2540.1589      6.9047
+ 3336.0923    3336.1272    3339.9720    3332.7477      1.8435
 </pre></div>
 </div>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  3.975 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-model-on-adreno-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/2387d8448da213eb625e6b3d916327d4/deploy_model_on_adreno.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_model_on_adreno.py</span></code></a></p>
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 79f684d735..134adab1cd 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -667,7 +667,7 @@ to the remote android device.</p>
 Evaluate inference time cost...
 Execution time summary:
  mean (ms)   median (ms)    max (ms)     min (ms)     std (ms)
-  16.1492      16.1049      16.6712      15.8424       0.2481
+  16.4196      16.4974      16.9973      15.5352       0.3802
 </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 b6b58e9bed..2368ea38b4 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -459,22 +459,26 @@ 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%|5         | 8.54M/170M [00:00&lt;00:01, 89.5MB/s]
- 13%|#2        | 21.6M/170M [00:00&lt;00:01, 117MB/s]
- 19%|#9        | 32.8M/170M [00:00&lt;00:01, 101MB/s]
- 28%|##7       | 47.1M/170M [00:00&lt;00:01, 118MB/s]
- 35%|###4      | 58.6M/170M [00:00&lt;00:01, 105MB/s]
- 41%|####1     | 70.0M/170M [00:00&lt;00:00, 109MB/s]
- 48%|####7     | 80.7M/170M [00:00&lt;00:00, 96.4MB/s]
- 53%|#####3    | 90.3M/170M [00:00&lt;00:00, 89.4MB/s]
- 61%|######1   | 104M/170M [00:01&lt;00:00, 101MB/s]
- 69%|######8   | 117M/170M [00:01&lt;00:00, 110MB/s]
- 75%|#######5  | 128M/170M [00:01&lt;00:00, 107MB/s]
- 81%|########1 | 138M/170M [00:01&lt;00:00, 105MB/s]
- 87%|########7 | 148M/170M [00:01&lt;00:00, 103MB/s]
- 93%|#########3| 158M/170M [00:01&lt;00:00, 102MB/s]
- 99%|#########8| 168M/170M [00:01&lt;00:00, 76.6MB/s]
-100%|##########| 170M/170M [00:01&lt;00:00, 97.3MB/s]
+  5%|4         | 7.99M/170M [00:00&lt;00:03, 53.5MB/s]
+  9%|9         | 16.1M/170M [00:00&lt;00:02, 68.4MB/s]
+ 14%|#3        | 23.3M/170M [00:00&lt;00:02, 71.2MB/s]
+ 19%|#8        | 32.0M/170M [00:00&lt;00:01, 73.3MB/s]
+ 27%|##6       | 45.3M/170M [00:00&lt;00:01, 95.1MB/s]
+ 33%|###2      | 56.0M/170M [00:00&lt;00:01, 89.9MB/s]
+ 38%|###8      | 64.8M/170M [00:00&lt;00:01, 69.9MB/s]
+ 46%|####6     | 78.1M/170M [00:01&lt;00:01, 86.7MB/s]
+ 51%|#####1    | 87.4M/170M [00:01&lt;00:01, 84.6MB/s]
+ 57%|#####6    | 96.1M/170M [00:01&lt;00:01, 69.9MB/s]
+ 61%|######1   | 104M/170M [00:01&lt;00:00, 70.9MB/s]
+ 66%|######5   | 112M/170M [00:01&lt;00:00, 74.1MB/s]
+ 70%|#######   | 120M/170M [00:01&lt;00:00, 74.7MB/s]
+ 75%|#######4  | 127M/170M [00:01&lt;00:00, 61.1MB/s]
+ 79%|#######8  | 133M/170M [00:01&lt;00:00, 61.9MB/s]
+ 83%|########2 | 140M/170M [00:02&lt;00:00, 57.1MB/s]
+ 87%|########6 | 148M/170M [00:02&lt;00:00, 61.9MB/s]
+ 92%|#########2| 156M/170M [00:02&lt;00:00, 69.4MB/s]
+ 96%|#########6| 163M/170M [00:02&lt;00:00, 63.8MB/s]
+100%|##########| 170M/170M [00:02&lt;00:00, 70.8MB/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=& [...]
@@ -572,7 +576,7 @@ 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  37.492 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes  33.293 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 ebaa75867b..11fcc01f75 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -500,8 +500,8 @@ 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]
- 72%|#######2  | 9.79M/13.6M [00:00&lt;00:00, 103MB/s]
-100%|##########| 13.6M/13.6M [00:00&lt;00:00, 118MB/s]
+ 59%|#####8    | 7.99M/13.6M [00:00&lt;00:00, 63.8MB/s]
+100%|##########| 13.6M/13.6M [00:00&lt;00:00, 90.4MB/s]
 </pre></div>
 </div>
 </div>
@@ -592,7 +592,7 @@ 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)
-  90.5422      90.5458      91.3528      90.1322       0.2162
+  90.4924      90.1736      105.9132     89.9654       1.6432
 </pre></div>
 </div>
 <div class="admonition note">
@@ -631,7 +631,7 @@ 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  17.475 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  16.438 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 1bf0e88485..d025159b19 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -585,7 +585,7 @@ 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)
-  120.6696     120.5876     128.2718     119.6104      0.9786
+  121.1309     121.1470     121.9582     120.3322      0.3616
 </pre></div>
 </div>
 <div class="admonition note">
@@ -613,7 +613,7 @@ 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  31.722 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes  36.366 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 9df8f4083a..d474bcd802 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -526,7 +526,7 @@ 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  34.853 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  31.299 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 4b7910ebcc..739bca701b 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -468,24 +468,25 @@ 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]
-  4%|3         | 5000/132723 [00:00&lt;00:02, 49996.95KB/s]
-  9%|9         | 12265/132723 [00:00&lt;00:01, 63320.20KB/s]
- 15%|#4        | 19793/132723 [00:00&lt;00:01, 68777.13KB/s]
- 20%|##        | 27158/132723 [00:00&lt;00:01, 70698.74KB/s]
- 26%|##5       | 34410/132723 [00:00&lt;00:01, 71321.28KB/s]
- 31%|###1      | 41754/132723 [00:00&lt;00:01, 72038.61KB/s]
- 38%|###7      | 49792/132723 [00:00&lt;00:01, 74762.70KB/s]
- 44%|####3     | 57759/132723 [00:00&lt;00:00, 76321.19KB/s]
- 49%|####9     | 65688/132723 [00:00&lt;00:00, 77247.43KB/s]
- 55%|#####5    | 73613/132723 [00:01&lt;00:00, 77861.54KB/s]
- 61%|######1   | 81568/132723 [00:01&lt;00:00, 78376.65KB/s]
- 67%|######7   | 89487/132723 [00:01&lt;00:00, 78615.64KB/s]
- 73%|#######3  | 97455/132723 [00:01&lt;00:00, 78934.85KB/s]
- 79%|#######9  | 105349/132723 [00:01&lt;00:00, 78816.63KB/s]
- 85%|########5 | 113231/132723 [00:01&lt;00:00, 77704.92KB/s]
- 91%|#########1| 121005/132723 [00:01&lt;00:00, 76952.44KB/s]
- 97%|#########6| 128704/132723 [00:01&lt;00:00, 76767.77KB/s]
-100%|##########| 132723/132723 [00:01&lt;00:00, 75349.98KB/s]
+  1%|          | 1287/132723 [00:00&lt;00:11, 11746.58KB/s]
+  5%|4         | 6075/132723 [00:00&lt;00:03, 32195.06KB/s]
+ 10%|#         | 13313/132723 [00:00&lt;00:02, 50151.41KB/s]
+ 16%|#5        | 21235/132723 [00:00&lt;00:01, 61477.69KB/s]
+ 22%|##1       | 29162/132723 [00:00&lt;00:01, 67838.64KB/s]
+ 28%|##7       | 37037/132723 [00:00&lt;00:01, 71524.28KB/s]
+ 34%|###4      | 45141/132723 [00:00&lt;00:01, 74619.84KB/s]
+ 40%|###9      | 52798/132723 [00:00&lt;00:01, 75233.16KB/s]
+ 46%|####5     | 60637/132723 [00:00&lt;00:00, 76215.30KB/s]
+ 52%|#####1    | 68393/132723 [00:01&lt;00:00, 76624.85KB/s]
+ 57%|#####7    | 76148/132723 [00:01&lt;00:00, 76905.73KB/s]
+ 63%|######3   | 83844/132723 [00:01&lt;00:00, 76866.09KB/s]
+ 69%|######9   | 91671/132723 [00:01&lt;00:00, 77289.32KB/s]
+ 75%|#######4  | 99403/132723 [00:01&lt;00:00, 77274.47KB/s]
+ 81%|########  | 107265/132723 [00:01&lt;00:00, 77674.88KB/s]
+ 87%|########6 | 115129/132723 [00:01&lt;00:00, 77964.58KB/s]
+ 93%|#########2| 122929/132723 [00:01&lt;00:00, 77970.50KB/s]
+ 99%|#########8| 130795/132723 [00:01&lt;00:00, 78176.64KB/s]
+100%|##########| 132723/132723 [00:01&lt;00:00, 72366.44KB/s]
 </pre></div>
 </div>
 <p>Create TVM runtime and do inference
@@ -524,7 +525,7 @@ 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  48.761 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  45.424 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 ee351ba7b6..643b02a72b 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>15:24.431</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>15:25.837</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 86%" />
@@ -354,39 +354,39 @@
 </colgroup>
 <tbody>
 <tr class="row-odd"><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>03:48.761</p></td>
+<td><p>03:45.424</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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>03:37.492</p></td>
+<td><p>03:33.293</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>02:31.722</p></td>
+<td><p>02:36.366</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>01:34.853</p></td>
+<td><p>01:31.299</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>01:17.475</p></td>
+<td><p>01:16.438</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="deploy_model_on_adreno.html#sphx-glr-how-to-deploy-models-deploy-model-on-adreno-py"><span class="std std-ref">Deploy the Pretrained Model on Adreno</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_adreno.py</span></code>)</p></td>
-<td><p>00:54.814</p></td>
+<td><p>01:03.975</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><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>00:43.000</p></td>
+<td><p>00:42.309</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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>00:28.356</p></td>
+<td><p>00:28.488</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><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>00:27.951</p></td>
+<td><p>00:28.239</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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 83f499a215..3f3a377bb8 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -624,7 +624,7 @@ 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.zip3e28c82f-ca1f-4e34-aaed-17f934dcf6d2 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.zip61858a39-05f3-4d4b-bcb1-53d26f89705d 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 40f74dc6ef..791a3ead97 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>00:55.057</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:53.636</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -354,19 +354,19 @@
 </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>00:51.172</p></td>
+<td><p>00:49.872</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>00:02.774</p></td>
+<td><p>00:02.685</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>00:01.102</p></td>
+<td><p>00:01.072</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>
-<td><p>00:00.009</p></td>
+<td><p>00:00.008</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index d9ec452887..5ad494ef79 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -531,10 +531,10 @@ 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: 22302us [22302us] (48.68%; 48.68%)
-FoldScaleAxis: 23510us [8us] (51.32%; 51.32%)
-        FoldConstant: 23501us [1725us] (51.30%; 99.96%)
-                InferType: 21777us [21777us] (47.53%; 92.66%)
+InferType: 22409us [22409us] (47.83%; 47.83%)
+FoldScaleAxis: 24439us [7us] (52.17%; 52.17%)
+        FoldConstant: 24432us [1696us] (52.15%; 99.97%)
+                InferType: 22736us [22736us] (48.53%; 93.06%)
 </pre></div>
 </div>
 </div>
@@ -556,10 +556,10 @@ 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: 21885us [21885us] (48.30%; 48.30%)
-FoldScaleAxis: 23427us [6us] (51.70%; 51.70%)
-        FoldConstant: 23421us [1749us] (51.69%; 99.97%)
-                InferType: 21672us [21672us] (47.83%; 92.53%)
+InferType: 21571us [21571us] (48.43%; 48.43%)
+FoldScaleAxis: 22968us [5us] (51.57%; 51.57%)
+        FoldConstant: 22963us [1704us] (51.56%; 99.98%)
+                InferType: 21259us [21259us] (47.73%; 92.58%)
 </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 4c405a3560..f25bc8e5b6 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -580,7 +580,7 @@ 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: 54.161792 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.239936 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 7e590169c9..db6b37aa4e 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -862,7 +862,7 @@ 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: 7.498425 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 6.683037 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 8d4b4d6634..f16ffde879 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -477,8 +477,8 @@ 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: 0.018414
-Baseline: 3.427426
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.018453
+Baseline: 3.409857
 </pre></div>
 </div>
 <p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -537,7 +537,7 @@ 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: 0.303755
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.295121
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -594,7 +594,7 @@ 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: 0.349998
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.332745
 </pre></div>
 </div>
 <p>Here is the generated IR after vectorization.</p>
@@ -649,7 +649,7 @@ 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: 0.115294
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.114450
 </pre></div>
 </div>
 <p>Here is the generated IR after loop permutation.</p>
@@ -726,7 +726,7 @@ flattening.</p>
 <span class="nb">print</span><span class="p">(</span><span class="s2">&quot;Opt4: </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>Opt4: 0.107893
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.109112
 </pre></div>
 </div>
 <p>Here is the generated IR after array packing.</p>
@@ -804,7 +804,7 @@ 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: 0.110806
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.111139
 </pre></div>
 </div>
 <p>Here is the generated IR after blocking.</p>
@@ -884,7 +884,7 @@ class Module:
 <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: 0.146479
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.146767
 </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 6fe604016f..63aa68ab81 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>00:35.107</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:34.737</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 83%" />
@@ -354,15 +354,15 @@
 </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>00:32.496</p></td>
+<td><p>00:32.165</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>00:01.499</p></td>
+<td><p>00:01.500</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>00:01.112</p></td>
+<td><p>00:01.072</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 69af211586..7e4031653a 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>09:59.540</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>09:57.787</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 85%" />
@@ -354,27 +354,27 @@
 </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>06:06.168</p></td>
+<td><p>06:08.646</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>01:42.432</p></td>
+<td><p>01:41.629</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>01:07.705</p></td>
+<td><p>01:07.266</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>00:35.267</p></td>
+<td><p>00:32.690</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>00:14.289</p></td>
+<td><p>00:14.002</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>00:13.678</p></td>
+<td><p>00:13.554</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 f6aed3df23..370d43e1b1 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
@@ -511,11 +511,11 @@ class Module:
     def main(data: T.Buffer((1, 512, 7, 7), &quot;float32&quot;), kernel: T.Buffer((512, 512, 3, 3), &quot;float32&quot;), bias: T.Buffer((1, 512, 1, 1), &quot;float32&quot;), compute: T.Buffer((1, 512, 7, 7), &quot;float32&quot;)):
         T.func_attr({&quot;from_legacy_te_schedule&quot;: True, &quot;global_symbol&quot;: &quot;main&quot;, &quot;tir.noalias&quot;: True})
         blockIdx_x = T.launch_thread(&quot;blockIdx.x&quot;, 28)
-        conv2d_nchw = T.allocate([7], &quot;float32&quot;, &quot;local&quot;)
-        pad_temp_shared = T.allocate([216], &quot;float32&quot;, &quot;shared&quot;)
-        kernel_shared = T.allocate([9216], &quot;float32&quot;, &quot;shared&quot;)
-        threadIdx_x = T.launch_thread(&quot;threadIdx.x&quot;, 128)
-        conv2d_nchw_1 = T.Buffer((7,), data=conv2d_nchw, scope=&quot;local&quot;, align=16)
+        conv2d_nchw = T.allocate([14], &quot;float32&quot;, &quot;local&quot;)
+        pad_temp_shared = T.allocate([72], &quot;float32&quot;, &quot;shared&quot;)
+        kernel_shared = T.allocate([3072], &quot;float32&quot;, &quot;shared&quot;)
+        threadIdx_x = T.launch_thread(&quot;threadIdx.x&quot;, 64)
+        conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope=&quot;local&quot;, align=32)
         conv2d_nchw_1[0] = T.float32(0)
         conv2d_nchw_1[1] = T.float32(0)
         conv2d_nchw_1[2] = T.float32(0)
@@ -523,563 +523,466 @@ class Module:
         conv2d_nchw_1[4] = T.float32(0)
         conv2d_nchw_1[5] = T.float32(0)
         conv2d_nchw_1[6] = T.float32(0)
-        for rc_outer_outer in range(64):
-            cse_var_1: T.int32 = rc_outer_outer * 72
+        conv2d_nchw_1[7] = T.float32(0)
+        conv2d_nchw_1[8] = T.float32(0)
+        conv2d_nchw_1[9] = T.float32(0)
+        conv2d_nchw_1[10] = T.float32(0)
+        conv2d_nchw_1[11] = T.float32(0)
+        conv2d_nchw_1[12] = T.float32(0)
+        conv2d_nchw_1[13] = T.float32(0)
+        for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
+            cse_var_2: T.int32 = rc_outer_outer * 72
+            cse_var_1: T.int32 = ry_outer_outer * 3
+            pad_temp_shared_1 = T.Buffer((72,), data=pad_temp_shared, scope=&quot;shared&quot;)
+            with T.launch_thread(&quot;threadIdx.x&quot;, 64) as threadIdx_x_1:
+                data_1 = T.Buffer((25088,), data=data.data)
+                if T.likely(threadIdx_x_1 &lt; 18):
+                    pad_temp_shared_1[threadIdx_x_1 * 4] = T.if_then_else(1 &lt;= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 &lt; 8 and 1 &lt;= threadIdx_x_1 * 4 % 9 and threadIdx_x_1 * 4 % 9 &lt; 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 * 4 // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + threadIdx_x_1 * 4 % 9 - 8], T.float32(0))
+                if T.likely(threadIdx_x_1 &lt; 18):
+                    pad_temp_shared_1[threadIdx_x_1 * 4 + 1] = T.if_then_else(1 &lt;= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 &lt; 8 and 1 &lt;= (threadIdx_x_1 * 4 + 1) % 9 and (threadIdx_x_1 * 4 + 1) % 9 &lt; 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 1) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 1) % 9 - 8], T.float32(0))
+                if T.likely(threadIdx_x_1 &lt; 18):
+                    pad_temp_shared_1[threadIdx_x_1 * 4 + 2] = T.if_then_else(1 &lt;= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 &lt; 8 and 1 &lt;= (threadIdx_x_1 * 4 + 2) % 9 and (threadIdx_x_1 * 4 + 2) % 9 &lt; 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 2) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 2) % 9 - 8], T.float32(0))
+                if T.likely(threadIdx_x_1 &lt; 18):
+                    pad_temp_shared_1[threadIdx_x_1 * 4 + 3] = T.if_then_else(1 &lt;= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 &lt; 8 and 1 &lt;= (threadIdx_x_1 * 4 + 3) % 9 and (threadIdx_x_1 * 4 + 3) % 9 &lt; 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 3) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 3) % 9 - 8], T.float32(0))
             threadIdx_x_1 = T.env_thread(&quot;threadIdx.x&quot;)
-            pad_temp_shared_1 = T.Buffer((216,), data=pad_temp_shared, scope=&quot;shared&quot;)
-            data_1 = T.Buffer((25088,), data=data.data)
-            with T.launch_thread(threadIdx_x_1, 128):
-                pad_temp_shared_1[threadIdx_x_1] = T.if_then_else(3 &lt;= threadIdx_x_1 % 27 and threadIdx_x_1 % 27 &lt; 24 and 1 &lt;= blockIdx_x % 7 + threadIdx_x_1 % 3 and blockIdx_x % 7 + threadIdx_x_1 % 3 &lt; 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 // 27 * 49 + threadIdx_x_1 % 27 // 3 * 7 + blockIdx_x % 7 + threadIdx_x_1 % 3 - 8], T.float32(0))
-            with T.launch_thread(threadIdx_x_1, 128):
-                if T.likely(threadIdx_x_1 &lt; 88):
-                    pad_temp_shared_1[threadIdx_x_1 + 128] = T.if_then_else(3 &lt;= (threadIdx_x_1 + 20) % 27 and (threadIdx_x_1 + 20) % 27 &lt; 24 and 1 &lt;= blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 and blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 &lt; 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 + 128) // 27 * 49 + (threadIdx_x_1 + 20) % 27 // 3 * 7 + blockIdx_x % 7 + (threadIdx_x_1 + 2) % 3 - 8], T.float32(0))
-            threadIdx_x_2 = T.env_thread(&quot;threadIdx.x&quot;)
-            kernel_shared_1 = T.Buffer((9216,), data=kernel_shared, scope=&quot;shared&quot;)
+            kernel_shared_1 = T.Buffer((3072,), data=kernel_shared, scope=&quot;shared&quot;)
             kernel_1 = T.Buffer((2359296,), data=kernel.data)
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4:threadIdx_x_2 * 4 + 4] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4:blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 512:threadIdx_x_2 * 4 + 512 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 512) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 512) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 1024:threadIdx_x_2 * 4 + 1024 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1024) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1024) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 1536:threadIdx_x_2 * 4 + 1536 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1536) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 1536) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 2048:threadIdx_x_2 * 4 + 2048 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2048) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2048) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 2560:threadIdx_x_2 * 4 + 2560 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2560) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 2560) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 3072:threadIdx_x_2 * 4 + 3072 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3072) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3072) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 3584:threadIdx_x_2 * 4 + 3584 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3584) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 3584) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 4096:threadIdx_x_2 * 4 + 4096 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 4096) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 4096) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 4608:threadIdx_x_2 * 4 + 4608 + 4] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 294912:blockIdx_x // 7 * 589824 + threadIdx_x_2 // 18 * 4608 + cse_var_1 + threadIdx_x_2 % 18 * 4 + 294912 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 5120:threadIdx_x_2 * 4 + 5120 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5120) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5120) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 8) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 5632:threadIdx_x_2 * 4 + 5632 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5632) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 5632) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 16) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 6144:threadIdx_x_2 * 4 + 6144 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6144) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6144) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 24) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 6656:threadIdx_x_2 * 4 + 6656 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6656) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 6656) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 32) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 7168:threadIdx_x_2 * 4 + 7168 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7168) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7168) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 40) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 7680:threadIdx_x_2 * 4 + 7680 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7680) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 7680) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 48) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 8192:threadIdx_x_2 * 4 + 8192 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8192) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8192) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 56) % 72 + 4]
-            with T.launch_thread(threadIdx_x_2, 128):
-                kernel_shared_1[threadIdx_x_2 * 4 + 8704:threadIdx_x_2 * 4 + 8704 + 4] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8704) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72:blockIdx_x // 7 * 589824 + (threadIdx_x_2 * 4 + 8704) // 72 * 4608 + cse_var_1 + (threadIdx_x_2 * 4 + 64) % 72 + 4]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[81] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[82] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[83] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[84] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[85] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[86] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[84] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[85] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[86] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[87] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[88] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[89] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[90] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[91] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[92] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[93] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[94] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[95] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[75] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[76] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[77] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[96] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[97] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[98] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[102] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[103] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[104] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 72]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 72 + 1]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 72 + 2]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 72 + 3]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 72 + 4]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 72 + 5]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 72 + 6]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 72 + 7]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 72 + 8]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 72 + 9]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 72 + 10]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 72 + 11]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 72 + 12]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 72 + 13]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 72 + 14]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 72 + 15]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 72 + 16]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 72 + 17]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[72] * kernel_shared_1[threadIdx_x * 72 + 18]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[73] * kernel_shared_1[threadIdx_x * 72 + 19]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[74] * kernel_shared_1[threadIdx_x * 72 + 20]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[75] * kernel_shared_1[threadIdx_x * 72 + 21]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[76] * kernel_shared_1[threadIdx_x * 72 + 22]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[77] * kernel_shared_1[threadIdx_x * 72 + 23]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[78] * kernel_shared_1[threadIdx_x * 72 + 24]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[79] * kernel_shared_1[threadIdx_x * 72 + 25]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[80] * kernel_shared_1[threadIdx_x * 72 + 26]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[99] * kernel_shared_1[threadIdx_x * 72 + 27]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[100] * kernel_shared_1[threadIdx_x * 72 + 28]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[101] * kernel_shared_1[threadIdx_x * 72 + 29]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[102] * kernel_shared_1[threadIdx_x * 72 + 30]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[103] * kernel_shared_1[threadIdx_x * 72 + 31]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[104] * kernel_shared_1[threadIdx_x * 72 + 32]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[105] * kernel_shared_1[threadIdx_x * 72 + 33]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[106] * kernel_shared_1[threadIdx_x * 72 + 34]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[107] * kernel_shared_1[threadIdx_x * 72 + 35]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[108] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[109] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[110] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[111] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[112] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[113] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[135] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[136] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[137] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[138] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[139] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[140] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[162] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[163] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[164] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[165] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[166] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[167] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[189] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[190] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[191] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[192] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[193] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[194] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[111] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[112] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[113] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[138] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[139] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[140] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[165] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[166] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[167] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[192] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[193] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[194] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[114] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[115] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[116] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[141] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[142] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[143] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[168] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[169] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[170] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[195] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[196] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[197] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[117] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[118] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[119] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[144] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[145] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[146] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[171] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[172] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[173] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[198] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[199] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[200] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[120] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[121] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[122] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[147] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[148] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[149] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[174] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[175] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[176] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[201] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[202] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[203] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[123] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[124] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[125] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[129] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[130] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[131] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[150] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[151] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[152] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[156] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[157] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[158] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[177] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[178] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[179] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[183] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[184] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[185] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[204] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[205] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[206] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[210] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[211] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[212] * kernel_shared_1[threadIdx_x * 72 + 71]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[126] * kernel_shared_1[threadIdx_x * 72 + 36]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[127] * kernel_shared_1[threadIdx_x * 72 + 37]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[128] * kernel_shared_1[threadIdx_x * 72 + 38]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[129] * kernel_shared_1[threadIdx_x * 72 + 39]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[130] * kernel_shared_1[threadIdx_x * 72 + 40]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[131] * kernel_shared_1[threadIdx_x * 72 + 41]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[132] * kernel_shared_1[threadIdx_x * 72 + 42]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[133] * kernel_shared_1[threadIdx_x * 72 + 43]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[134] * kernel_shared_1[threadIdx_x * 72 + 44]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[153] * kernel_shared_1[threadIdx_x * 72 + 45]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[154] * kernel_shared_1[threadIdx_x * 72 + 46]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[155] * kernel_shared_1[threadIdx_x * 72 + 47]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[156] * kernel_shared_1[threadIdx_x * 72 + 48]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[157] * kernel_shared_1[threadIdx_x * 72 + 49]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[158] * kernel_shared_1[threadIdx_x * 72 + 50]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[159] * kernel_shared_1[threadIdx_x * 72 + 51]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[160] * kernel_shared_1[threadIdx_x * 72 + 52]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[161] * kernel_shared_1[threadIdx_x * 72 + 53]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[180] * kernel_shared_1[threadIdx_x * 72 + 54]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[181] * kernel_shared_1[threadIdx_x * 72 + 55]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[182] * kernel_shared_1[threadIdx_x * 72 + 56]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[183] * kernel_shared_1[threadIdx_x * 72 + 57]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[184] * kernel_shared_1[threadIdx_x * 72 + 58]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[185] * kernel_shared_1[threadIdx_x * 72 + 59]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[186] * kernel_shared_1[threadIdx_x * 72 + 60]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[187] * kernel_shared_1[threadIdx_x * 72 + 61]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[188] * kernel_shared_1[threadIdx_x * 72 + 62]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[207] * kernel_shared_1[threadIdx_x * 72 + 63]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[208] * kernel_shared_1[threadIdx_x * 72 + 64]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[209] * kernel_shared_1[threadIdx_x * 72 + 65]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[210] * kernel_shared_1[threadIdx_x * 72 + 66]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[211] * kernel_shared_1[threadIdx_x * 72 + 67]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[212] * kernel_shared_1[threadIdx_x * 72 + 68]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[213] * kernel_shared_1[threadIdx_x * 72 + 69]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[214] * kernel_shared_1[threadIdx_x * 72 + 70]
-            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[215] * kernel_shared_1[threadIdx_x * 72 + 71]
-        for i2_inner in range(7):
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+            with T.launch_thread(threadIdx_x_1, 64):
+                kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+            conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+            conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
+        for i1_inner, i3_inner in T.grid(2, 7):
             compute_1 = T.Buffer((25088,), data=compute.data)
             bias_1 = T.Buffer((512,), data=bias.data)
-            compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 49 + i2_inner * 7 + blockIdx_x % 7] = T.max(conv2d_nchw_1[i2_inner] + bias_1[blockIdx_x // 7 * 128 + threadIdx_x], T.float32(0))
+            compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 98 + i1_inner * 49 + blockIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x // 7 * 128 + threadIdx_x * 2 + i1_inner], T.float32(0))
 </pre></div>
 </div>
 </div>
@@ -1113,7 +1016,7 @@ class Module:
 <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: 0.198 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.364 ms
 </pre></div>
 </div>
 </div>
@@ -1143,34 +1046,34 @@ conv2d_nchw_nn_o_o_i, conv2d_nchw_nn_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o
 conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
 conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
 conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
-conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
-conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=128)
+conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=2)
+conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=64)
 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=7)
+conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
 conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
 conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
 conv2d_nchw_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_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
 conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=1)
 conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
-conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
-conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=2)
-conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=3)
+conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=2)
+conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=4)
+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_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=3)
-conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+conv2d_nchw_rx_o_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=3)
 s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2d_nc [...]
 compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
 compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
 compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=128)
+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=64)
 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_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
 compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
 compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
-compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
+compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=7)
 compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=1)
 compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
 s[compute].reorder(compute_i0_o_o_o, compute_i1_o_o_o, compute_i2_o_o_o, compute_i3_o_o_o, compute_i0_o_o_i, compute_i1_o_o_i, compute_i2_o_o_i, compute_i3_o_o_i, compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i, compute_i0_i, compute_i1_i, compute_i2_i, compute_i3_i)
@@ -1189,16 +1092,16 @@ 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=128)
+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=64)
 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=4)
 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=128)
+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=64)
 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:
@@ -1216,10 +1119,10 @@ CUDA source code:
   #define int64_t long long
   #define uint64_t unsigned long long
 #endif
-extern &quot;C&quot; __global__ void __launch_bounds__(128) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
-  float conv2d_nchw[7];
-  __shared__ float pad_temp_shared[216];
-  __shared__ float kernel_shared[9216];
+extern &quot;C&quot; __global__ void __launch_bounds__(64) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+  float conv2d_nchw[14];
+  __shared__ float pad_temp_shared[72];
+  __shared__ float kernel_shared[3072];
   conv2d_nchw[0] = 0.000000e+00f;
   conv2d_nchw[1] = 0.000000e+00f;
   conv2d_nchw[2] = 0.000000e+00f;
@@ -1227,538 +1130,419 @@ extern &quot;C&quot; __global__ void __launch_bounds__(128) default_function_ker
   conv2d_nchw[4] = 0.000000e+00f;
   conv2d_nchw[5] = 0.000000e+00f;
   conv2d_nchw[6] = 0.000000e+00f;
+  conv2d_nchw[7] = 0.000000e+00f;
+  conv2d_nchw[8] = 0.000000e+00f;
+  conv2d_nchw[9] = 0.000000e+00f;
+  conv2d_nchw[10] = 0.000000e+00f;
+  conv2d_nchw[11] = 0.000000e+00f;
+  conv2d_nchw[12] = 0.000000e+00f;
+  conv2d_nchw[13] = 0.000000e+00f;
   for (int rc_outer_outer = 0; rc_outer_outer &lt; 64; ++rc_outer_outer) {
-    __syncthreads();
-    pad_temp_shared[((int)threadIdx.x)] = (((((3 &lt;= (((int)threadIdx.x) % 27)) &amp;&amp; ((((int)threadIdx.x) % 27) &lt; 24)) &amp;&amp; (1 &lt;= ((((int)blockIdx.x) % 7) + (((int)threadIdx.x) % 3)))) &amp;&amp; (((((int)blockIdx.x) % 7) + (((int)threadIdx.x) % 3)) &lt; 8)) ? data[((((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 27) * 49)) + (((((int)threadIdx.x) % 27) / 3) * 7)) + (((int)blockIdx.x) % 7)) + (((int)threadIdx.x) % 3)) - 8)] : 0.000000e+00f);
-    if (((int)threadIdx.x) &lt; 88) {
-      pad_temp_shared[(((int)threadIdx.x) + 128)] = (((((3 &lt;= ((((int)threadIdx.x) + 20) % 27)) &amp;&amp; (((((int)threadIdx.x) + 20) % 27) &lt; 24)) &amp;&amp; (1 &lt;= ((((int)blockIdx.x) % 7) + ((((int)threadIdx.x) + 2) % 3)))) &amp;&amp; (((((int)blockIdx.x) % 7) + ((((int)threadIdx.x) + 2) % 3)) &lt; 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 128) / 27) * 49)) + ((((((int)threadIdx.x) + 20) % 27) / 3) * 7)) + (((int)blockIdx.x) % 7)) + ((((int)threadIdx.x)  [...]
+    for (int ry_outer_outer = 0; ry_outer_outer &lt; 3; ++ry_outer_outer) {
+      __syncthreads();
+      if (((int)threadIdx.x) &lt; 18) {
+        pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 &lt;= (ry_outer_outer + (((int)blockIdx.x) % 7))) &amp;&amp; ((ry_outer_outer + (((int)blockIdx.x) % 7)) &lt; 8)) &amp;&amp; (1 &lt;= ((((int)threadIdx.x) * 4) % 9))) &amp;&amp; (((((int)threadIdx.x) * 4) % 9) &lt; 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
+      }
+      if (((int)threadIdx.x) &lt; 18) {
+        pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 &lt;= (ry_outer_outer + (((int)blockIdx.x) % 7))) &amp;&amp; ((ry_outer_outer + (((int)blockIdx.x) % 7)) &lt; 8)) &amp;&amp; (1 &lt;= (((((int)threadIdx.x) * 4) + 1) % 9))) &amp;&amp; ((((((int)threadIdx.x) * 4) + 1) % 9) &lt; 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
+      }
+      if (((int)threadIdx.x) &lt; 18) {
+        pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 &lt;= (ry_outer_outer + (((int)blockIdx.x) % 7))) &amp;&amp; ((ry_outer_outer + (((int)blockIdx.x) % 7)) &lt; 8)) &amp;&amp; (1 &lt;= (((((int)threadIdx.x) * 4) + 2) % 9))) &amp;&amp; ((((((int)threadIdx.x) * 4) + 2) % 9) &lt; 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
+      }
+      if (((int)threadIdx.x) &lt; 18) {
+        pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 &lt;= (ry_outer_outer + (((int)blockIdx.x) % 7))) &amp;&amp; ((ry_outer_outer + (((int)blockIdx.x) % 7)) &lt; 8)) &amp;&amp; (1 &lt;= (((((int)threadIdx.x) * 4) + 3) % 9))) &amp;&amp; ((((((int)threadIdx.x) * 4) + 3) % 9) &lt; 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
+      }
+      kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 64) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 128) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
+      kernel_shared[(((((((int)threadIdx.x) + 256) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 320) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
+      kernel_shared[(((((((int)threadIdx.x) + 448) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 512) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
+      kernel_shared[(((((((int)threadIdx.x) + 640) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 704) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
+      kernel_shared[(((((((int)threadIdx.x) + 832) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 896) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
+      kernel_shared[(((((((int)threadIdx.x) + 1024) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 1088) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
+      kernel_shared[(((((((int)threadIdx.x) + 1216) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 1280) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
+      kernel_shared[(((((((int)threadIdx.x) + 1408) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 1472) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
+      kernel_shared[(((((((int)threadIdx.x) + 1600) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 1664) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
+      kernel_shared[(((((((int)threadIdx.x) + 1792) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 1856) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
+      kernel_shared[(((((((int)threadIdx.x) + 1984) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 2048) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
+      kernel_shared[(((((((int)threadIdx.x) + 2176) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 2240) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
+      kernel_shared[(((((((int)threadIdx.x) + 2368) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 2432) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
+      kernel_shared[(((((((int)threadIdx.x) + 2560) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 2624) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
+      kernel_shared[(((((((int)threadIdx.x) + 2752) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 2816) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
+      kernel_shared[(((((((int)threadIdx.x) + 2944) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+      kernel_shared[(((((((int)threadIdx.x) + 3008) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+      __syncthreads();
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[1] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[2] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[4] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[5] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 48)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[0] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+      conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+      conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
     }
-    *(float4*)(kernel_shared + (((int)threadIdx.x) * 4)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 18) * 4608)) + (rc_outer_outer * 72)) + ((((int)threadIdx.x) % 18) * 4)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 512)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 512) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 8) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 1024)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 1024) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 16) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 1536)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 1536) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 24) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 2048)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 2048) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 32) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 2560)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 2560) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 40) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 3072)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 3072) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 48) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 3584)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 3584) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 56) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 4096)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 4096) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 64) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 4608)) = *(float4*)(kernel + ((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 18) * 4608)) + (rc_outer_outer * 72)) + ((((int)threadIdx.x) % 18) * 4)) + 294912));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 5120)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 5120) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 8) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 5632)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 5632) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 16) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 6144)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 6144) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 24) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 6656)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 6656) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 32) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 7168)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 7168) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 40) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 7680)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 7680) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 48) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 8192)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 8192) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 56) % 72)));
-    *(float4*)(kernel_shared + ((((int)threadIdx.x) * 4) + 8704)) = *(float4*)(kernel + (((((((int)blockIdx.x) / 7) * 589824) + ((((((int)threadIdx.x) * 4) + 8704) / 72) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) * 4) + 64) % 72)));
-    __syncthreads();
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[81] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[82] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[83] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[84] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[85] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[86] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[84] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[85] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[86] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[87] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[88] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[89] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[9] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[90] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[91] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[92] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[12] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[93] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[94] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[95] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[75] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[76] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[77] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[96] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[97] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[98] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[102] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[103] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[104] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[18] * kernel_shared[(((int)threadIdx.x) * 72)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 72) + 1)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 72) + 2)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 72) + 3)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 72) + 4)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 72) + 5)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 72) + 6)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 72) + 7)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 72) + 8)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 72) + 9)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 72) + 10)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 72) + 11)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 72) + 12)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 72) + 13)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 72) + 14)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 72) + 15)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 72) + 16)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 72) + 17)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[72] * kernel_shared[((((int)threadIdx.x) * 72) + 18)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[73] * kernel_shared[((((int)threadIdx.x) * 72) + 19)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[74] * kernel_shared[((((int)threadIdx.x) * 72) + 20)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[75] * kernel_shared[((((int)threadIdx.x) * 72) + 21)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[76] * kernel_shared[((((int)threadIdx.x) * 72) + 22)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[77] * kernel_shared[((((int)threadIdx.x) * 72) + 23)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[78] * kernel_shared[((((int)threadIdx.x) * 72) + 24)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[79] * kernel_shared[((((int)threadIdx.x) * 72) + 25)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[80] * kernel_shared[((((int)threadIdx.x) * 72) + 26)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[99] * kernel_shared[((((int)threadIdx.x) * 72) + 27)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[100] * kernel_shared[((((int)threadIdx.x) * 72) + 28)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[101] * kernel_shared[((((int)threadIdx.x) * 72) + 29)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[102] * kernel_shared[((((int)threadIdx.x) * 72) + 30)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[103] * kernel_shared[((((int)threadIdx.x) * 72) + 31)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[104] * kernel_shared[((((int)threadIdx.x) * 72) + 32)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[105] * kernel_shared[((((int)threadIdx.x) * 72) + 33)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[106] * kernel_shared[((((int)threadIdx.x) * 72) + 34)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[107] * kernel_shared[((((int)threadIdx.x) * 72) + 35)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[108] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[109] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[110] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[111] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[112] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[113] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[135] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[136] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[137] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[138] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[139] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[140] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[162] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[163] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[164] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[165] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[166] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[167] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[189] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[190] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[191] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[192] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[193] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[194] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[111] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[112] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[113] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[138] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[139] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[140] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[165] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[166] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[167] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[192] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[193] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[194] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[114] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[115] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[116] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[141] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[142] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[143] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[168] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[169] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[170] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[195] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[196] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[197] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[117] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[118] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[119] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[144] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[145] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[146] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[171] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[172] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[173] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[198] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[199] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[200] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[120] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[121] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[122] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[147] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[148] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[149] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[174] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[175] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[176] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[201] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[202] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[203] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[123] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[124] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[125] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[129] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[130] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[131] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[150] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[151] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[152] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[156] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[157] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[158] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[177] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[178] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[179] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[183] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[184] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[185] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[204] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[205] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[206] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[210] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[211] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[212] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[126] * kernel_shared[((((int)threadIdx.x) * 72) + 36)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[127] * kernel_shared[((((int)threadIdx.x) * 72) + 37)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[128] * kernel_shared[((((int)threadIdx.x) * 72) + 38)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[129] * kernel_shared[((((int)threadIdx.x) * 72) + 39)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[130] * kernel_shared[((((int)threadIdx.x) * 72) + 40)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[131] * kernel_shared[((((int)threadIdx.x) * 72) + 41)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[132] * kernel_shared[((((int)threadIdx.x) * 72) + 42)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[133] * kernel_shared[((((int)threadIdx.x) * 72) + 43)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[134] * kernel_shared[((((int)threadIdx.x) * 72) + 44)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[153] * kernel_shared[((((int)threadIdx.x) * 72) + 45)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[154] * kernel_shared[((((int)threadIdx.x) * 72) + 46)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[155] * kernel_shared[((((int)threadIdx.x) * 72) + 47)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[156] * kernel_shared[((((int)threadIdx.x) * 72) + 48)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[157] * kernel_shared[((((int)threadIdx.x) * 72) + 49)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[158] * kernel_shared[((((int)threadIdx.x) * 72) + 50)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[159] * kernel_shared[((((int)threadIdx.x) * 72) + 51)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[160] * kernel_shared[((((int)threadIdx.x) * 72) + 52)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[161] * kernel_shared[((((int)threadIdx.x) * 72) + 53)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[180] * kernel_shared[((((int)threadIdx.x) * 72) + 54)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[181] * kernel_shared[((((int)threadIdx.x) * 72) + 55)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[182] * kernel_shared[((((int)threadIdx.x) * 72) + 56)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[183] * kernel_shared[((((int)threadIdx.x) * 72) + 57)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[184] * kernel_shared[((((int)threadIdx.x) * 72) + 58)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[185] * kernel_shared[((((int)threadIdx.x) * 72) + 59)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[186] * kernel_shared[((((int)threadIdx.x) * 72) + 60)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[187] * kernel_shared[((((int)threadIdx.x) * 72) + 61)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[188] * kernel_shared[((((int)threadIdx.x) * 72) + 62)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[207] * kernel_shared[((((int)threadIdx.x) * 72) + 63)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[208] * kernel_shared[((((int)threadIdx.x) * 72) + 64)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[209] * kernel_shared[((((int)threadIdx.x) * 72) + 65)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[210] * kernel_shared[((((int)threadIdx.x) * 72) + 66)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[211] * kernel_shared[((((int)threadIdx.x) * 72) + 67)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[212] * kernel_shared[((((int)threadIdx.x) * 72) + 68)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[213] * kernel_shared[((((int)threadIdx.x) * 72) + 69)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[214] * kernel_shared[((((int)threadIdx.x) * 72) + 70)]));
-    conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[215] * kernel_shared[((((int)threadIdx.x) * 72) + 71)]));
   }
-  for (int i2_inner = 0; i2_inner &lt; 7; ++i2_inner) {
-    compute[(((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 49)) + (i2_inner * 7)) + (((int)blockIdx.x) % 7))] = max((conv2d_nchw[i2_inner] + bias[(((((int)blockIdx.x) / 7) * 128) + ((int)threadIdx.x))]), 0.000000e+00f);
+  for (int i1_inner = 0; i1_inner &lt; 2; ++i1_inner) {
+    for (int i3_inner = 0; i3_inner &lt; 7; ++i3_inner) {
+      compute[((((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 98)) + (i1_inner * 49)) + ((((int)blockIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[((((((int)blockIdx.x) / 7) * 128) + (((int)threadIdx.x) * 2)) + i1_inner)]), 0.000000e+00f);
+    }
   }
 }
 </pre></div>
@@ -1793,7 +1577,7 @@ 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> ( 6 minutes  6.168 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 6 minutes  8.646 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_cuda.html b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
index a5e517c6da..cff0eeba83 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -921,7 +921,7 @@ 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)
-   7.9169       7.9170       7.9266       7.9072       0.0079
+   7.9036       7.9046       7.9074       7.8987       0.0036
 </pre></div>
 </div>
 </div>
@@ -943,7 +943,7 @@ 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  7.705 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  7.266 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_x86.html b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
index 3b7b7b8a18..eae0e0f8ed 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -940,7 +940,7 @@ 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)
-  749.0074     749.4644     749.6283     747.9293      0.7652
+  756.5266     754.0870     762.1017     753.3910      3.9525
 </pre></div>
 </div>
 </div>
@@ -962,7 +962,7 @@ 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  42.432 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  41.629 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 7e5a51a27a..23458dfe60 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -637,74 +637,75 @@ class Module:
     @T.prim_func
     def main(placeholder: T.Buffer((128, 256), &quot;float32&quot;), placeholder_1: T.Buffer((4916, 16, 1), &quot;float32&quot;), placeholder_2: T.Buffer((4916,), &quot;int32&quot;), placeholder_3: T.Buffer((33,), &quot;int32&quot;), placeholder_4: T.Buffer((128, 512), &quot;float32&quot;), compute: T.Buffer((128, 512), &quot;float32&quot;)):
         T.func_attr({&quot;from_legacy_te_schedule&quot;: True, &quot;global_symbol&quot;: &quot;main&quot;, &quot;tir.noalias&quot;: True})
-        for i0_outer_i1_outer_fused in T.parallel(16):
-            compute_1 = T.allocate([4096], &quot;float32&quot;, &quot;global&quot;)
-            compute_2 = T.Buffer((4096,), data=compute_1)
-            for i_outer_inner, nb_j_inner in T.grid(4, 2):
-                for i_inner_init in range(32):
-                    cse_var_1: T.int32 = i_outer_inner * 1024 + i_inner_init * 32 + nb_j_inner * 16
-                    compute_2[cse_var_1] = T.float32(0)
-                    compute_2[cse_var_1 + 1] = T.float32(0)
-                    compute_2[cse_var_1 + 2] = T.float32(0)
-                    compute_2[cse_var_1 + 3] = T.float32(0)
-                    compute_2[cse_var_1 + 4] = T.float32(0)
-                    compute_2[cse_var_1 + 5] = T.float32(0)
-                    compute_2[cse_var_1 + 6] = T.float32(0)
-                    compute_2[cse_var_1 + 7] = T.float32(0)
-                    compute_2[cse_var_1 + 8] = T.float32(0)
-                    compute_2[cse_var_1 + 9] = T.float32(0)
-                    compute_2[cse_var_1 + 10] = T.float32(0)
-                    compute_2[cse_var_1 + 11] = T.float32(0)
-                    compute_2[cse_var_1 + 12] = T.float32(0)
-                    compute_2[cse_var_1 + 13] = T.float32(0)
-                    compute_2[cse_var_1 + 14] = T.float32(0)
-                    compute_2[cse_var_1 + 15] = T.float32(0)
-                for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i0_outer_i1_outer_fused * 2 + nb_j_inner}), 32):
-                    cse_var_2 = T.int32()
-                    placeholder_5 = T.Buffer((33,), &quot;int32&quot;, data=placeholder_3.data)
-                    cse_var_21: T.int32 = elem_idx * 16
-                    cse_var_20: T.int32 = i0_outer_i1_outer_fused * 2 + nb_j_inner
-                    cse_var_19: T.int32 = i_outer_inner * 8192 + i_inner * 256
-                    cse_var_18: T.int32 = i_outer_inner * 1024 + i_inner * 32 + nb_j_inner * 16
-                    cse_var_17: T.int32 = cse_var_18 + 9
-                    cse_var_16: T.int32 = cse_var_18 + 8
-                    cse_var_15: T.int32 = cse_var_18 + 7
-                    cse_var_14: T.int32 = cse_var_18 + 6
-                    cse_var_13: T.int32 = cse_var_18 + 5
-                    cse_var_12: T.int32 = cse_var_18 + 4
-                    cse_var_11: T.int32 = cse_var_18 + 3
-                    cse_var_10: T.int32 = cse_var_18 + 2
-                    cse_var_9: T.int32 = cse_var_18 + 15
-                    cse_var_8: T.int32 = cse_var_18 + 14
-                    cse_var_7: T.int32 = cse_var_18 + 13
-                    cse_var_6: T.int32 = cse_var_18 + 12
-                    cse_var_5: T.int32 = cse_var_18 + 11
-                    cse_var_4: T.int32 = cse_var_18 + 10
-                    cse_var_3: T.int32 = cse_var_18 + 1
-                    placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
-                    placeholder_7 = T.Buffer((32768,), data=placeholder.data)
-                    placeholder_8 = T.Buffer((4916,), &quot;int32&quot;, data=placeholder_2.data)
-                    compute_2[cse_var_18] = compute_2[cse_var_18] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-                    compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
-            for i0_inner in range(128):
-                cse_var_22: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 32
-                compute_3 = T.Buffer((65536,), data=compute.data)
-                placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
-                compute_3[cse_var_22:cse_var_22 + 32] = T.max(compute_2[i0_inner * 32:i0_inner * 32 + 32] + placeholder_5[cse_var_22:cse_var_22 + 32], T.Broadcast(T.float32(0), 32))
+        for i0_outer in T.parallel(8):
+            compute_1 = T.allocate([512], &quot;float32&quot;, &quot;global&quot;)
+            for i1_outer in range(16):
+                compute_2 = T.Buffer((512,), data=compute_1)
+                for nb_j_inner in range(2):
+                    for i_inner_init in range(16):
+                        cse_var_1: T.int32 = i_inner_init * 32 + nb_j_inner * 16
+                        compute_2[cse_var_1] = T.float32(0)
+                        compute_2[cse_var_1 + 1] = T.float32(0)
+                        compute_2[cse_var_1 + 2] = T.float32(0)
+                        compute_2[cse_var_1 + 3] = T.float32(0)
+                        compute_2[cse_var_1 + 4] = T.float32(0)
+                        compute_2[cse_var_1 + 5] = T.float32(0)
+                        compute_2[cse_var_1 + 6] = T.float32(0)
+                        compute_2[cse_var_1 + 7] = T.float32(0)
+                        compute_2[cse_var_1 + 8] = T.float32(0)
+                        compute_2[cse_var_1 + 9] = T.float32(0)
+                        compute_2[cse_var_1 + 10] = T.float32(0)
+                        compute_2[cse_var_1 + 11] = T.float32(0)
+                        compute_2[cse_var_1 + 12] = T.float32(0)
+                        compute_2[cse_var_1 + 13] = T.float32(0)
+                        compute_2[cse_var_1 + 14] = T.float32(0)
+                        compute_2[cse_var_1 + 15] = T.float32(0)
+                    for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i1_outer * 2 + nb_j_inner}), 16):
+                        cse_var_2 = T.int32()
+                        placeholder_5 = T.Buffer((33,), &quot;int32&quot;, data=placeholder_3.data)
+                        cse_var_21: T.int32 = elem_idx * 16
+                        cse_var_20: T.int32 = i1_outer * 2 + nb_j_inner
+                        cse_var_19: T.int32 = i0_outer * 4096 + i_inner * 256
+                        cse_var_18: T.int32 = i_inner * 32 + nb_j_inner * 16
+                        cse_var_17: T.int32 = cse_var_18 + 9
+                        cse_var_16: T.int32 = cse_var_18 + 8
+                        cse_var_15: T.int32 = cse_var_18 + 7
+                        cse_var_14: T.int32 = cse_var_18 + 6
+                        cse_var_13: T.int32 = cse_var_18 + 5
+                        cse_var_12: T.int32 = cse_var_18 + 4
+                        cse_var_11: T.int32 = cse_var_18 + 3
+                        cse_var_10: T.int32 = cse_var_18 + 2
+                        cse_var_9: T.int32 = cse_var_18 + 15
+                        cse_var_8: T.int32 = cse_var_18 + 14
+                        cse_var_7: T.int32 = cse_var_18 + 13
+                        cse_var_6: T.int32 = cse_var_18 + 12
+                        cse_var_5: T.int32 = cse_var_18 + 11
+                        cse_var_4: T.int32 = cse_var_18 + 10
+                        cse_var_3: T.int32 = cse_var_18 + 1
+                        placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
+                        placeholder_7 = T.Buffer((32768,), data=placeholder.data)
+                        placeholder_8 = T.Buffer((4916,), &quot;int32&quot;, data=placeholder_2.data)
+                        compute_2[cse_var_18] = compute_2[cse_var_18] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                        compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_19 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
+                for i0_inner in range(16):
+                    cse_var_22: T.int32 = i0_outer * 8192 + i0_inner * 512 + i1_outer * 32
+                    compute_3 = T.Buffer((65536,), data=compute.data)
+                    placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
+                    compute_3[cse_var_22:cse_var_22 + 32] = T.max(compute_2[i0_inner * 32:i0_inner * 32 + 32] + placeholder_5[cse_var_22:cse_var_22 + 32], T.Broadcast(T.float32(0), 32))
 </pre></div>
 </div>
 </div>
@@ -738,7 +739,7 @@ class Module:
 <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: 1.740 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.667 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 7f4d811cee..81ddb57dae 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>00:32.201</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:34.245</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -354,11 +354,11 @@
 </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>00:32.167</p></td>
+<td><p>00:34.212</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>00:00.021</p></td>
+<td><p>00:00.019</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>
@@ -366,7 +366,7 @@
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></td>
-<td><p>00:00.004</p></td>
+<td><p>00:00.005</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><td><p><a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></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 d4bc02d103..61ebc44f2c 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -573,7 +573,8 @@ for this template</p>
 waiting for device...
 device available
 Get devices for measurement successfully!
-No: 1   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+No: 1   GFLOPS: 67.10/67.10     result: MeasureResult(costs=(0.0034500863142857144,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.1896886825561523, timestamp=1678262153.3454974)      [(&#39;tile_f&#39;, [-1, 4, 8, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#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,1283341
+No: 2   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -695,8 +696,8 @@ 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 875, 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, 4, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 16, 4]), (&#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,7440299
-No: 2   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 1, 16]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 256]), (&#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;, 0)],None,4252327
+No: 3   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -818,8 +819,9 @@ 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 875, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 32, 1, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 256, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9937101
-No: 3   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 4, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 64, 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,9575209
+No: 4   GFLOPS: 30.59/67.10     result: MeasureResult(costs=(0.007568822571428571,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.9978384971618652, timestamp=1678262157.216696)        [(&#39;tile_f&#39;, [-1, 8, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 2]), (&#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;, 1)],None,8170023
+No: 5   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -941,8 +943,8 @@ 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 875, 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, 16, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 2, 256]), (&#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,3479826
-No: 4   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 2, 128]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 256, 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;, 0)],None,3128394
+No: 6   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1064,8 +1066,8 @@ 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 875, 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, 16, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 1, 128]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,8692675
-No: 5   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 2, 128]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 32]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4985634
+No: 7   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1187,8 +1189,8 @@ 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 875, 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, 128]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 64, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4706012
-No: 6   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 4, 16]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 32]), (&#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;, 0)],None,2080495
+No: 8   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1310,8 +1312,8 @@ 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 875, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 32, 2, 2]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 64]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5777709
-No: 7   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 1, 512]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 64, 1]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4864419
+No: 9   GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1433,8 +1435,9 @@ 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 875, 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, 16]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 32]), (&#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,724858
-No: 8   GFLOPS: 0.00/0.00       result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 1, 64]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 128]), (&#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,3660343
+No: 10  GFLOPS: 25.07/67.10     result: MeasureResult(costs=(0.009234245909090909,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.9820995330810547, timestamp=1678262163.0424588)       [(&#39;tile_f&#39;, [-1, 4, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 64, 2]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4703482
+No: 11  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1556,10 +1559,8 @@ 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 875, 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, 128]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 2, 128]), (&#39;tile_ry&#39;, [-1, 3, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,1534274
-No: 9   GFLOPS: 81.89/81.89     result: MeasureResult(costs=(0.0028270755277777777,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.0235955715179443, timestamp=1678247077.446814)       [(&#39;tile_f&#39;, [-1, 4, 2, 2]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 16, 1]), (&#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;, 0)],None,4080186
-No: 10  GFLOPS: 7.31/81.89      result: MeasureResult(costs=(0.03166355125,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.513221263885498, timestamp=1678247078.2887018)       [(&#39;tile_f&#39;, [-1, 1, 1, 4]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 2, 4]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4718000
-No: 11  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 2, 128]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 32, 16]), (&#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,6335773
+No: 12  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1681,8 +1682,8 @@ 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 875, 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, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 64]), (&#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,2678508
-No: 12  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 256, 1, 1]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 8, 32]), (&#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,732828
+No: 13  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1804,8 +1805,8 @@ 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 875, 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, 2]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 2, 128]), (&#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,2695065
-No: 13  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 32, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 1, 256]), (&#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,7348130
+No: 14  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -1927,8 +1928,8 @@ 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 875, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 32, 16]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#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,9570404
-No: 14  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 16, 4, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 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;, 1500), (&#39;unroll_explicit&#39;, 0)],None,4563799
+No: 15  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -2050,9 +2051,8 @@ 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 875, 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, 64, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 8, 1]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 3, 1]), (&#39;auto_unroll_max_step&#39;, 512), (&#39;unroll_explicit&#39;, 1)],None,7951594
-No: 15  GFLOPS: 2.34/81.89      result: MeasureResult(costs=(0.09909812150000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.931825876235962, timestamp=1678247082.4376202) [(&#39;tile_f&#39;, [-1, 16, 8, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 1, 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;, 1)],None,8714765
-No: 16  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 2, 32]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 512]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 1)],None,5804673
+No: 16  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -2174,8 +2174,9 @@ 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 875, 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, 1]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 16, 16]), (&#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,1878382
-No: 17  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 1, 4, 8]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 64, 2]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,446089
+No: 17  GFLOPS: 4.26/67.10      result: MeasureResult(costs=(0.05434708775,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.2146730422973633, timestamp=1678262166.633485)       [(&#39;tile_f&#39;, [-1, 32, 1, 4]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 64]), (&#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,7128545
+No: 18  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -2297,8 +2298,8 @@ 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 875, 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, 128, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 7]), (&#39;tile_rc&#39;, [-1, 4, 4]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,9950955
-No: 18  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 8, 1, 64]), (&#39;tile_y&#39;, [-1, 1, 1, 7]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 1, 512]), (&#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;, 0)],None,384543
+No: 19  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -2420,8 +2421,8 @@ 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 875, 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, 2, 16]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 256, 1]), (&#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,7193512
-No: 19  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 2, 4, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 7, 1, 1]), (&#39;tile_rc&#39;, [-1, 512, 1]), (&#39;tile_ry&#39;, [-1, 1, 3]), (&#39;tile_rx&#39;, [-1, 1, 1]), (&#39;auto_unroll_max_step&#39;, 0), (&#39;unroll_explicit&#39;, 0)],None,420053
+No: 20  GFLOPS: 0.00/67.10      result: Traceback (most recent call last):
   File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
@@ -2543,130 +2544,7 @@ 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 875, in verify_pass
     raise InstantiationError(&quot;Skipped because of invalid gpu kernel&quot;)
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel        [(&#39;tile_f&#39;, [-1, 32, 2, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 4, 64]), (&#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,7522353
-No: 20  GFLOPS: 0.00/81.89      result: Traceback (most recent call last):
-  File &quot;/workspace/python/tvm/autotvm/measure/measure_methods.py&quot;, line 592, 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 544, in _build_func_common
-    func = build(s, args, target=target, 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:1734
-  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:1674
-  19: run&lt;&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  18: run&lt;tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  17: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  16: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  15: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  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:1649
-  13: operator()
-        at ../src/driver/driver_api.cc:402
-  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:388
-  11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array&lt;tvm::transform::Pass, void&gt;)
-        at ../src/driver/driver_api.cc:283
-  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:451
-  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:1753
-  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:1697
-  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:1621
-  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 875, 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:1734
-  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:1674
-  19: run&lt;&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  18: run&lt;tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  17: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  16: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  15: run&lt;tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_&gt;
-        at ../include/tvm/runtime/packed_func.h:1634
-  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:1649
-  13: operator()
-        at ../src/driver/driver_api.cc:402
-  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:388
-  11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array&lt;tvm::transform::Pass, void&gt;)
-        at ../src/driver/driver_api.cc:283
-  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:451
-  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:1753
-  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:1697
-  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:1621
-  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 875, 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, 4]), (&#39;tile_y&#39;, [-1, 1, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#39;tile_rc&#39;, [-1, 64, 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,7027804
+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, 1, 64]), (&#39;tile_ry&#39;, [-1, 1, 1]), (&#39;tile_rx&#39;, [-1, 1, 3]), (&#39;auto_unroll_max_step&#39;, 1500), (&#39;unroll_explicit&#39;, 1)],None,10032685
 </pre></div>
 </div>
 <p>Finally we can inspect the best config from log file, check correctness,
@@ -2705,9 +2583,9 @@ 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, 4, 2, 2]), (&#39;tile_y&#39;, [-1, 1, 7, 1]), (&#39;tile_x&#39;, [-1, 1, 1, 1]), (&#39;tile_rc&#39;, [-1, 16, 1]), (&#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;, 0)],None,4080186
+[(&#39;tile_f&#39;, [-1, 4, 8, 2]), (&#39;tile_y&#39;, [-1, 7, 1, 1]), (&#39;tile_x&#39;, [-1, 1, 7, 1]), (&#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,1283341
 Finish loading 20 records
-Time cost of this operator: 0.001595
+Time cost of this operator: 0.003360
 </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 28343f4ed6..d8d5f1e02b 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -648,10 +648,10 @@ 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  317.2     98.752   (1, 2, 10, 10, 3)  2       1        [317.2]
-tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.055     0.951    (1, 6, 10, 10)     1       1        [3.055]
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.954     0.297    (1, 1, 10, 10, 3)  1       1        [0.954]
-Total_time                                    -                                             321.209   -        -                  -       -        -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  316.0     98.734   (1, 2, 10, 10, 3)  2       1        [316.0]
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       3.099     0.968    (1, 6, 10, 10)     1       1        [3.099]
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.954     0.298    (1, 1, 10, 10, 3)  1       1        [0.954]
+Total_time                                    -                                             320.053   -        -                  -       -        -
 </pre></div>
 </div>
 </div>
@@ -703,13 +703,13 @@ 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  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       1.796     1.702    (1, 6, 10, 10)     1       1        [1.796]
-tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.855     0.81     (1, 3, 10, 10, 1)  1       1        [0.855]
-Total_time                                    -                                             105.552   -        -                  -       -        -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc  tvmgen_default_fused_nn_contrib_conv2d_NCHWc  114.1     97.661   (1, 6, 10, 10, 1)  2       1        [114.1]
+tvmgen_default_fused_layout_transform_1       tvmgen_default_fused_layout_transform_1       1.769     1.514    (1, 6, 10, 10)     1       1        [1.769]
+tvmgen_default_fused_layout_transform         tvmgen_default_fused_layout_transform         0.965     0.826    (1, 1, 10, 10, 3)  1       1        [0.965]
+Total_time                                    -                                             116.833   -        -                  -       -        -
 </pre></div>
 </div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  21.564 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  20.055 seconds)</p>
 <div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-autotune-py">
 <div class="sphx-glr-download sphx-glr-download-python docutils container">
 <p><a class="reference download internal" download="" href="../../_downloads/9ccca8fd489a1486ac71b55a55c320c5/micro_autotune.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_autotune.py</span></code></a></p>
diff --git a/docs/how_to/work_with_microtvm/micro_pytorch.html b/docs/how_to/work_with_microtvm/micro_pytorch.html
index 6d061b144f..ff2d370b6f 100644
--- a/docs/how_to/work_with_microtvm/micro_pytorch.html
+++ b/docs/how_to/work_with_microtvm/micro_pytorch.html
@@ -459,7 +459,8 @@ 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]
-100%|##########| 3.42M/3.42M [00:00&lt;00:00, 114MB/s]
+ 61%|######    | 2.09M/3.42M [00:00&lt;00:00, 20.2MB/s]
+100%|##########| 3.42M/3.42M [00:00&lt;00:00, 31.3MB/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.
@@ -585,7 +586,7 @@ 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  20.710 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes  18.750 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 acf2c3b641..50690b04e5 100644
--- a/docs/how_to/work_with_microtvm/micro_train.html
+++ b/docs/how_to/work_with_microtvm/micro_train.html
@@ -528,7 +528,7 @@ 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/tmprd4ob6f2/images/random&#39;
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&#39;/tmp/tmpsk87vt3y/images/random&#39;
 </pre></div>
 </div>
 </div>
@@ -588,8 +588,8 @@ 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], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmprd4ob6f2/images/target contains 8144 images
-/tmp/tmprd4ob6f2/images/random contains 5000 images
+<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [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/tmpsk87vt3y/images/target contains 8144 images
+/tmp/tmpsk87vt3y/images/random contains 5000 images
 </pre></div>
 </div>
 </div>
@@ -701,13 +701,13 @@ 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: 0.2333 - accuracy: 0.9228 - val_loss: 0.1396 - val_accuracy: 0.9490 - 47s/epoch - 145ms/step
+328/328 - 47s - loss: 0.2111 - accuracy: 0.9258 - val_loss: 0.1170 - val_accuracy: 0.9603 - 47s/epoch - 143ms/step
 Epoch 2/3
-328/328 - 43s - loss: 0.1045 - accuracy: 0.9619 - val_loss: 0.0988 - val_accuracy: 0.9679 - 43s/epoch - 132ms/step
+328/328 - 43s - loss: 0.1019 - accuracy: 0.9631 - val_loss: 0.1391 - val_accuracy: 0.9547 - 43s/epoch - 132ms/step
 Epoch 3/3
-328/328 - 43s - loss: 0.0714 - accuracy: 0.9730 - val_loss: 0.0917 - val_accuracy: 0.9675 - 43s/epoch - 132ms/step
+328/328 - 43s - loss: 0.0651 - accuracy: 0.9761 - val_loss: 0.1452 - val_accuracy: 0.9528 - 43s/epoch - 132ms/step
 
-&lt;keras.callbacks.History object at 0x7f2c4652b910&gt;
+&lt;keras.callbacks.History object at 0x7f131c467750&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  30.774 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes  32.244 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 1c090a850d..fbbe995a47 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>07:30.654</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>07:28.310</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 82%" />
@@ -354,23 +354,23 @@
 </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">5. 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>04:30.774</p></td>
+<td><p>04:32.244</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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">6. Model Tuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></td>
-<td><p>01:21.564</p></td>
+<td><p>01:20.055</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-odd"><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">4. microTVM PyTorch Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_pytorch.py</span></code>)</p></td>
-<td><p>01:20.710</p></td>
+<td><p>01:18.750</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">3. microTVM Ahead-of-Time (AOT) Compilation</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_aot.py</span></code>)</p></td>
-<td><p>00:10.263</p></td>
+<td><p>00:10.165</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">2. microTVM TFLite Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></td>
-<td><p>00:07.343</p></td>
+<td><p>00:07.096</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><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">7. 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 ff049e5a48..430e74e151 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>00:46.404</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:45.343</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 84%" />
@@ -354,15 +354,15 @@
 </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>00:33.930</p></td>
+<td><p>00:33.413</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>00:10.589</p></td>
+<td><p>00:10.146</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>00:01.879</p></td>
+<td><p>00:01.779</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 7e24ae98d3..c4a361b2ce 100644
--- a/docs/how_to/work_with_schedules/intrin_math.html
+++ b/docs/how_to/work_with_schedules/intrin_math.html
@@ -540,7 +540,7 @@ 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 0x7f2aea840440&gt;
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>&lt;function my_cuda_math_rule at 0x7f11cb1e0d40&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 64120e4793..ea3b340de5 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -345,7 +345,7 @@
             
   <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>00:08.284</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:07.845</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
 <table class="docutils align-default">
 <colgroup>
 <col style="width: 83%" />
@@ -354,23 +354,23 @@
 </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>00:05.786</p></td>
+<td><p>00:05.463</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>00:01.135</p></td>
+<td><p>00:01.031</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>00:00.573</p></td>
+<td><p>00:00.572</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>00:00.556</p></td>
+<td><p>00:00.552</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>
-<td><p>00:00.121</p></td>
+<td><p>00:00.116</p></td>
 <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>
@@ -378,11 +378,11 @@
 <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>
-<td><p>00:00.033</p></td>
+<td><p>00:00.032</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 <tr class="row-even"><td><p><a class="reference internal" href="tuple_inputs.html#sphx-glr-how-to-work-with-schedules-tuple-inputs-py"><span class="std std-ref">Compute and Reduce with Tuple Inputs</span></a> (<code class="docutils literal notranslate"><span class="pre">tuple_inputs.py</span></code>)</p></td>
-<td><p>00:00.027</p></td>
+<td><p>00:00.026</p></td>
 <td><p>0.0 MB</p></td>
 </tr>
 </tbody>
diff --git a/docs/objects.inv b/docs/objects.inv
index c389acb56e..271b9fdeb2 100644
Binary files a/docs/objects.inv and b/docs/objects.inv differ
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
index 513a58385c..33f96ab6d8 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
@@ -129,6 +129,8 @@ $(function() {
   <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a0d492efee331e2239a093f4b2017c10f">ref_counter_</a></td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">protected</span></td></tr>
   <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a55549a6c23987890246248682560a03d">RefCounterType</a> typedef</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"></td></tr>
   <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5">ReIndex</a>(const BlockRV &amp;block_rv, int buffer_index, BufferIndexType buffer_index_type)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+  <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">ReindexCacheRead</a>(const BlockRV &amp;block_rv, int read_buffer_index, const String &amp;storage_scope, const IndexMap &amp;index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+  <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">ReindexCacheWrite</a>(const BlockRV &amp;block_rv, int write_buffer_index, const String &amp;storage_scope, const IndexMap &amp;index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
   <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a70d353bb52f6fa29fedeb90a6ff872d5">RemoveRV</a>(const BlockRV &amp;block_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
   <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a7c44d4f4ea662291ccb9d79383b6fefe">RemoveRV</a>(const LoopRV &amp;loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
   <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a00fcf343d2bc8f36f170c04e5e29d2dc">RemoveRV</a>(const ExprRV &amp;expr_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
index bcbe06b1af..530721f21e 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
@@ -207,6 +207,12 @@ Public Member Functions</h2></td></tr>
 <tr class="memitem:a9f43b2ddd6e3b3bab64671345ecba72d"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9f43b2ddd6e3b3bab64671345ecba72d">CacheWrite</a> (const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;block_rv, int write_buffer_index, const <a class="el" href="classtvm_1_1runtime [...]
 <tr class="memdesc:a9f43b2ddd6e3b3bab64671345ecba72d"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a block that writes a buffer region into a write cache. It requires: 1) There is only one block who writes the target buffer. 2) The scope block have stage-pipeline property.  <a href="#a9f43b2ddd6e3b3bab64671345ecba72d">More...</a><br /></td></tr>
 <tr class="separator:a9f43b2ddd6e3b3bab64671345ecba72d"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:a2625f87c74fe0cbc95006c763bb172b3"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">ReindexCacheRead</a> (const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;block_rv, int read_buffer_index, const <a class="el" href="classtvm_1_1ru [...]
+<tr class="memdesc:a2625f87c74fe0cbc95006c763bb172b3"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a block that reads a buffer region into a read cache. It requires: 1) There is at most one block who writes the buffer in the scope. 2) The scope block have stage-pipeline property. Compared to cache read, the indices to access allocated cache buffer is customized by user.  <a href="#a2625f87c74fe0cbc95006c763bb172b3">More...</a><br /></td></tr>
+<tr class="separator:a2625f87c74fe0cbc95006c763bb172b3"><td class="memSeparator" colspan="2">&#160;</td></tr>
+<tr class="memitem:af3983a7f5d094529049d83ce22d7b729"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a>&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">ReindexCacheWrite</a> (const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;block_rv, int write_buffer_index, const <a class="el" href="classtvm_1_1 [...]
+<tr class="memdesc:af3983a7f5d094529049d83ce22d7b729"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create a block that writes a buffer region into a write cache. It requires: 1) There is only one block who writes the target buffer. 2) The scope block have stage-pipeline property. Compared to cache write, the indices to access allocated cache buffer is customized by user.  <a href="#af3983a7f5d094529049d83ce22d7b729">More...</a><br /></td></tr>
+<tr class="separator:af3983a7f5d094529049d83ce22d7b729"><td class="memSeparator" colspan="2">&#160;</td></tr>
 <tr class="memitem:abefe28d441bad8fc4b276e1bb35f9818"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1runtime_1_1Array.html">Array</a>&lt; <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &gt;&#160;</td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abefe28d441bad8fc4b276e1bb35f9818">CacheInplace</a> (const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;blo [...]
 <tr class="memdesc:abefe28d441bad8fc4b276e1bb35f9818"><td class="mdescLeft">&#160;</td><td class="mdescRight">Create 2 blocks that read&amp;write a buffer region into a read/write cache. It requires the the target block both read &amp; write the target buffer.  <a href="#abefe28d441bad8fc4b276e1bb35f9818">More...</a><br /></td></tr>
 <tr class="separator:abefe28d441bad8fc4b276e1bb35f9818"><td class="memSeparator" colspan="2">&#160;</td></tr>
@@ -1878,6 +1884,126 @@ Additional Inherited Members</h2></td></tr>
 </dl>
 <dl class="section return"><dt>Returns</dt><dd>The reindex stage block. </dd></dl>
 
+</div>
+</div>
+<a id="a2625f87c74fe0cbc95006c763bb172b3"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a2625f87c74fe0cbc95006c763bb172b3">&#9670;&nbsp;</a></span>ReindexCacheRead()</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">virtual <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> tvm::tir::ScheduleNode::ReindexCacheRead </td>
+          <td>(</td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;&#160;</td>
+          <td class="paramname"><em>block_rv</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>read_buffer_index</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> &amp;&#160;</td>
+          <td class="paramname"><em>storage_scope</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1tir_1_1IndexMap.html">IndexMap</a> &amp;&#160;</td>
+          <td class="paramname"><em>index_map</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">pure virtual</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Create a block that reads a buffer region into a read cache. It requires: 1) There is at most one block who writes the buffer in the scope. 2) The scope block have stage-pipeline property. Compared to cache read, the indices to access allocated cache buffer is customized by user. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">block_rv</td><td>The consumer block of the target buffer. </td></tr>
+    <tr><td class="paramname">read_buffer_index</td><td>The index of the buffer in block's read region. </td></tr>
+    <tr><td class="paramname">storage_scope</td><td>The target storage scope. </td></tr>
+    <tr><td class="paramname">index_map</td><td>User defined indices to access allocated cache buffer, maps from block iter vars. </td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>The cache stage block. </dd></dl>
+
+</div>
+</div>
+<a id="af3983a7f5d094529049d83ce22d7b729"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#af3983a7f5d094529049d83ce22d7b729">&#9670;&nbsp;</a></span>ReindexCacheWrite()</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+  <tr>
+  <td class="mlabels-left">
+      <table class="memname">
+        <tr>
+          <td class="memname">virtual <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> tvm::tir::ScheduleNode::ReindexCacheWrite </td>
+          <td>(</td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &amp;&#160;</td>
+          <td class="paramname"><em>block_rv</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">int&#160;</td>
+          <td class="paramname"><em>write_buffer_index</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> &amp;&#160;</td>
+          <td class="paramname"><em>storage_scope</em>, </td>
+        </tr>
+        <tr>
+          <td class="paramkey"></td>
+          <td></td>
+          <td class="paramtype">const <a class="el" href="classtvm_1_1tir_1_1IndexMap.html">IndexMap</a> &amp;&#160;</td>
+          <td class="paramname"><em>index_map</em>&#160;</td>
+        </tr>
+        <tr>
+          <td></td>
+          <td>)</td>
+          <td></td><td></td>
+        </tr>
+      </table>
+  </td>
+  <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">pure virtual</span></span>  </td>
+  </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Create a block that writes a buffer region into a write cache. It requires: 1) There is only one block who writes the target buffer. 2) The scope block have stage-pipeline property. Compared to cache write, the indices to access allocated cache buffer is customized by user. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+  <table class="params">
+    <tr><td class="paramname">block_rv</td><td>The producer of the buffer </td></tr>
+    <tr><td class="paramname">write_buffer_index</td><td>The index of the buffer in block's write region </td></tr>
+    <tr><td class="paramname">storage_scope</td><td>The target storage scope </td></tr>
+    <tr><td class="paramname">index_map</td><td>User defined indices to access allocated cache buffer, maps from block iter vars. </td></tr>
+  </table>
+  </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>The cache stage block. </dd></dl>
+
 </div>
 </div>
 <a id="a70d353bb52f6fa29fedeb90a6ff872d5"></a>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
index 842d8fb94b..d28a3b3908 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
@@ -27,7 +27,7 @@
 <text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Seed()</text>
 <text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ForkSeed()</text>
 <text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Get()</text>
-<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 55 more...</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 57 more...</text>
 </g>
 <!-- Node3 -->
 <g id="node2" class="node">
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
index 3dd6487331..e0b39ed8f9 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
@@ -27,7 +27,7 @@
 <text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Seed()</text>
 <text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ForkSeed()</text>
 <text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Get()</text>
-<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 55 more...</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 57 more...</text>
 </g>
 <!-- Node1 -->
 <g id="node2" class="node">
diff --git a/docs/reference/api/doxygen/database_8h_source.html b/docs/reference/api/doxygen/database_8h_source.html
index 96025bcfea..0590ed81c4 100644
--- a/docs/reference/api/doxygen/database_8h_source.html
+++ b/docs/reference/api/doxygen/database_8h_source.html
@@ -85,7 +85,7 @@ $(function() {
 <div class="ttc" id="structtvm_1_1meta__schedule_1_1WorkloadHash_html_a7cb09ddc6c76d9d00ddbeab8502d97cb"><div class="ttname"><a href="structtvm_1_1meta__schedule_1_1WorkloadHash.html#a7cb09ddc6c76d9d00ddbeab8502d97cb">tvm::meta_schedule::WorkloadHash::operator()</a></div><div class="ttdeci">size_t operator()(const Workload &amp;a) const</div><div class="ttdef"><b>Definition:</b> database.h:96</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1PyDatabaseNode_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyDatabaseNode.html">tvm::meta_schedule::PyDatabaseNode</a></div><div class="ttdoc">The database with customized methods on the python-side. </div><div class="ttdef"><b>Definition:</b> database.h:277</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="structtvm_1_1meta__schedule_1_1WorkloadEqual_html_a49b0e137a01a278469ff63729afae804"><div class="ttname"><a href="structtvm_1_1meta__schedule_1_1WorkloadEqual.html#a49b0e137a01a278469ff63729afae804">tvm::meta_schedule::WorkloadEqual::WorkloadEqual</a></div><div class="ttdeci">WorkloadEqual(const ModuleEquality &amp;mod_eq)</div><div class="ttdef"><b>Definition:</b> database.h:101</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1TuningRecordNode_html_a8cc2d64f796593a1a774eef259f17b29"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a8cc2d64f796593a1a774eef259f17b29">tvm::meta_schedule::TuningRecordNode::trace</a></div><div class="ttdeci">tir::Trace trace</div><div class="ttdoc">The trace tuned. </div><div class="ttdef"><b>Definition:</b> database.h:117</div></div>
 <div class="ttc" id="arg__info_8h_html"><div class="ttname"><a href="arg__info_8h.html">arg_info.h</a></div></div>
diff --git a/docs/reference/api/doxygen/functions_func_r.html b/docs/reference/api/doxygen/functions_func_r.html
index f960446c49..cfb6eb96d2 100644
--- a/docs/reference/api/doxygen/functions_func_r.html
+++ b/docs/reference/api/doxygen/functions_func_r.html
@@ -169,6 +169,12 @@ $(function() {
 <li>ReIndex()
 : <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5">tvm::tir::ScheduleNode</a>
 </li>
+<li>ReindexCacheRead()
+: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">tvm::tir::ScheduleNode</a>
+</li>
+<li>ReindexCacheWrite()
+: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">tvm::tir::ScheduleNode</a>
+</li>
 <li>Reinterpret()
 : <a class="el" href="classtvm_1_1BaseValueHash.html#a5ca1332d212969bb1bf6c865d5703b39">tvm::BaseValueHash</a>
 </li>
@@ -310,7 +316,7 @@ $(function() {
 </li>
 <li>Rewrite_()
 : <a class="el" href="classtvm_1_1relay_1_1ExprRewriter.html#a867da2f78446c33e201da79e9e1a0b2e">tvm::relay::ExprRewriter</a>
-, <a class="el" href="classtvm_1_1relay_1_1MixedModeMutator.html#a3b53908f4b8cc3708ca75892e47f0929">tvm::relay::MixedModeMutator</a>
+, <a class="el" href="classtvm_1_1relay_1_1MixedModeMutator.html#a2424d6590fceb087cb1624ab8d3348a1">tvm::relay::MixedModeMutator</a>
 </li>
 <li>RewriteCooperativeFetch()
 : <a class="el" href="classtvm_1_1meta__schedule_1_1Postproc.html#a8a8e8e047dcdcf89ad9d96eed47c293a">tvm::meta_schedule::Postproc</a>
diff --git a/docs/reference/api/doxygen/functions_func_u.html b/docs/reference/api/doxygen/functions_func_u.html
index 611cae9ff1..4b4e0f203d 100644
--- a/docs/reference/api/doxygen/functions_func_u.html
+++ b/docs/reference/api/doxygen/functions_func_u.html
@@ -106,7 +106,7 @@ $(function() {
 , <a class="el" href="classtvm_1_1auto__scheduler_1_1CostModelNode.html#ae35b2b678760b8da57a43d3ae9c24da5">tvm::auto_scheduler::CostModelNode</a>
 , <a class="el" href="classtvm_1_1auto__scheduler_1_1PythonBasedModelNode.html#a2d7849df6c7dbe93bf363c1d9f860a26">tvm::auto_scheduler::PythonBasedModelNode</a>
 , <a class="el" href="classtvm_1_1auto__scheduler_1_1RandomModelNode.html#a7febac6c05d8e2d407f466467769ee32">tvm::auto_scheduler::RandomModelNode</a>
-, <a class="el" href="classtvm_1_1IRModuleNode.html#abdd8936c6fca33ef9b7c086f8fd58f84">tvm::IRModuleNode</a>
+, <a class="el" href="classtvm_1_1IRModuleNode.html#a94a93385e64ce844299729af6a573015">tvm::IRModuleNode</a>
 , <a class="el" href="classtvm_1_1meta__schedule_1_1CostModelNode.html#a1bba32eba84db583fe90d1a5bce085f1">tvm::meta_schedule::CostModelNode</a>
 , <a class="el" href="classtvm_1_1meta__schedule_1_1PyCostModelNode.html#a970b00b0eb1bf6b88eea2711b58c4d1d">tvm::meta_schedule::PyCostModelNode</a>
 </li>
diff --git a/docs/reference/api/doxygen/functions_r.html b/docs/reference/api/doxygen/functions_r.html
index 37c492b872..d9fad78c2f 100644
--- a/docs/reference/api/doxygen/functions_r.html
+++ b/docs/reference/api/doxygen/functions_r.html
@@ -281,6 +281,12 @@ $(function() {
 <li>ReIndex()
 : <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5">tvm::tir::ScheduleNode</a>
 </li>
+<li>ReindexCacheRead()
+: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">tvm::tir::ScheduleNode</a>
+</li>
+<li>ReindexCacheWrite()
+: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">tvm::tir::ScheduleNode</a>
+</li>
 <li>Reinterpret()
 : <a class="el" href="classtvm_1_1BaseValueHash.html#a5ca1332d212969bb1bf6c865d5703b39">tvm::BaseValueHash</a>
 </li>
@@ -498,7 +504,7 @@ $(function() {
 , <a class="el" href="classtvm_1_1tir_1_1IndexDataTypeNormalizer.html#afc2a6d4cb922111b859d75e3ba43c150">tvm::tir::IndexDataTypeNormalizer</a>
 </li>
 <li>Rewrite_()
-: <a class="el" href="classtvm_1_1relay_1_1ExprRewriter.html#a956ed40ff5f64ecef5e6034056184cd7">tvm::relay::ExprRewriter</a>
+: <a class="el" href="classtvm_1_1relay_1_1ExprRewriter.html#a0ce4f1f1a3abf18ee99addd3de09e99e">tvm::relay::ExprRewriter</a>
 , <a class="el" href="classtvm_1_1relay_1_1MixedModeMutator.html#a2424d6590fceb087cb1624ab8d3348a1">tvm::relay::MixedModeMutator</a>
 </li>
 <li>rewrite_once
diff --git a/docs/reference/api/doxygen/functions_s.html b/docs/reference/api/doxygen/functions_s.html
index 46ad572f77..018d2c76e2 100644
--- a/docs/reference/api/doxygen/functions_s.html
+++ b/docs/reference/api/doxygen/functions_s.html
@@ -949,7 +949,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1runtime_1_1String.html#a88aa19338db83ae838183e7478f09e60">tvm::runtime::String</a>
 </li>
 <li>Stage()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#af0643fe8c1298451c9a322f915c48843">tvm::auto_scheduler::Stage</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#a39ffbb1b4e189180bc4067e74965f42b">tvm::auto_scheduler::Stage</a>
 , <a class="el" href="classtvm_1_1te_1_1Stage.html#a1ecdc9a000be62c9cc26a96d4c33e36e">tvm::te::Stage</a>
 </li>
 <li>stage_id
diff --git a/docs/reference/api/doxygen/functions_t.html b/docs/reference/api/doxygen/functions_t.html
index da4324a2ea..2e0ece61cc 100644
--- a/docs/reference/api/doxygen/functions_t.html
+++ b/docs/reference/api/doxygen/functions_t.html
@@ -202,7 +202,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1te_1_1Stage.html#ab5fe485e1d730c36b096c060b8d2ef9d">tvm::te::Stage</a>
 </li>
 <li>Tensorize()
-: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928">tvm::tir::ScheduleNode</a>
+: <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf">tvm::tir::ScheduleNode</a>
 </li>
 <li>TensorNode()
 : <a class="el" href="classtvm_1_1te_1_1TensorNode.html#a153569448cb1bf9d2924d35639c3b8b8">tvm::te::TensorNode</a>
@@ -1538,7 +1538,7 @@ $(function() {
 : <a class="el" href="classtvm_1_1TypeRelation.html#ac26b1897eab8197ed26606ab81b7403b">tvm::TypeRelation</a>
 </li>
 <li>TypeReporter()
-: <a class="el" href="classtvm_1_1TypeReporter.html#a8e7e05a07f9f7ad9bea91f27afac9051">tvm::TypeReporter</a>
+: <a class="el" href="classtvm_1_1TypeReporter.html#aa3dc38a3c84d324d0b3a9f358460a091">tvm::TypeReporter</a>
 </li>
 <li>types
 : <a class="el" href="classtvm_1_1TupleAffineTypeNode.html#a30c834b7e1cb64467e6587ac16ebb187">tvm::TupleAffineTypeNode</a>
diff --git a/docs/reference/api/doxygen/functions_v.html b/docs/reference/api/doxygen/functions_v.html
index 9d88f5d3f0..bd23544c17 100644
--- a/docs/reference/api/doxygen/functions_v.html
+++ b/docs/reference/api/doxygen/functions_v.html
@@ -664,7 +664,7 @@ $(function() {
 </li>
 <li>VisitType_()
 : <a class="el" href="classtvm_1_1TypeFunctor_3_01R_07const_01Type_01_6n_00_01Args_8_8_8_08_4.html#a0e715c54558934e4504c366ff803d8e1">tvm::TypeFunctor&lt; R(const Type &amp;n, Args...)&gt;</a>
-, <a class="el" href="classtvm_1_1TypeMutator.html#a4c7667d35d0a9a28c957165b65536c93">tvm::TypeMutator</a>
+, <a class="el" href="classtvm_1_1TypeMutator.html#ab270cb7a0d9af50362137c1d68d32276">tvm::TypeMutator</a>
 , <a class="el" href="classtvm_1_1TypeVisitor.html#ae699be9a6ed94a635c315506e0c2a6d2">tvm::TypeVisitor</a>
 </li>
 <li>VisitTypeDefault_()
diff --git a/docs/reference/api/doxygen/measure__candidate_8h_source.html b/docs/reference/api/doxygen/measure__candidate_8h_source.html
index 5cc2da1bc2..88002f2bb5 100644
--- a/docs/reference/api/doxygen/measure__candidate_8h_source.html
+++ b/docs/reference/api/doxygen/measure__candidate_8h_source.html
@@ -71,7 +71,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html_a99858dbe74082cc52938ac942523d792"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html#a99858dbe74082cc52938ac942523d792">tvm::meta_schedule::MeasureCandidateNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(tvm::AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> measure_candidate.h:40</div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html_a6891e92cac8712bb690401ed121ae7e8"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html#a6891e92cac8712bb690401ed121ae7e8">tvm::meta_schedule::MeasureCandidateNode::args_info</a></div><div class="ttdeci">Array&lt; ArgInfo &gt; args_info</div><div class="ttdoc">The argument information, e.g., (shape, dtype) for tensors. </div><div class="ttdef"><b>Definition:</b> measure_candidate. [...]
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="arg__info_8h_html"><div class="ttname"><a href="arg__info_8h.html">arg_info.h</a></div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html">tvm::meta_schedule::MeasureCandidateNode</a></div><div class="ttdoc">The schedule (with input shapes) to be measured. </div><div class="ttdef"><b>Definition:</b> measure_candidate.h:33</div></div>
 <div class="ttc" id="array_8h_html"><div class="ttname"><a href="array_8h.html">array.h</a></div><div class="ttdoc">Runtime Array container types. </div></div>
diff --git a/docs/reference/api/doxygen/postproc_8h_source.html b/docs/reference/api/doxygen/postproc_8h_source.html
index 7500f04c00..ed97838b67 100644
--- a/docs/reference/api/doxygen/postproc_8h_source.html
+++ b/docs/reference/api/doxygen/postproc_8h_source.html
@@ -75,7 +75,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1PyPostprocNode_html_a3771e585727ef6dfecc502ffe57fd2a2"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyPostprocNode.html#a3771e585727ef6dfecc502ffe57fd2a2">tvm::meta_schedule::PyPostprocNode::f_apply</a></div><div class="ttdeci">FApply f_apply</div><div class="ttdoc">The packed function to the Apply function. </div><div class="ttdef"><b>Definition:</b> postproc.h:190</div></div>
 <div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1TuneContext_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuneContext.html">tvm::meta_schedule::TuneContext</a></div><div class="ttdoc">Managed reference to TuneContextNode. </div><div class="ttdef"><b>Definition:</b> tune_context.h:95</div></div>
 <div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1Postproc_html_a1b95aac48704d0c0740ede2040b942bb"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1Postproc.html#a1b95aac48704d0c0740ede2040b942bb">tvm::meta_schedule::Postproc::FAsString</a></div><div class="ttdeci">runtime::TypedPackedFunc&lt; String()&gt; FAsString</div><div class="ttdoc">Get the postprocessor function as string with name. </div><div class="ttdef"><b>Definition:</b> postproc.h:94</div></div>
diff --git a/docs/reference/api/doxygen/schedule__rule_8h_source.html b/docs/reference/api/doxygen/schedule__rule_8h_source.html
index 09a9725ea7..2c304d485a 100644
--- a/docs/reference/api/doxygen/schedule__rule_8h_source.html
+++ b/docs/reference/api/doxygen/schedule__rule_8h_source.html
@@ -80,7 +80,7 @@ $(function() {
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
 <div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1PyScheduleRuleNode_html_a752192bcb5385b1ba72b7c1856c6f360"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyScheduleRuleNode.html#a752192bcb5385b1ba72b7c1856c6f360">tvm::meta_schedule::PyScheduleRuleNode::f_apply</a></div><div class="ttdeci">FApply f_apply</div><div class="ttdoc">The packed function to the Apply function. </div><div class="ttdef"><b>Definition:</b> schedule_rule.h:318</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="classtvm_1_1meta__schedule_1_1TuneContext_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuneContext.html">tvm::meta_schedule::TuneContext</a></div><div class="ttdoc">Managed reference to TuneContextNode. </div><div class="ttdef"><b>Definition:</b> tune_context.h:95</div></div>
 <div class="ttc" id="array_8h_html"><div class="ttname"><a href="array_8h.html">array.h</a></div><div class="ttdoc">Runtime Array container types. </div></div>
 <div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
diff --git a/docs/reference/api/doxygen/search/all_11.js b/docs/reference/api/doxygen/search/all_11.js
index 4ec5ff350e..b454b90c2b 100644
--- a/docs/reference/api/doxygen/search/all_11.js
+++ b/docs/reference/api/doxygen/search/all_11.js
@@ -42,7 +42,7 @@ var searchData=
   ['parallelizevectorizeunroll',['ParallelizeVectorizeUnroll',['../classtvm_1_1meta__schedule_1_1ScheduleRule.html#a0ef9b604081db7a8bf960f3fbfd3a804',1,'tvm::meta_schedule::ScheduleRule']]],
   ['param_5fdevice_5findexes',['param_device_indexes',['../structtvm_1_1runtime_1_1vm_1_1VMFunction.html#afff8cae6bf6100376c4275b301a11828',1,'tvm::runtime::vm::VMFunction']]],
   ['params',['params',['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#a6530833b23371eaeee737cc891b160b9',1,'tvm::meta_schedule::BuilderInputNode::params()'],['../classtvm_1_1relay_1_1FunctionPatternNode.html#a0773abf2bbd77c732a3b098d68dcd92d',1,'tvm::relay::FunctionPatternNode::params()'],['../classtvm_1_1relay_1_1FunctionNode.html#a689c5078fb501798aa1d57d1825e3b5a',1,'tvm::relay::FunctionNode::params()'],['../structtvm_1_1runtime_1_1vm_1_1VMFunction.html#a6d480483112097888c7150 [...]
-  ['parent',['Parent',['../classtvm_1_1tir_1_1IndexDataTypeRewriter.html#a0e7deada1e51a604e4f7bc6c8a115955',1,'tvm::tir::IndexDataTypeRewriter::Parent()'],['../classtvm_1_1tir_1_1IndexDataTypeNormalizer.html#a8d04a64a80c9fbdd28c7cb9e00c02120',1,'tvm::tir::IndexDataTypeNormalizer::Parent()'],['../classtvm_1_1relay_1_1DominatorPatternNode.html#ab4720ce40ecd4f89c7eb6ff469c40375',1,'tvm::relay::DominatorPatternNode::parent()'],['../classtvm_1_1te_1_1SplitNode.html#aff313d36c09c8d93ba34aa4a25 [...]
+  ['parent',['parent',['../classtvm_1_1relay_1_1DominatorPatternNode.html#ab4720ce40ecd4f89c7eb6ff469c40375',1,'tvm::relay::DominatorPatternNode::parent()'],['../classtvm_1_1te_1_1SplitNode.html#aff313d36c09c8d93ba34aa4a251d6a9f',1,'tvm::te::SplitNode::parent()'],['../classtvm_1_1te_1_1RebaseNode.html#a4fea0a62970e2831dd92c58aee14a1f4',1,'tvm::te::RebaseNode::parent()'],['../classtvm_1_1tir_1_1StmtSRefNode.html#aa69a77c8bee12ecc0c17fe75656e392d',1,'tvm::tir::StmtSRefNode::parent()'],['.. [...]
   ['parentnode',['ParentNode',['../classtvm_1_1ObjectPathNode.html#a4a06a6588b427640e2f775a6307ccab5',1,'tvm::ObjectPathNode']]],
   ['parse_5fauto_5fscheduler_5flayout',['parse_auto_scheduler_layout',['../namespacetvm_1_1topi.html#a6fec3a88d419cd018cf1041cf3d99204',1,'tvm::topi']]],
   ['parsecustomdatatype',['ParseCustomDatatype',['../namespacetvm_1_1runtime.html#ac6d167947285dff3c780bd5d01ace9b9',1,'tvm::runtime']]],
diff --git a/docs/reference/api/doxygen/search/all_13.js b/docs/reference/api/doxygen/search/all_13.js
index 4c6f2e5a34..70193e3cbc 100644
--- a/docs/reference/api/doxygen/search/all_13.js
+++ b/docs/reference/api/doxygen/search/all_13.js
@@ -88,6 +88,8 @@ var searchData=
   ['registry_2eh',['registry.h',['../registry_8h.html',1,'']]],
   ['regname',['RegName',['../namespacetvm_1_1runtime_1_1vm.html#a3bbbf700719e9dc3dda2bc25210c18ae',1,'tvm::runtime::vm']]],
   ['reindex',['ReIndex',['../classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5',1,'tvm::tir::ScheduleNode']]],
+  ['reindexcacheread',['ReindexCacheRead',['../classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3',1,'tvm::tir::ScheduleNode']]],
+  ['reindexcachewrite',['ReindexCacheWrite',['../classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729',1,'tvm::tir::ScheduleNode']]],
   ['reinterpret',['Reinterpret',['../classtvm_1_1BaseValueHash.html#a5ca1332d212969bb1bf6c865d5703b39',1,'tvm::BaseValueHash::Reinterpret()'],['../namespacetvm_1_1tir_1_1builtin.html#a7b555bc5cca2f5e7b26c1037bc0001ce',1,'tvm::tir::builtin::reinterpret()'],['../namespacetvm.html#a34084606675cd2c73c6b0f10e1618280',1,'tvm::reinterpret()'],['../namespacetvm_1_1topi.html#a25239505894bdae140e53f4abc146f92',1,'tvm::topi::reinterpret()']]],
   ['relations',['relations',['../classtvm_1_1arith_1_1IntConstraintsNode.html#a078c29fba655311710227460312e78b5',1,'tvm::arith::IntConstraintsNode::relations()'],['../classtvm_1_1te_1_1StageNode.html#ad1c0f8dc1f0f406a2abcd05fdad8fad5',1,'tvm::te::StageNode::relations()']]],
   ['relay_5fdebug',['RELAY_DEBUG',['../relay_2base_8h.html#aa85debc943daeee5d755b1eab7c9d108',1,'base.h']]],
@@ -151,7 +153,7 @@ var searchData=
   ['reserve',['reserve',['../classtvm_1_1runtime_1_1Array.html#a1a7727b86efaf35c58a5198ab1c139c8',1,'tvm::runtime::Array']]],
   ['reserveglobalvar',['ReserveGlobalVar',['../classtvm_1_1GlobalVarSupplyNode.html#a29185b94238fc62c928346a004c43b16',1,'tvm::GlobalVarSupplyNode']]],
   ['reservename',['ReserveName',['../classtvm_1_1NameSupplyNode.html#a9feb960ebeeee03fb9c5105655a8da17',1,'tvm::NameSupplyNode']]],
-  ['reset',['Reset',['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#a73b14ea360a9902c291d5bf6e97636cd',1,'tvm::auto_scheduler::ProgramMeasurerNode::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Unframer.html#ae6279154fe70e9eb85937b51e70a4bf8',1,'tvm::runtime::micro_rpc::Unframer::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Framer.html#a44ff9650ecca8785e33c25c369d2570a',1,'tvm::runtime::micro_rpc::Framer::Reset()'],['../classtvm_1_1tir_1_1StmtSRefNode.html#a0a81 [...]
+  ['reset',['reset',['../classtvm_1_1runtime_1_1NDArray.html#af2a8ccab95d432d1ecad7a389e11bcd3',1,'tvm::runtime::NDArray::reset()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#ac4461465ba0e785794794e0405c96590',1,'tvm::runtime::ObjectPtr::reset()'],['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#a73b14ea360a9902c291d5bf6e97636cd',1,'tvm::auto_scheduler::ProgramMeasurerNode::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Unframer.html#ae6279154fe70e9eb85937b51e70a4bf8',1, [...]
   ['reset_5fattr',['reset_attr',['../classtvm_1_1OpRegEntry.html#a67628f8d3d6dea5b0a47e462c06b7790',1,'tvm::OpRegEntry']]],
   ['resetthreadpool',['ResetThreadPool',['../namespacetvm_1_1runtime_1_1threading.html#aafdb21c00248ff146b614a7e888b4fd7',1,'tvm::runtime::threading']]],
   ['reshape',['reshape',['../namespacetvm_1_1topi.html#a3aad65f2505802109ba7d05359ce9005',1,'tvm::topi']]],
diff --git a/docs/reference/api/doxygen/search/all_14.js b/docs/reference/api/doxygen/search/all_14.js
index bfa646967c..b6e3402203 100644
--- a/docs/reference/api/doxygen/search/all_14.js
+++ b/docs/reference/api/doxygen/search/all_14.js
@@ -99,7 +99,7 @@ var searchData=
   ['selectshashreduce_3c_20t_2c_20traitname_2c_20false_20_3e',['SelectSHashReduce&lt; T, TraitName, false &gt;',['../structtvm_1_1detail_1_1SelectSHashReduce_3_01T_00_01TraitName_00_01false_01_4.html',1,'tvm::detail']]],
   ['selectvisitattrs',['SelectVisitAttrs',['../structtvm_1_1detail_1_1SelectVisitAttrs.html',1,'tvm::detail']]],
   ['selectvisitattrs_3c_20t_2c_20traitname_2c_20false_20_3e',['SelectVisitAttrs&lt; T, TraitName, false &gt;',['../structtvm_1_1detail_1_1SelectVisitAttrs_3_01T_00_01TraitName_00_01false_01_4.html',1,'tvm::detail']]],
-  ['self',['Self',['../classtvm_1_1runtime_1_1InplaceArrayBase.html#ae447f7c7a742fb3f5613a632706509df',1,'tvm::runtime::InplaceArrayBase::Self()'],['../classtvm_1_1runtime_1_1MapNode_1_1iterator.html#a5bac4439279428fb3c0d44aa6b1cc798',1,'tvm::runtime::MapNode::iterator::self()']]],
+  ['self',['self',['../classtvm_1_1runtime_1_1MapNode_1_1iterator.html#a5bac4439279428fb3c0d44aa6b1cc798',1,'tvm::runtime::MapNode::iterator::self()'],['../classtvm_1_1runtime_1_1InplaceArrayBase.html#ae447f7c7a742fb3f5613a632706509df',1,'tvm::runtime::InplaceArrayBase::Self()']]],
   ['sendbodychunk',['SendBodyChunk',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a37b77101825145283cced6cd05eb502c',1,'tvm::runtime::micro_rpc::Session']]],
   ['sendmessage',['SendMessage',['../classtvm_1_1runtime_1_1micro__rpc_1_1Session.html#a6e540521a7e9188564da712c0641619c',1,'tvm::runtime::micro_rpc::Session']]],
   ['seq',['seq',['../classtvm_1_1tir_1_1SeqStmtNode.html#a0e548955529d35c56e646fcaac38f865',1,'tvm::tir::SeqStmtNode']]],
@@ -260,7 +260,7 @@ var searchData=
   ['solvelinearequations',['SolveLinearEquations',['../namespacetvm_1_1arith.html#ae0290f04432523ab8e5f76edde80071a',1,'tvm::arith']]],
   ['solvelinearinequalities',['SolveLinearInequalities',['../namespacetvm_1_1arith.html#ac59d63560e04431f108e81457b212fdc',1,'tvm::arith']]],
   ['sorted',['sorted',['../structtvm_1_1relay_1_1UniqueAttrs.html#aef434799646533ec9d796393ba01db44',1,'tvm::relay::UniqueAttrs']]],
-  ['source',['Source',['../classtvm_1_1Source.html',1,'tvm::Source'],['../classtvm_1_1Source.html#a794f5989e4e89cb39b638a5cd837eaa9',1,'tvm::Source::Source()'],['../classtvm_1_1arith_1_1IterMarkNode.html#a8b885a675c88e5a5d142fa68bcba048a',1,'tvm::arith::IterMarkNode::source()'],['../classtvm_1_1arith_1_1IterSplitExprNode.html#a7a129dc9b432359a07c1a1e286c3c66f',1,'tvm::arith::IterSplitExprNode::source()'],['../classtvm_1_1SourceNode.html#a6a31f8a2bf412edf7a529c5931185529',1,'tvm::SourceNo [...]
+  ['source',['Source',['../classtvm_1_1Source.html',1,'tvm::Source'],['../classtvm_1_1arith_1_1IterMarkNode.html#a8b885a675c88e5a5d142fa68bcba048a',1,'tvm::arith::IterMarkNode::source()'],['../classtvm_1_1arith_1_1IterSplitExprNode.html#a7a129dc9b432359a07c1a1e286c3c66f',1,'tvm::arith::IterSplitExprNode::source()'],['../classtvm_1_1SourceNode.html#a6a31f8a2bf412edf7a529c5931185529',1,'tvm::SourceNode::source()'],['../structtvm_1_1tir_1_1MemCpyDetails.html#adf01e30a65e219e5303f535f793dbaf [...]
   ['source_5fmap',['source_map',['../classtvm_1_1IRModuleNode.html#abc7bd8d734f093ecb6abe8969d6dbcba',1,'tvm::IRModuleNode::source_map()'],['../classtvm_1_1SourceMapNode.html#abd44bcd7252397ade5584c0387845342',1,'tvm::SourceMapNode::source_map()']]],
   ['source_5fmap_2eh',['source_map.h',['../source__map_8h.html',1,'']]],
   ['source_5fname',['source_name',['../classtvm_1_1DiagnosticBuilder.html#a92d320e1ede24fe5ff47862365002691',1,'tvm::DiagnosticBuilder::source_name()'],['../classtvm_1_1SpanNode.html#ad573167f93facbfbee19983b08bbba3d',1,'tvm::SpanNode::source_name()'],['../classtvm_1_1SourceNode.html#ad4002d5ddb822b7340fd16c460a1ec6f',1,'tvm::SourceNode::source_name()']]],
@@ -277,7 +277,7 @@ var searchData=
   ['spacegeneratornode',['SpaceGeneratorNode',['../classtvm_1_1meta__schedule_1_1SpaceGeneratorNode.html',1,'tvm::meta_schedule']]],
   ['spacegeneratorunion',['SpaceGeneratorUnion',['../classtvm_1_1meta__schedule_1_1SpaceGenerator.html#a44828204c6ae3b7f390b9a9c3fdb9aa7',1,'tvm::meta_schedule::SpaceGenerator']]],
   ['spacetobatchndattrs',['SpaceToBatchNDAttrs',['../structtvm_1_1relay_1_1SpaceToBatchNDAttrs.html',1,'tvm::relay']]],
-  ['span',['Span',['../classtvm_1_1support_1_1Span.html',1,'tvm::support::Span&lt; T, W &gt;'],['../classtvm_1_1Span.html',1,'tvm::Span'],['../classtvm_1_1Span.html#a5216631b639e8c802263d87d3fe9e5f6',1,'tvm::Span::Span()'],['../classtvm_1_1support_1_1Span.html#a77653730a2542edf93b7c4413a72f3ec',1,'tvm::support::Span::Span(T *begin, int num_elements)'],['../classtvm_1_1support_1_1Span.html#a3c22dd06856e7029e7107adf38eb72f5',1,'tvm::support::Span::Span(T *begin, T *end)'],['../classtvm_1_1 [...]
+  ['span',['Span',['../classtvm_1_1support_1_1Span.html',1,'tvm::support::Span&lt; T, W &gt;'],['../classtvm_1_1Span.html',1,'tvm::Span'],['../classtvm_1_1AffineTypeNode.html#aa45c91e3c8ebcff609d10f6a921f3fa2',1,'tvm::AffineTypeNode::span()'],['../classtvm_1_1DiagnosticNode.html#af5469f228f87711ad8bd3f4f78f3bb54',1,'tvm::DiagnosticNode::span()'],['../classtvm_1_1DiagnosticBuilder.html#a52d9cc3cb33e655c5d82af47daa74c66',1,'tvm::DiagnosticBuilder::span()'],['../classtvm_1_1BaseExprNode.htm [...]
   ['span_2eh',['span.h',['../span_8h.html',1,'']]],
   ['spannode',['SpanNode',['../classtvm_1_1SpanNode.html',1,'tvm::SpanNode'],['../namespacetvm_1_1relay.html#a7d0fa6578e97d0d64b08865f94f04827',1,'tvm::relay::SpanNode()']]],
   ['sparse_5flhs',['sparse_lhs',['../structtvm_1_1relay_1_1SparseDenseAttrs.html#ae52d5465cb3421f342607abcc1cb1d5c',1,'tvm::relay::SparseDenseAttrs']]],
@@ -335,7 +335,7 @@ var searchData=
   ['stagenode',['StageNode',['../classtvm_1_1te_1_1StageNode.html',1,'tvm::te::StageNode'],['../classtvm_1_1auto__scheduler_1_1StageNode.html',1,'tvm::auto_scheduler::StageNode']]],
   ['stages',['stages',['../classtvm_1_1auto__scheduler_1_1StateNode.html#a881e14990bf228ee3fddb3721c451b9e',1,'tvm::auto_scheduler::StateNode::stages()'],['../classtvm_1_1te_1_1ScheduleNode.html#ab5649969db603d6b7b4d155c0d09cdd5',1,'tvm::te::ScheduleNode::stages()']]],
   ['stagetoaxesmap',['StageToAxesMap',['../namespacetvm_1_1auto__scheduler.html#a8f12e558fc4b8fbb990e7e204c06beeb',1,'tvm::auto_scheduler']]],
-  ['start',['start',['../structtvm_1_1relay_1_1ArangeAttrs.html#ae8ae5bc1551b406a4f52395af343c2ce',1,'tvm::relay::ArangeAttrs::start()'],['../classtvm_1_1script_1_1printer_1_1SliceDocNode.html#a16de0189a979a6cf9d8f14b39cb5fb54',1,'tvm::script::printer::SliceDocNode::start()'],['../classtvm_1_1runtime_1_1TimerNode.html#aa11fc338c39ee2137448e54a10efe0ae',1,'tvm::runtime::TimerNode::Start()'],['../classtvm_1_1runtime_1_1Timer.html#a89bcaa433499bc68902cb473d5eba6ca',1,'tvm::runtime::Timer::S [...]
+  ['start',['Start',['../classtvm_1_1runtime_1_1TimerNode.html#aa11fc338c39ee2137448e54a10efe0ae',1,'tvm::runtime::TimerNode::Start()'],['../classtvm_1_1runtime_1_1Timer.html#a89bcaa433499bc68902cb473d5eba6ca',1,'tvm::runtime::Timer::Start()'],['../classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#a44fadfb7b0f961a7fb2275e3b5dbcd88',1,'tvm::runtime::profiling::MetricCollectorNode::Start()'],['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#aee5452075c8e022b8aaa6fb365f68e14 [...]
   ['start_5findex',['start_index',['../namespacetvm_1_1topi_1_1nn.html#a752c4130dac73fd2de0390c5f6b24b15',1,'tvm::topi::nn']]],
   ['start_5fprofile_5fintrinsic',['start_profile_intrinsic',['../namespacetvm_1_1tir_1_1builtin.html#aa63c16281d1010a82a06eb771268b4e2',1,'tvm::tir::builtin']]],
   ['startcall',['StartCall',['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#a1fe322f7ba92be44d7e7c8cb184f3833',1,'tvm::runtime::profiling::Profiler']]],
@@ -378,9 +378,9 @@ var searchData=
   ['stmtsref',['StmtSRef',['../classtvm_1_1tir_1_1StmtSRef.html',1,'tvm::tir::StmtSRef'],['../classtvm_1_1tir_1_1StmtSRef.html#a31687ace5dc4fe487ffb87d658d86412',1,'tvm::tir::StmtSRef::StmtSRef()']]],
   ['stmtsrefnode',['StmtSRefNode',['../classtvm_1_1tir_1_1StmtSRefNode.html',1,'tvm::tir']]],
   ['stmtvisitor',['StmtVisitor',['../classtvm_1_1tir_1_1StmtVisitor.html',1,'tvm::tir']]],
-  ['stop',['stop',['../structtvm_1_1relay_1_1ArangeAttrs.html#a1eadf1f3964ca83dade8edeae7d6d7cf',1,'tvm::relay::ArangeAttrs::stop()'],['../classtvm_1_1script_1_1printer_1_1SliceDocNode.html#aaeb98937e7617cb76fb9662616b89e81',1,'tvm::script::printer::SliceDocNode::stop()'],['../classtvm_1_1runtime_1_1TimerNode.html#a67eb764f2c9e3fb7c2708f01c0c35683',1,'tvm::runtime::TimerNode::Stop()'],['../classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#aca9679dd49dfbc886b9dc99539cbf0e6',1,' [...]
+  ['stop',['Stop',['../classtvm_1_1runtime_1_1TimerNode.html#a67eb764f2c9e3fb7c2708f01c0c35683',1,'tvm::runtime::TimerNode::Stop()'],['../classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#aca9679dd49dfbc886b9dc99539cbf0e6',1,'tvm::runtime::profiling::MetricCollectorNode::Stop()'],['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#aa2000d8cd1970b5d29139ab1831394f0',1,'tvm::runtime::profiling::Profiler::Stop()'],['../structtvm_1_1relay_1_1ArangeAttrs.html#a1eadf1f3964ca83dad [...]
   ['stopcall',['StopCall',['../classtvm_1_1runtime_1_1profiling_1_1Profiler.html#ad5e6a8e8c9d915c80f494138eedfec3f',1,'tvm::runtime::profiling::Profiler']]],
-  ['storage',['Storage',['../classtvm_1_1runtime_1_1vm_1_1Storage.html',1,'tvm::runtime::vm::Storage'],['../classtvm_1_1runtime_1_1vm_1_1Storage.html#aff0c1264864e6205cfa468f069f62f55',1,'tvm::runtime::vm::Storage::Storage()'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a3412cabd3b4f42f106f56fc22257f6ca',1,'tvm::runtime::vm::Instruction::storage()']]],
+  ['storage',['Storage',['../classtvm_1_1runtime_1_1vm_1_1Storage.html',1,'tvm::runtime::vm::Storage'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a3412cabd3b4f42f106f56fc22257f6ca',1,'tvm::runtime::vm::Instruction::storage()'],['../classtvm_1_1runtime_1_1vm_1_1Storage.html#aff0c1264864e6205cfa468f069f62f55',1,'tvm::runtime::vm::Storage::Storage()']]],
   ['storage_5falign',['storage_align',['../classtvm_1_1auto__scheduler_1_1State.html#ab006690418e43cc9b7ad021c02657ed6',1,'tvm::auto_scheduler::State::storage_align()'],['../classtvm_1_1te_1_1Stage.html#aa73e3a269d84c3b4f0a1994371d67bab',1,'tvm::te::Stage::storage_align()']]],
   ['storage_5falignment',['storage_alignment',['../namespacetvm_1_1tir_1_1attr.html#af27d464f2065dc5f77408df7b94d4bb6',1,'tvm::tir::attr']]],
   ['storage_5fid',['storage_id',['../structTVMGraphExecutorGraphAttr.html#a8a0d6d05adcffbf499aafb6a6700c400',1,'TVMGraphExecutorGraphAttr']]],
@@ -398,7 +398,7 @@ var searchData=
   ['store_5fpredicate',['store_predicate',['../classtvm_1_1te_1_1StageNode.html#a8f4ba7f2931b3541c12734af511600a7',1,'tvm::te::StageNode']]],
   ['storenode',['StoreNode',['../classtvm_1_1tir_1_1StoreNode.html',1,'tvm::tir']]],
   ['str',['Str',['../classtvm_1_1script_1_1printer_1_1LiteralDoc.html#a8423a0ede5a29b24f6db5d9f6574a200',1,'tvm::script::printer::LiteralDoc::Str()'],['../classtvm_1_1TargetNode.html#a30cd67db46a9c4b098a8ba38fff22e26',1,'tvm::TargetNode::str()']]],
-  ['stream',['Stream',['../classtvm_1_1ReprLegacyPrinter.html#a8234b759c989ddac7fb24253d2e1e83d',1,'tvm::ReprLegacyPrinter::Stream()'],['../classtvm_1_1ReprPrinter.html#a036409dcdcf6f0ac5c6d7d27ec60ed94',1,'tvm::ReprPrinter::stream()']]],
+  ['stream',['stream',['../classtvm_1_1ReprPrinter.html#a036409dcdcf6f0ac5c6d7d27ec60ed94',1,'tvm::ReprPrinter::stream()'],['../classtvm_1_1ReprLegacyPrinter.html#a8234b759c989ddac7fb24253d2e1e83d',1,'tvm::ReprLegacyPrinter::Stream()']]],
   ['streamsync',['StreamSync',['../classtvm_1_1runtime_1_1DeviceAPI.html#ac29b9295c432a87658392872c644864f',1,'tvm::runtime::DeviceAPI']]],
   ['stride',['stride',['../structtvm_1_1relay_1_1YoloReorgAttrs.html#a720c79759a2f2b603b006b0bfcab5fe0',1,'tvm::relay::YoloReorgAttrs::stride()'],['../classtvm_1_1tir_1_1RampNode.html#a618c78e39d18f999dbdd3fe01436f447',1,'tvm::tir::RampNode::stride()']]],
   ['stride1',['stride1',['../structtvm_1_1relay_1_1CorrelationAttrs.html#aab5bf45d1aa0972bc1ce5cdb715546ff',1,'tvm::relay::CorrelationAttrs']]],
diff --git a/docs/reference/api/doxygen/search/all_15.js b/docs/reference/api/doxygen/search/all_15.js
index c3afb0d255..fa57a2ef3d 100644
--- a/docs/reference/api/doxygen/search/all_15.js
+++ b/docs/reference/api/doxygen/search/all_15.js
@@ -77,7 +77,7 @@ var searchData=
   ['te',['te',['../namespacetvm_1_1te.html',1,'tvm']]],
   ['tempexpr',['TempExpr',['../classtvm_1_1relay_1_1TempExpr.html',1,'tvm::relay']]],
   ['tempexprnode',['TempExprNode',['../classtvm_1_1relay_1_1TempExprNode.html',1,'tvm::relay']]],
-  ['tensor',['Tensor',['../classtvm_1_1te_1_1Tensor.html',1,'tvm::te::Tensor'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a22de469ea5521ba12e14f1e8181bae56',1,'tvm::runtime::vm::Instruction::tensor()'],['../classtvm_1_1te_1_1Tensor.html#afc8d8e74d1c840359661b39514d6fecf',1,'tvm::te::Tensor::Tensor()']]],
+  ['tensor',['Tensor',['../classtvm_1_1te_1_1Tensor.html',1,'tvm::te::Tensor'],['../classtvm_1_1te_1_1Tensor.html#afc8d8e74d1c840359661b39514d6fecf',1,'tvm::te::Tensor::Tensor()'],['../structtvm_1_1runtime_1_1vm_1_1Instruction.html#a22de469ea5521ba12e14f1e8181bae56',1,'tvm::runtime::vm::Instruction::tensor()']]],
   ['tensor_2eh',['tensor.h',['../tensor_8h.html',1,'']]],
   ['tensor_5fintrin',['tensor_intrin',['../classtvm_1_1te_1_1IterVarAttrNode.html#a6a0d96bbebfd716f851b2ad01738cb3f',1,'tvm::te::IterVarAttrNode']]],
   ['tensor_5fintrin_2eh',['tensor_intrin.h',['../tensor__intrin_8h.html',1,'']]],
@@ -96,7 +96,7 @@ var searchData=
   ['tensorintrincall',['TensorIntrinCall',['../classtvm_1_1te_1_1TensorIntrinCall.html',1,'tvm::te::TensorIntrinCall'],['../classtvm_1_1te_1_1TensorIntrinCall.html#a91c10074ce6babeba78fe72a0aab4b52',1,'tvm::te::TensorIntrinCall::TensorIntrinCall()']]],
   ['tensorintrincallnode',['TensorIntrinCallNode',['../classtvm_1_1te_1_1TensorIntrinCallNode.html',1,'tvm::te']]],
   ['tensorintrinnode',['TensorIntrinNode',['../classtvm_1_1te_1_1TensorIntrinNode.html',1,'tvm::te::TensorIntrinNode'],['../classtvm_1_1tir_1_1TensorIntrinNode.html',1,'tvm::tir::TensorIntrinNode'],['../classtvm_1_1te_1_1TensorIntrinNode.html#ad59e7f2b881fc798a8c64fd3959f929c',1,'tvm::te::TensorIntrinNode::TensorIntrinNode()']]],
-  ['tensorize',['Tensorize',['../classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf',1,'tvm::tir::ScheduleNode::Tensorize(const LoopRV &amp;loop_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928',1,'tvm::tir::ScheduleNode::Tensorize(const BlockRV &amp;block_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1te_1_1Stage.html#ab5fe485e1d730c36b096c06 [...]
+  ['tensorize',['tensorize',['../classtvm_1_1te_1_1Stage.html#ab5fe485e1d730c36b096c060b8d2ef9d',1,'tvm::te::Stage::tensorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf',1,'tvm::tir::ScheduleNode::Tensorize(const LoopRV &amp;loop_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928',1,'tvm::tir::ScheduleNode::Tensorize(const BlockRV &amp;block_rv, const String &amp [...]
   ['tensornode',['TensorNode',['../classtvm_1_1te_1_1TensorNode.html',1,'tvm::te::TensorNode'],['../classtvm_1_1te_1_1TensorNode.html#a153569448cb1bf9d2924d35639c3b8b8',1,'tvm::te::TensorNode::TensorNode()']]],
   ['tensors',['tensors',['../classtvm_1_1auto__scheduler_1_1ComputeDAGNode.html#afc71b9ecc0d6b82a5c2ab3250f01514b',1,'tvm::auto_scheduler::ComputeDAGNode::tensors()'],['../classtvm_1_1te_1_1TensorIntrinCallNode.html#a92b543750ea55b9cfd6852139e2ddbd6',1,'tvm::te::TensorIntrinCallNode::tensors()']]],
   ['tensortype',['TensorType',['../classtvm_1_1TensorType.html',1,'tvm::TensorType'],['../classtvm_1_1TensorType.html#ade4460e9b02b42757a83808dec478b87',1,'tvm::TensorType::TensorType()'],['../namespacetvm_1_1relay.html#a52c13723bba53f4953dfd10c34d480f8',1,'tvm::relay::TensorType()']]],
diff --git a/docs/reference/api/doxygen/search/all_16.js b/docs/reference/api/doxygen/search/all_16.js
index ceaaba7045..bd11de3019 100644
--- a/docs/reference/api/doxygen/search/all_16.js
+++ b/docs/reference/api/doxygen/search/all_16.js
@@ -40,7 +40,7 @@ var searchData=
   ['unionregion',['UnionRegion',['../namespacetvm_1_1arith.html#ad27c4f216e41eb8e81296fb7ec4b9453',1,'tvm::arith']]],
   ['unionregionlowerbound',['UnionRegionLowerBound',['../namespacetvm_1_1arith.html#a4c3dedfa4cba4ad39c953eb51eb83e4d',1,'tvm::arith']]],
   ['unipolar',['unipolar',['../structtvm_1_1relay_1_1BinaryConv2DAttrs.html#a7e0ad68dce226079b769a678aa01dc49',1,'tvm::relay::BinaryConv2DAttrs::unipolar()'],['../structtvm_1_1relay_1_1BinaryDenseAttrs.html#af21cdb9dac67ab9ecea5a19642658d8a',1,'tvm::relay::BinaryDenseAttrs::unipolar()']]],
-  ['unique',['unique',['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()'],['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()']]],
+  ['unique',['Unique',['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()'],['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()']]],
   ['uniqueattrs',['UniqueAttrs',['../structtvm_1_1relay_1_1UniqueAttrs.html',1,'tvm::relay']]],
   ['uniqueglobalfor',['UniqueGlobalFor',['../classtvm_1_1GlobalVarSupplyNode.html#af67bad5d9d93381c440a7886cbef430a',1,'tvm::GlobalVarSupplyNode']]],
   ['unit_5fbits',['unit_bits',['../classtvm_1_1MemoryInfoNode.html#a505c2f2dd0dd0c28a12b9113e2176a8d',1,'tvm::MemoryInfoNode']]],
@@ -49,7 +49,7 @@ var searchData=
   ['unknownattributeaccesspathnode',['UnknownAttributeAccessPathNode',['../classtvm_1_1UnknownAttributeAccessPathNode.html',1,'tvm::UnknownAttributeAccessPathNode'],['../classtvm_1_1UnknownAttributeAccessPathNode.html#a1882e9e591466a2785acc761dc63d56e',1,'tvm::UnknownAttributeAccessPathNode::UnknownAttributeAccessPathNode()']]],
   ['unmatchedcases',['UnmatchedCases',['../namespacetvm_1_1relay.html#aa3a8cace40f8056fd6412f39c3eaa605',1,'tvm::relay']]],
   ['unravel_5findex',['unravel_index',['../namespacetvm_1_1topi.html#a8811a02532bbe3047986bf1a8449ac0e',1,'tvm::topi']]],
-  ['unroll',['Unroll',['../classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552',1,'tvm::tir::ScheduleNode::Unroll()'],['../classtvm_1_1auto__scheduler_1_1State.html#aa68a9d2e226bae38a36e4be4af1d1ae4',1,'tvm::auto_scheduler::State::unroll()'],['../classtvm_1_1te_1_1Stage.html#af83ad8672660403504f472228b044b33',1,'tvm::te::Stage::unroll()'],['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3',1,'tvm::script::ir_builder::tir::Unroll()']]],
+  ['unroll',['unroll',['../classtvm_1_1auto__scheduler_1_1State.html#aa68a9d2e226bae38a36e4be4af1d1ae4',1,'tvm::auto_scheduler::State::unroll()'],['../classtvm_1_1te_1_1Stage.html#af83ad8672660403504f472228b044b33',1,'tvm::te::Stage::unroll()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552',1,'tvm::tir::ScheduleNode::Unroll()'],['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3',1,'tvm::script::ir_builder::tir::Unroll()']]],
   ['unrollloop',['UnrollLoop',['../namespacetvm_1_1tir_1_1transform.html#ab2f279e91071fa96a1edb24fa004ea6a',1,'tvm::tir::transform']]],
   ['update',['update',['../classtvm_1_1te_1_1ScanOpNode.html#ace2bf7e43cd4197324ec6363626fc60a',1,'tvm::te::ScanOpNode::update()'],['../classtvm_1_1arith_1_1ConstIntBoundAnalyzer.html#a5ae0699196c4bbc754bbdd4c3a6c7ca7',1,'tvm::arith::ConstIntBoundAnalyzer::Update()'],['../classtvm_1_1arith_1_1ModularSetAnalyzer.html#a04156fac580981f3005af3b8e676720d',1,'tvm::arith::ModularSetAnalyzer::Update()'],['../classtvm_1_1arith_1_1RewriteSimplifier.html#a5e6752c0702dc2d3e4235797d9d3ac7b',1,'tvm::a [...]
   ['update_5ffunc',['update_func',['../classtvm_1_1auto__scheduler_1_1PythonBasedModelNode.html#ade9364c152a36501d4f24fa4f0111519',1,'tvm::auto_scheduler::PythonBasedModelNode']]],
diff --git a/docs/reference/api/doxygen/search/all_17.js b/docs/reference/api/doxygen/search/all_17.js
index a1364d4101..b7a861b1bd 100644
--- a/docs/reference/api/doxygen/search/all_17.js
+++ b/docs/reference/api/doxygen/search/all_17.js
@@ -33,7 +33,7 @@ var searchData=
   ['vector_5funit_5fbytes',['vector_unit_bytes',['../classtvm_1_1auto__scheduler_1_1HardwareParamsNode.html#a6f2dd9161fdb3233417a9912c8854434',1,'tvm::auto_scheduler::HardwareParamsNode']]],
   ['vectorcombine',['vectorcombine',['../namespacetvm_1_1tir_1_1builtin.html#a30dff65bc2c142b57fae7f60e378ff43',1,'tvm::tir::builtin']]],
   ['vectorhigh',['vectorhigh',['../namespacetvm_1_1tir_1_1builtin.html#a45bf65ca7ca01d2016e0b609117d7e25',1,'tvm::tir::builtin']]],
-  ['vectorize',['vectorize',['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()']]],
+  ['vectorize',['Vectorize',['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()'],['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()']]],
   ['vectorized',['Vectorized',['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2b3df6013772bdd23dfa872b81555272',1,'tvm::script::ir_builder::tir']]],
   ['vectorizeloop',['VectorizeLoop',['../namespacetvm_1_1tir_1_1transform.html#af3cecb50a8b8fc8021f6a87bc27587da',1,'tvm::tir::transform']]],
   ['vectorizer',['Vectorizer',['../classtvm_1_1tir_1_1BufferLoadNode.html#a842a72b9d02a9f8541b512478932fece',1,'tvm::tir::BufferLoadNode']]],
diff --git a/docs/reference/api/doxygen/search/all_7.js b/docs/reference/api/doxygen/search/all_7.js
index 4608b7fed7..7d02830beb 100644
--- a/docs/reference/api/doxygen/search/all_7.js
+++ b/docs/reference/api/doxygen/search/all_7.js
@@ -300,7 +300,7 @@ var searchData=
   ['funcname',['FuncName',['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6abb6566a443b59385ec26213336f1c0',1,'tvm::script::ir_builder::tir']]],
   ['funcret',['FuncRet',['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4d377c370c4d836034937d558f6fba11',1,'tvm::script::ir_builder::tir']]],
   ['funcs',['funcs',['../structTVMFuncRegistry.html#a25badb00e205aaa5c317bd61a4b88d96',1,'TVMFuncRegistry']]],
-  ['function',['Function',['../classtvm_1_1relay_1_1Function.html',1,'tvm::relay::Function'],['../classtvm_1_1relay_1_1DFPatternCallbackNode.html#a878e6e49af2466c49ffd9fcfe7f609fa',1,'tvm::relay::DFPatternCallbackNode::function()'],['../classtvm_1_1relay_1_1Function.html#a11ee77c0df8aa1c2c072c7cf613b9238',1,'tvm::relay::Function::Function()']]],
+  ['function',['Function',['../classtvm_1_1relay_1_1Function.html',1,'tvm::relay::Function'],['../classtvm_1_1relay_1_1Function.html#a11ee77c0df8aa1c2c072c7cf613b9238',1,'tvm::relay::Function::Function()'],['../classtvm_1_1relay_1_1DFPatternCallbackNode.html#a878e6e49af2466c49ffd9fcfe7f609fa',1,'tvm::relay::DFPatternCallbackNode::function()']]],
   ['function_2eh',['function.h',['../ir_2function_8h.html',1,'(Global Namespace)'],['../relay_2function_8h.html',1,'(Global Namespace)'],['../tir_2function_8h.html',1,'(Global Namespace)']]],
   ['functiondoc',['FunctionDoc',['../classtvm_1_1script_1_1printer_1_1FunctionDoc.html',1,'tvm::script::printer::FunctionDoc'],['../classtvm_1_1script_1_1printer_1_1FunctionDoc.html#ac7ed2ed1c4c3cf89ff1b9bd58583c79d',1,'tvm::script::printer::FunctionDoc::FunctionDoc()']]],
   ['functiondocnode',['FunctionDocNode',['../classtvm_1_1script_1_1printer_1_1FunctionDocNode.html',1,'tvm::script::printer']]],
diff --git a/docs/reference/api/doxygen/search/all_e.js b/docs/reference/api/doxygen/search/all_e.js
index 7fccd89ba6..f28d30a05e 100644
--- a/docs/reference/api/doxygen/search/all_e.js
+++ b/docs/reference/api/doxygen/search/all_e.js
@@ -73,7 +73,7 @@ var searchData=
   ['matmulattrs',['MatmulAttrs',['../structtvm_1_1relay_1_1MatmulAttrs.html',1,'tvm::relay']]],
   ['matrix_5fset_5fdiag',['matrix_set_diag',['../namespacetvm_1_1topi.html#aead477c6c9d4f4589d22b8acff82040c',1,'tvm::topi']]],
   ['matrixsetdiagattrs',['MatrixSetDiagAttrs',['../structtvm_1_1relay_1_1MatrixSetDiagAttrs.html',1,'tvm::relay']]],
-  ['max',['Max',['../classtvm_1_1tir_1_1Max.html',1,'tvm::tir::Max'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, [...]
+  ['max',['Max',['../classtvm_1_1tir_1_1Max.html',1,'tvm::tir::Max'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, [...]
   ['max_5fcontinuous_5ferror',['max_continuous_error',['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#abdc38da91bcdf77be765c1e3d5af3648',1,'tvm::auto_scheduler::ProgramMeasurerNode']]],
   ['max_5fdisplacement',['max_displacement',['../structtvm_1_1relay_1_1CorrelationAttrs.html#ad1d16e2ba537736c8baee2553e1e32bf',1,'tvm::relay::CorrelationAttrs']]],
   ['max_5ffunctions',['max_functions',['../structTVMMutableFuncRegistry.html#a41745f8e0f73f8e4fb2074f5b154b49c',1,'TVMMutableFuncRegistry']]],
@@ -177,7 +177,7 @@ var searchData=
   ['microtvmruntimegetoutput',['MicroTVMRuntimeGetOutput',['../microtvm__runtime_8h.html#a76129be7b6de972791a3f9a1b312acfa',1,'microtvm_runtime.h']]],
   ['microtvmruntimerun',['MicroTVMRuntimeRun',['../microtvm__runtime_8h.html#ac43a544f675dd716e8c279c3e41f6e45',1,'microtvm_runtime.h']]],
   ['microtvmruntimesetinput',['MicroTVMRuntimeSetInput',['../microtvm__runtime_8h.html#aa593edc600f4356f2b560702aa01b113',1,'microtvm_runtime.h']]],
-  ['min',['Min',['../classtvm_1_1tir_1_1Min.html',1,'tvm::tir::Min'],['../classtvm_1_1tir_1_1Min.html#a3a4403aec40029a5206e22cd334e356b',1,'tvm::tir::Min::Min()'],['../classtvm_1_1RangeNode.html#a43d2fb12bb61cf05936a1972d0158b49',1,'tvm::RangeNode::min()'],['../classtvm_1_1tir_1_1ForNode.html#a1d1aa2006328bd84e4911f6d43ceca5c',1,'tvm::tir::ForNode::min()'],['../classtvm_1_1arith_1_1IntSet.html#ae5517de2862e93a801224eed98a57001',1,'tvm::arith::IntSet::min()'],['../classtvm_1_1support_1_1L [...]
+  ['min',['Min',['../classtvm_1_1tir_1_1Min.html',1,'tvm::tir::Min'],['../classtvm_1_1RangeNode.html#a43d2fb12bb61cf05936a1972d0158b49',1,'tvm::RangeNode::min()'],['../classtvm_1_1tir_1_1ForNode.html#a1d1aa2006328bd84e4911f6d43ceca5c',1,'tvm::tir::ForNode::min()'],['../classtvm_1_1arith_1_1IntSet.html#ae5517de2862e93a801224eed98a57001',1,'tvm::arith::IntSet::min()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#aec5f11b588fa3a12294a46c945c34411',1,'tvm::support::LinearCongrue [...]
   ['min_5frepeat_5fms',['min_repeat_ms',['../classtvm_1_1auto__scheduler_1_1ProgramRunnerNode.html#a39a865216db9ed6f57dfb22160cae1ff',1,'tvm::auto_scheduler::ProgramRunnerNode']]],
   ['min_5fvalue',['min_value',['../classtvm_1_1arith_1_1ConstIntBoundNode.html#a0761897bf16ab73b848bf360e9b195a3',1,'tvm::arith::ConstIntBoundNode::min_value()'],['../namespacetvm.html#a3b37fa55ea93d6868751a2441996b072',1,'tvm::min_value()']]],
   ['minimum',['minimum',['../namespacetvm_1_1topi.html#a7ac1dc0d99ce93090a4cdf90ab19d4b8',1,'tvm::topi::minimum(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a0e19dc06a2b1ecbb83b0942fdf836169',1,'tvm::topi::minimum(const tvm::te::Tensor &amp;A, const tvm::te::Tensor &amp;B, std::string name=&quot;T_&quot; &quot;minimum&quot;, std::string tag=kBroadcast)'],['../namespacetvm_1_1topi.html#a28d4ef4b3426bff237215ce356dd5681',1,'tvm::topi::minimum(con [...]
@@ -195,7 +195,7 @@ var searchData=
   ['mixedmodulepassmanager',['MixedModulePassManager',['../namespacetvm.html#abc01352eff102d4902632d097adc0e08',1,'tvm']]],
   ['mma_5ffill',['mma_fill',['../namespacetvm_1_1tir_1_1builtin.html#a307667c449c54cef747d781771f79bab',1,'tvm::tir::builtin']]],
   ['mma_5fstore',['mma_store',['../namespacetvm_1_1tir_1_1builtin.html#a772fb68f083e71e635c50bb503903f22',1,'tvm::tir::builtin']]],
-  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html',1,'tvm::tir::Mod'],['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#ab2fb058ca54af03b5bc47bf4fac23cf7',1,'tvm::meta_schedule::BuilderInputNode::mod()'],['../classtvm_1_1meta__schedule_1_1WorkloadNode.html#a3929f2761c168c25de6be2247b913911',1,'tvm::meta_schedule::WorkloadNode::mod()'],['../classtvm_1_1meta__schedule_1_1ExtractedTaskNode.html#a50c40aa8beb57d0f31c36ef360042be6',1,'tvm::meta_schedule::ExtractedTaskNode::mod()'],['../c [...]
+  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html',1,'tvm::tir::Mod'],['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../classtvm_1_1meta__schedule_1_1BuilderInputNode.html#ab2fb058ca54af03b5bc47bf4fac23cf7',1,'tvm::meta_schedule::BuilderInputNode::mod()'],['../classtvm_1_1meta__schedule_1_1WorkloadNode.html#a3929f2761c168c25de6be2247b913911',1,'tvm::meta_schedule::WorkloadNode::mod()'],['../classtvm_1_1meta__schedule_1_1ExtractedTaskNode.ht [...]
   ['mod_5fname',['mod_name',['../structTVMMetadata.html#a32e45fcae0f9328e944a35a885d94276',1,'TVMMetadata::mod_name()'],['../classtvm_1_1runtime_1_1metadata_1_1MetadataNode.html#a1c05bb5eb88b5d55b3abeeb2de263191',1,'tvm::runtime::metadata::MetadataNode::mod_name()']]],
   ['mode',['mode',['../structtvm_1_1relay_1_1MirrorPadAttrs.html#af5381d72f1d9c9abcb9d2e522966ad86',1,'tvm::relay::MirrorPadAttrs::mode()'],['../structtvm_1_1relay_1_1SubPixelAttrs.html#a6f0822aa1ad7672a18ab73c64e83fa99',1,'tvm::relay::SubPixelAttrs::mode()'],['../structtvm_1_1relay_1_1ScatterNDAttrs.html#ab13eeaa700fe7e41666ac04179e0fd62',1,'tvm::relay::ScatterNDAttrs::mode()'],['../structtvm_1_1relay_1_1TakeAttrs.html#a0bf9d25ced9bfc91e766494e5f641e70',1,'tvm::relay::TakeAttrs::mode()' [...]
   ['modnode',['ModNode',['../classtvm_1_1tir_1_1ModNode.html',1,'tvm::tir']]],
diff --git a/docs/reference/api/doxygen/search/all_f.js b/docs/reference/api/doxygen/search/all_f.js
index 54ce9182aa..a8bb513026 100644
--- a/docs/reference/api/doxygen/search/all_f.js
+++ b/docs/reference/api/doxygen/search/all_f.js
@@ -3,7 +3,7 @@ var searchData=
   ['n_5ffft',['n_fft',['../structtvm_1_1relay_1_1StftAttrs.html#a4363adcb567c32e1b814ece68d321b3e',1,'tvm::relay::StftAttrs']]],
   ['n_5fparallel',['n_parallel',['../classtvm_1_1auto__scheduler_1_1ProgramBuilderNode.html#aaa056bbd3887751729ca207de8c6c792',1,'tvm::auto_scheduler::ProgramBuilderNode::n_parallel()'],['../classtvm_1_1auto__scheduler_1_1RPCRunnerNode.html#a33dc20205dc39f9b23a4d25939298923',1,'tvm::auto_scheduler::RPCRunnerNode::n_parallel()']]],
   ['n_5fsplit',['n_split',['../classtvm_1_1auto__scheduler_1_1FollowSplitStepNode.html#a365ea1b81b21e382ac350a66326f1b86',1,'tvm::auto_scheduler::FollowSplitStepNode']]],
-  ['name',['name',['../classtvm_1_1auto__scheduler_1_1IteratorNode.html#afb44a9dd3077356de0545daf7899b402',1,'tvm::auto_scheduler::IteratorNode::name()'],['../classtvm_1_1AttrFieldInfoNode.html#ad918a5375ed2f8db7557ba50f95e026b',1,'tvm::AttrFieldInfoNode::name()'],['../classtvm_1_1EnvFuncNode.html#a02d22636502da6754661d3698ea338ad',1,'tvm::EnvFuncNode::name()'],['../classtvm_1_1instrument_1_1PassInstrumentNode.html#abb8ed0c496c5f9d85b33903a10954cbc',1,'tvm::instrument::PassInstrumentNode [...]
+  ['name',['Name',['../classtvm_1_1script_1_1ir__builder_1_1IRBuilder.html#ace475c7a85ef508d912beba48ef4183a',1,'tvm::script::ir_builder::IRBuilder::Name()'],['../classtvm_1_1script_1_1ir__builder_1_1details_1_1Namer.html#a22da1264beaa8681bab998a5c597369f',1,'tvm::script::ir_builder::details::Namer::Name()'],['../classtvm_1_1auto__scheduler_1_1IteratorNode.html#afb44a9dd3077356de0545daf7899b402',1,'tvm::auto_scheduler::IteratorNode::name()'],['../classtvm_1_1AttrFieldInfoNode.html#ad918a [...]
   ['name_5f',['name_',['../classtvm_1_1runtime_1_1Registry.html#a4d8221b67729bafee4c2c5b424ed80ea',1,'tvm::runtime::Registry::name_()'],['../classtvm_1_1GenericFuncNode.html#ade1da360d3e314360fd5399b2d76d3a1',1,'tvm::GenericFuncNode::name_()']]],
   ['name_5fhint',['name_hint',['../classtvm_1_1ConstructorNode.html#ad94b373e9c4669fc5d472a9194483d66',1,'tvm::ConstructorNode::name_hint()'],['../classtvm_1_1GlobalVarNode.html#ab82974132026f07d89afcf409a2ca616',1,'tvm::GlobalVarNode::name_hint()'],['../structtvm_1_1ConstantInfoNode.html#a054eb75fb628ae03d5cb4c4c7f7e8846',1,'tvm::ConstantInfoNode::name_hint()'],['../classtvm_1_1TypeVarNode.html#a8f040892b1484a503c58870b4f0b70f6',1,'tvm::TypeVarNode::name_hint()'],['../classtvm_1_1Global [...]
   ['name_5fsupply_2eh',['name_supply.h',['../name__supply_8h.html',1,'']]],
diff --git a/docs/reference/api/doxygen/search/functions_12.js b/docs/reference/api/doxygen/search/functions_12.js
index 256f6fa8a0..2a01386a61 100644
--- a/docs/reference/api/doxygen/search/functions_12.js
+++ b/docs/reference/api/doxygen/search/functions_12.js
@@ -36,6 +36,8 @@ var searchData=
   ['registerorget',['RegisterOrGet',['../classtvm_1_1OpRegEntry.html#a39a4d3e7f905eb4e29ca464bcedb05bd',1,'tvm::OpRegEntry::RegisterOrGet()'],['../classtvm_1_1relay_1_1ExecutorRegEntry.html#a03347a2b68269b853a7c0399994951ef',1,'tvm::relay::ExecutorRegEntry::RegisterOrGet()'],['../classtvm_1_1relay_1_1RuntimeRegEntry.html#ae8b479159ccd8b35b75950fcda58dd9d',1,'tvm::relay::RuntimeRegEntry::RegisterOrGet()'],['../classtvm_1_1TargetTagRegEntry.html#a07e0631600484dc0985ca62b1620461c',1,'tvm::T [...]
   ['registry',['Registry',['../classtvm_1_1ReflectionVTable_1_1Registry.html#ac8f4637640aa9dffed745303a4cfa827',1,'tvm::ReflectionVTable::Registry']]],
   ['reindex',['ReIndex',['../classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5',1,'tvm::tir::ScheduleNode']]],
+  ['reindexcacheread',['ReindexCacheRead',['../classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3',1,'tvm::tir::ScheduleNode']]],
+  ['reindexcachewrite',['ReindexCacheWrite',['../classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729',1,'tvm::tir::ScheduleNode']]],
   ['reinterpret',['Reinterpret',['../classtvm_1_1BaseValueHash.html#a5ca1332d212969bb1bf6c865d5703b39',1,'tvm::BaseValueHash::Reinterpret()'],['../namespacetvm_1_1tir_1_1builtin.html#a7b555bc5cca2f5e7b26c1037bc0001ce',1,'tvm::tir::builtin::reinterpret()'],['../namespacetvm.html#a34084606675cd2c73c6b0f10e1618280',1,'tvm::reinterpret()'],['../namespacetvm_1_1topi.html#a25239505894bdae140e53f4abc146f92',1,'tvm::topi::reinterpret()']]],
   ['relayreftype',['RelayRefType',['../classtvm_1_1RelayRefType.html#a36d7d0022c0bd3218269356bd368b9fd',1,'tvm::RelayRefType']]],
   ['relayreftypenode',['RelayRefTypeNode',['../classtvm_1_1RelayRefTypeNode.html#a23840592da71d366e7f7dbaaf42d906d',1,'tvm::RelayRefTypeNode']]],
@@ -75,7 +77,7 @@ var searchData=
   ['reserve',['reserve',['../classtvm_1_1runtime_1_1Array.html#a1a7727b86efaf35c58a5198ab1c139c8',1,'tvm::runtime::Array']]],
   ['reserveglobalvar',['ReserveGlobalVar',['../classtvm_1_1GlobalVarSupplyNode.html#a29185b94238fc62c928346a004c43b16',1,'tvm::GlobalVarSupplyNode']]],
   ['reservename',['ReserveName',['../classtvm_1_1NameSupplyNode.html#a9feb960ebeeee03fb9c5105655a8da17',1,'tvm::NameSupplyNode']]],
-  ['reset',['Reset',['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#a73b14ea360a9902c291d5bf6e97636cd',1,'tvm::auto_scheduler::ProgramMeasurerNode::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Unframer.html#ae6279154fe70e9eb85937b51e70a4bf8',1,'tvm::runtime::micro_rpc::Unframer::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Framer.html#a44ff9650ecca8785e33c25c369d2570a',1,'tvm::runtime::micro_rpc::Framer::Reset()'],['../classtvm_1_1tir_1_1StmtSRefNode.html#a0a81 [...]
+  ['reset',['reset',['../classtvm_1_1runtime_1_1NDArray.html#af2a8ccab95d432d1ecad7a389e11bcd3',1,'tvm::runtime::NDArray::reset()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#ac4461465ba0e785794794e0405c96590',1,'tvm::runtime::ObjectPtr::reset()'],['../classtvm_1_1auto__scheduler_1_1ProgramMeasurerNode.html#a73b14ea360a9902c291d5bf6e97636cd',1,'tvm::auto_scheduler::ProgramMeasurerNode::Reset()'],['../classtvm_1_1runtime_1_1micro__rpc_1_1Unframer.html#ae6279154fe70e9eb85937b51e70a4bf8',1, [...]
   ['reset_5fattr',['reset_attr',['../classtvm_1_1OpRegEntry.html#a67628f8d3d6dea5b0a47e462c06b7790',1,'tvm::OpRegEntry']]],
   ['resetthreadpool',['ResetThreadPool',['../namespacetvm_1_1runtime_1_1threading.html#aafdb21c00248ff146b614a7e888b4fd7',1,'tvm::runtime::threading']]],
   ['reshape',['reshape',['../namespacetvm_1_1topi.html#a3aad65f2505802109ba7d05359ce9005',1,'tvm::topi']]],
diff --git a/docs/reference/api/doxygen/search/functions_14.js b/docs/reference/api/doxygen/search/functions_14.js
index 5f8a2ae072..75dc4a7b6f 100644
--- a/docs/reference/api/doxygen/search/functions_14.js
+++ b/docs/reference/api/doxygen/search/functions_14.js
@@ -21,7 +21,7 @@ var searchData=
   ['tensorintrin',['TensorIntrin',['../classtvm_1_1te_1_1TensorIntrin.html#a4ff4237911227bf80b3076906dc3b7ea',1,'tvm::te::TensorIntrin::TensorIntrin()'],['../classtvm_1_1tir_1_1TensorIntrin.html#af5a94c7b098b56056e02eaf187e6871c',1,'tvm::tir::TensorIntrin::TensorIntrin()']]],
   ['tensorintrincall',['TensorIntrinCall',['../classtvm_1_1te_1_1TensorIntrinCall.html#a91c10074ce6babeba78fe72a0aab4b52',1,'tvm::te::TensorIntrinCall']]],
   ['tensorintrinnode',['TensorIntrinNode',['../classtvm_1_1te_1_1TensorIntrinNode.html#ad59e7f2b881fc798a8c64fd3959f929c',1,'tvm::te::TensorIntrinNode']]],
-  ['tensorize',['Tensorize',['../classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf',1,'tvm::tir::ScheduleNode::Tensorize(const LoopRV &amp;loop_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928',1,'tvm::tir::ScheduleNode::Tensorize(const BlockRV &amp;block_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1te_1_1Stage.html#ab5fe485e1d730c36b096c06 [...]
+  ['tensorize',['tensorize',['../classtvm_1_1te_1_1Stage.html#ab5fe485e1d730c36b096c060b8d2ef9d',1,'tvm::te::Stage::tensorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf',1,'tvm::tir::ScheduleNode::Tensorize(const LoopRV &amp;loop_rv, const String &amp;intrin, bool preserve_unit_iters=true)=0'],['../classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928',1,'tvm::tir::ScheduleNode::Tensorize(const BlockRV &amp;block_rv, const String &amp [...]
   ['tensornode',['TensorNode',['../classtvm_1_1te_1_1TensorNode.html#a153569448cb1bf9d2924d35639c3b8b8',1,'tvm::te::TensorNode']]],
   ['tensortype',['TensorType',['../classtvm_1_1TensorType.html#ade4460e9b02b42757a83808dec478b87',1,'tvm::TensorType']]],
   ['terminalrenderer',['TerminalRenderer',['../namespacetvm.html#a69a0e3f559d3a3b98d42701117d93ed0',1,'tvm']]],
diff --git a/docs/reference/api/doxygen/search/functions_15.js b/docs/reference/api/doxygen/search/functions_15.js
index c968031b1f..d28dad9eed 100644
--- a/docs/reference/api/doxygen/search/functions_15.js
+++ b/docs/reference/api/doxygen/search/functions_15.js
@@ -37,12 +37,12 @@ var searchData=
   ['unionlowerbound',['UnionLowerBound',['../namespacetvm_1_1arith.html#ab22d7fd95abb5fa372843a40e19d80c5',1,'tvm::arith']]],
   ['unionregion',['UnionRegion',['../namespacetvm_1_1arith.html#ad27c4f216e41eb8e81296fb7ec4b9453',1,'tvm::arith']]],
   ['unionregionlowerbound',['UnionRegionLowerBound',['../namespacetvm_1_1arith.html#a4c3dedfa4cba4ad39c953eb51eb83e4d',1,'tvm::arith']]],
-  ['unique',['unique',['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()'],['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()']]],
+  ['unique',['Unique',['../classtvm_1_1VirtualDeviceCache.html#a25ba1351484aa58a2cc7cef8f8e4423c',1,'tvm::VirtualDeviceCache::Unique()'],['../classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7',1,'tvm::runtime::Object::unique()'],['../classtvm_1_1runtime_1_1ObjectPtr.html#af95c6c6fcd89da0f62b93f1167b72314',1,'tvm::runtime::ObjectPtr::unique()'],['../classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7',1,'tvm::runtime::ObjectRef::unique()']]],
   ['uniqueglobalfor',['UniqueGlobalFor',['../classtvm_1_1GlobalVarSupplyNode.html#af67bad5d9d93381c440a7886cbef430a',1,'tvm::GlobalVarSupplyNode']]],
   ['unknownattributeaccesspathnode',['UnknownAttributeAccessPathNode',['../classtvm_1_1UnknownAttributeAccessPathNode.html#a1882e9e591466a2785acc761dc63d56e',1,'tvm::UnknownAttributeAccessPathNode']]],
   ['unmatchedcases',['UnmatchedCases',['../namespacetvm_1_1relay.html#aa3a8cace40f8056fd6412f39c3eaa605',1,'tvm::relay']]],
   ['unravel_5findex',['unravel_index',['../namespacetvm_1_1topi.html#a8811a02532bbe3047986bf1a8449ac0e',1,'tvm::topi']]],
-  ['unroll',['Unroll',['../classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552',1,'tvm::tir::ScheduleNode::Unroll()'],['../classtvm_1_1auto__scheduler_1_1State.html#aa68a9d2e226bae38a36e4be4af1d1ae4',1,'tvm::auto_scheduler::State::unroll()'],['../classtvm_1_1te_1_1Stage.html#af83ad8672660403504f472228b044b33',1,'tvm::te::Stage::unroll()'],['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3',1,'tvm::script::ir_builder::tir::Unroll()']]],
+  ['unroll',['unroll',['../classtvm_1_1auto__scheduler_1_1State.html#aa68a9d2e226bae38a36e4be4af1d1ae4',1,'tvm::auto_scheduler::State::unroll()'],['../classtvm_1_1te_1_1Stage.html#af83ad8672660403504f472228b044b33',1,'tvm::te::Stage::unroll()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552',1,'tvm::tir::ScheduleNode::Unroll()'],['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3',1,'tvm::script::ir_builder::tir::Unroll()']]],
   ['unrollloop',['UnrollLoop',['../namespacetvm_1_1tir_1_1transform.html#ab2f279e91071fa96a1edb24fa004ea6a',1,'tvm::tir::transform']]],
   ['update',['Update',['../classtvm_1_1arith_1_1ConstIntBoundAnalyzer.html#a5ae0699196c4bbc754bbdd4c3a6c7ca7',1,'tvm::arith::ConstIntBoundAnalyzer::Update()'],['../classtvm_1_1arith_1_1ModularSetAnalyzer.html#a04156fac580981f3005af3b8e676720d',1,'tvm::arith::ModularSetAnalyzer::Update()'],['../classtvm_1_1arith_1_1RewriteSimplifier.html#a5e6752c0702dc2d3e4235797d9d3ac7b',1,'tvm::arith::RewriteSimplifier::Update()'],['../classtvm_1_1arith_1_1CanonicalSimplifier.html#a790c032e12c7d93e9e940 [...]
   ['updatecostmodel',['UpdateCostModel',['../classtvm_1_1meta__schedule_1_1MeasureCallback.html#afdf5503c6e6f53767de132d91a7b53f9',1,'tvm::meta_schedule::MeasureCallback']]],
diff --git a/docs/reference/api/doxygen/search/functions_16.js b/docs/reference/api/doxygen/search/functions_16.js
index f2ecc56a0e..63330900b8 100644
--- a/docs/reference/api/doxygen/search/functions_16.js
+++ b/docs/reference/api/doxygen/search/functions_16.js
@@ -8,7 +8,7 @@ var searchData=
   ['vector',['Vector',['../classtvm_1_1arith_1_1IntSet.html#a29b6f1e60f4b328fcfabc514e0c10f17',1,'tvm::arith::IntSet']]],
   ['vectorcombine',['vectorcombine',['../namespacetvm_1_1tir_1_1builtin.html#a30dff65bc2c142b57fae7f60e378ff43',1,'tvm::tir::builtin']]],
   ['vectorhigh',['vectorhigh',['../namespacetvm_1_1tir_1_1builtin.html#a45bf65ca7ca01d2016e0b609117d7e25',1,'tvm::tir::builtin']]],
-  ['vectorize',['vectorize',['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()'],['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()']]],
+  ['vectorize',['Vectorize',['../classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee',1,'tvm::tir::ScheduleNode::Vectorize()'],['../classtvm_1_1auto__scheduler_1_1State.html#a97b8a21210d63bea241dbab085d89b53',1,'tvm::auto_scheduler::State::vectorize()'],['../classtvm_1_1te_1_1Stage.html#a44d33e3920106e75dc7c68272f880812',1,'tvm::te::Stage::vectorize()']]],
   ['vectorized',['Vectorized',['../namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2b3df6013772bdd23dfa872b81555272',1,'tvm::script::ir_builder::tir']]],
   ['vectorizeloop',['VectorizeLoop',['../namespacetvm_1_1tir_1_1transform.html#af3cecb50a8b8fc8021f6a87bc27587da',1,'tvm::tir::transform']]],
   ['vectorjacobianproduct',['VectorJacobianProduct',['../namespacetvm_1_1te.html#a547183f5a311af53ab598faba423fd64',1,'tvm::te']]],
diff --git a/docs/reference/api/doxygen/search/functions_d.js b/docs/reference/api/doxygen/search/functions_d.js
index ba58483ed7..87821c714d 100644
--- a/docs/reference/api/doxygen/search/functions_d.js
+++ b/docs/reference/api/doxygen/search/functions_d.js
@@ -38,7 +38,7 @@ var searchData=
   ['matchrange',['MatchRange',['../classtvm_1_1arith_1_1IntSet.html#a2f2999336fbba4f436b66bdddce5c57a',1,'tvm::arith::IntSet']]],
   ['matmul',['matmul',['../namespacetvm_1_1topi.html#adae7dcb7e951109ba72192202d182994',1,'tvm::topi']]],
   ['matrix_5fset_5fdiag',['matrix_set_diag',['../namespacetvm_1_1topi.html#aead477c6c9d4f4589d22b8acff82040c',1,'tvm::topi']]],
-  ['max',['Max',['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
+  ['max',['max',['../classtvm_1_1arith_1_1IntSet.html#ac215840d3e9fb2817f1e5648e31317c5',1,'tvm::arith::IntSet::max()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#a2c5ea87b1155aa7810e0beb3b69b955b',1,'tvm::support::LinearCongruentialEngine::max()'],['../classtvm_1_1tir_1_1Max.html#a7dff11b4dea01bfc7a03eacd077f0729',1,'tvm::tir::Max::Max()'],['../namespacetvm.html#a0df5ca82d2c566f628ebb2f1e84a3fcb',1,'tvm::max(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
   ['max_5fvalue',['max_value',['../namespacetvm.html#a4f1398024c0af23699447ef910b654b8',1,'tvm']]],
   ['maxconcurrency',['MaxConcurrency',['../namespacetvm_1_1runtime_1_1threading.html#af8c1c389a74e67bcc3680555288219f8',1,'tvm::runtime::threading']]],
   ['maximum',['maximum',['../namespacetvm_1_1topi.html#afd64bc3e27dfc97002d3add5d7ce4174',1,'tvm::topi::maximum(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a5338e9297463bc745027fca67daa2ebb',1,'tvm::topi::maximum(const tvm::te::Tensor &amp;A, const tvm::te::Tensor &amp;B, std::string name=&quot;T_&quot; &quot;maximum&quot;, std::string tag=kBroadcast)'],['../namespacetvm_1_1topi.html#a4076a8d6a2b243c548d741e9f6bcfe69',1,'tvm::topi::maximum(con [...]
@@ -66,7 +66,7 @@ var searchData=
   ['microtvmruntimegetoutput',['MicroTVMRuntimeGetOutput',['../microtvm__runtime_8h.html#a76129be7b6de972791a3f9a1b312acfa',1,'microtvm_runtime.h']]],
   ['microtvmruntimerun',['MicroTVMRuntimeRun',['../microtvm__runtime_8h.html#ac43a544f675dd716e8c279c3e41f6e45',1,'microtvm_runtime.h']]],
   ['microtvmruntimesetinput',['MicroTVMRuntimeSetInput',['../microtvm__runtime_8h.html#aa593edc600f4356f2b560702aa01b113',1,'microtvm_runtime.h']]],
-  ['min',['Min',['../classtvm_1_1tir_1_1Min.html#a3a4403aec40029a5206e22cd334e356b',1,'tvm::tir::Min::Min()'],['../classtvm_1_1arith_1_1IntSet.html#ae5517de2862e93a801224eed98a57001',1,'tvm::arith::IntSet::min()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#aec5f11b588fa3a12294a46c945c34411',1,'tvm::support::LinearCongruentialEngine::min()'],['../namespacetvm.html#aac2abc149c1a47944c37b560181b15c0',1,'tvm::min(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
+  ['min',['min',['../classtvm_1_1arith_1_1IntSet.html#ae5517de2862e93a801224eed98a57001',1,'tvm::arith::IntSet::min()'],['../classtvm_1_1support_1_1LinearCongruentialEngine.html#aec5f11b588fa3a12294a46c945c34411',1,'tvm::support::LinearCongruentialEngine::min()'],['../classtvm_1_1tir_1_1Min.html#a3a4403aec40029a5206e22cd334e356b',1,'tvm::tir::Min::Min()'],['../namespacetvm.html#aac2abc149c1a47944c37b560181b15c0',1,'tvm::min(PrimExpr a, PrimExpr b, Span span=Span())'],['../namespacetvm.ht [...]
   ['min_5fvalue',['min_value',['../namespacetvm.html#a3b37fa55ea93d6868751a2441996b072',1,'tvm']]],
   ['minimum',['minimum',['../namespacetvm_1_1topi.html#a7ac1dc0d99ce93090a4cdf90ab19d4b8',1,'tvm::topi::minimum(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a0e19dc06a2b1ecbb83b0942fdf836169',1,'tvm::topi::minimum(const tvm::te::Tensor &amp;A, const tvm::te::Tensor &amp;B, std::string name=&quot;T_&quot; &quot;minimum&quot;, std::string tag=kBroadcast)'],['../namespacetvm_1_1topi.html#a28d4ef4b3426bff237215ce356dd5681',1,'tvm::topi::minimum(con [...]
   ['minop',['MinOp',['../namespacetvm_1_1topi.html#aea9a989b0aaa2aef03fe8ee237d8257e',1,'tvm::topi']]],
@@ -79,7 +79,7 @@ var searchData=
   ['mixedmodulepassmanager',['MixedModulePassManager',['../namespacetvm.html#abc01352eff102d4902632d097adc0e08',1,'tvm']]],
   ['mma_5ffill',['mma_fill',['../namespacetvm_1_1tir_1_1builtin.html#a307667c449c54cef747d781771f79bab',1,'tvm::tir::builtin']]],
   ['mma_5fstore',['mma_store',['../namespacetvm_1_1tir_1_1builtin.html#a772fb68f083e71e635c50bb503903f22',1,'tvm::tir::builtin']]],
-  ['mod',['mod',['../classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12',1,'tvm::tir::ScheduleNode::mod()'],['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../namespacetvm_1_1topi.html#aaa95d3ad68932ab206efbe0a326db6a2',1,'tvm::topi::mod(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a4eb4b5a58cf4c5dbbdd4413cfd166882',1,'tvm::topi::mod(const tvm::te::Tensor &amp;A, const tvm::te: [...]
+  ['mod',['Mod',['../classtvm_1_1tir_1_1Mod.html#a8bb56b57ed569d8f357c4439fd8a2f13',1,'tvm::tir::Mod::Mod()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12',1,'tvm::tir::ScheduleNode::mod()'],['../namespacetvm_1_1topi.html#aaa95d3ad68932ab206efbe0a326db6a2',1,'tvm::topi::mod(const tvm::PrimExpr &amp;a, const tvm::PrimExpr &amp;b)'],['../namespacetvm_1_1topi.html#a4eb4b5a58cf4c5dbbdd4413cfd166882',1,'tvm::topi::mod(const tvm::te::Tensor &amp;A, const tvm::te: [...]
   ['mod_5fname',['mod_name',['../classtvm_1_1runtime_1_1metadata_1_1MetadataNode.html#a1c05bb5eb88b5d55b3abeeb2de263191',1,'tvm::runtime::metadata::MetadataNode']]],
   ['modularset',['ModularSet',['../classtvm_1_1arith_1_1ModularSet.html#a9f54896d98169246c6a24cc338fde500',1,'tvm::arith::ModularSet']]],
   ['module',['Module',['../classtvm_1_1runtime_1_1Module.html#abfbc619b3b3166d63ec52e399c24bed9',1,'tvm::runtime::Module::Module()'],['../classtvm_1_1runtime_1_1Module.html#abd1380b3f813c2b6acefca3aaef425f4',1,'tvm::runtime::Module::Module(ObjectPtr&lt; Object &gt; n)']]],
diff --git a/docs/reference/api/doxygen/search/functions_e.js b/docs/reference/api/doxygen/search/functions_e.js
index 410daa8eb5..8314e40f47 100644
--- a/docs/reference/api/doxygen/search/functions_e.js
+++ b/docs/reference/api/doxygen/search/functions_e.js
@@ -1,6 +1,6 @@
 var searchData=
 [
-  ['name',['name',['../classtvm_1_1runtime_1_1metadata_1_1TensorInfoNode.html#a2b889fc061097f72dcab73f090574c65',1,'tvm::runtime::metadata::TensorInfoNode::name()'],['../classtvm_1_1tir_1_1LayoutAxis.html#a0e1b9f9e05162787ccc40a7daa54ac2e',1,'tvm::tir::LayoutAxis::name()'],['../classtvm_1_1tir_1_1Layout.html#a8beb5a8b6259d9c0e916185e9e5eee52',1,'tvm::tir::Layout::name()'],['../classtvm_1_1script_1_1ir__builder_1_1IRBuilder.html#ace475c7a85ef508d912beba48ef4183a',1,'tvm::script::ir_builde [...]
+  ['name',['Name',['../classtvm_1_1script_1_1ir__builder_1_1IRBuilder.html#ace475c7a85ef508d912beba48ef4183a',1,'tvm::script::ir_builder::IRBuilder::Name()'],['../classtvm_1_1script_1_1ir__builder_1_1details_1_1Namer.html#a22da1264beaa8681bab998a5c597369f',1,'tvm::script::ir_builder::details::Namer::Name()'],['../classtvm_1_1runtime_1_1metadata_1_1TensorInfoNode.html#a2b889fc061097f72dcab73f090574c65',1,'tvm::runtime::metadata::TensorInfoNode::name()'],['../classtvm_1_1tir_1_1LayoutAxis. [...]
   ['name_5fhint',['name_hint',['../classtvm_1_1relay_1_1VarPatternNode.html#aa2e698d6bb7d29f6db7aa07070029d42',1,'tvm::relay::VarPatternNode::name_hint()'],['../classtvm_1_1relay_1_1VarNode.html#a8e10dd556c96138aa5320df2fe947199',1,'tvm::relay::VarNode::name_hint()'],['../classtvm_1_1runtime_1_1metadata_1_1ConstantInfoMetadataNode.html#aae9679a8d5d63547239addeb13ba9c17',1,'tvm::runtime::metadata::ConstantInfoMetadataNode::name_hint()']]],
   ['namesupply',['NameSupply',['../classtvm_1_1NameSupply.html#ad6c6c3d3a4698ee50ab2fd2062566820',1,'tvm::NameSupply']]],
   ['namesupplynode',['NameSupplyNode',['../classtvm_1_1NameSupplyNode.html#aba5bb3284648f9f8740eb76500499a12',1,'tvm::NameSupplyNode::NameSupplyNode()=default'],['../classtvm_1_1NameSupplyNode.html#a06784e30bad359af91cb7cd65ec79ca7',1,'tvm::NameSupplyNode::NameSupplyNode(const String &amp;prefix, std::unordered_map&lt; std::string, int &gt; name_map)']]],
diff --git a/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html b/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
index 4b6049d610..ef33367ef8 100644
--- a/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
+++ b/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
@@ -66,7 +66,7 @@ $(function() {
 <div class="title">schedule.h</div>  </div>
 </div><!--header-->
 <div class="contents">
-<a href="tir_2schedule_2schedule_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 [...]
+<a href="tir_2schedule_2schedule_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 [...]
 <div class="ttc" id="namespacetvm_1_1script_1_1ir__builder_1_1tir_html_acd41556b0c4088d0f309ef5495aaebe3"><div class="ttname"><a href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3">tvm::script::ir_builder::tir::Unroll</a></div><div class="ttdeci">ForFrame Unroll(PrimExpr start, PrimExpr stop, Optional&lt; Map&lt; String, ObjectRef &gt;&gt; annotations=NullOpt)</div><div class="ttdoc">The unrolled For statement. </div></div>
 <div class="ttc" id="classtvm_1_1tir_1_1StmtNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1StmtNode.html">tvm::tir::StmtNode</a></div><div class="ttdoc">Base node of all statements. </div><div class="ttdef"><b>Definition:</b> stmt.h:38</div></div>
 <div class="ttc" id="classtvm_1_1tir_1_1BlockRVNode_html_af90b398c502892d19ff3bdf6463d32ab"><div class="ttname"><a href="classtvm_1_1tir_1_1BlockRVNode.html#af90b398c502892d19ff3bdf6463d32ab">tvm::tir::BlockRVNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(tvm::AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> schedule.h:53</div></div>
@@ -84,7 +84,7 @@ $(function() {
 <div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
 <div class="ttc" id="namespacetvm_1_1tir_html_a9ae244600a5e56c4adc9faf6d88f931e"><div class="ttname"><a href="namespacetvm_1_1tir.html#a9ae244600a5e56c4adc9faf6d88f931e">tvm::tir::ScheduleErrorRenderLevel</a></div><div class="ttdeci">ScheduleErrorRenderLevel</div><div class="ttdoc">The level of detailed error message rendering. </div><div class="ttdef"><b>Definition:</b> schedule.h:31</div></div>
 <div class="ttc" id="namespacetvm_1_1tir_html_a9ae244600a5e56c4adc9faf6d88f931ead6733547bb237ce06cddf96357f1b66b"><div class="ttname"><a href="namespacetvm_1_1tir.html#a9ae244600a5e56c4adc9faf6d88f931ead6733547bb237ce06cddf96357f1b66b">tvm::tir::ScheduleErrorRenderLevel::kDetail</a></div><div class="ttdoc">Render a detailed error message. </div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="index__map_8h_html"><div class="ttname"><a href="index__map_8h.html">index_map.h</a></div><div class="ttdoc">Defines a remapping of buffer indices. </div></div>
 <div class="ttc" id="classtvm_1_1support_1_1LinearCongruentialEngine_html_a4d3a3a94a3f3d2dfab4b5ccb1a7e97de"><div class="ttname"><a href="classtvm_1_1support_1_1LinearCongruentialEngine.html#a4d3a3a94a3f3d2dfab4b5ccb1a7e97de">tvm::support::LinearCongruentialEngine::TRandState</a></div><div class="ttdeci">int64_t TRandState</div><div class="ttdef"><b>Definition:</b> random_engine.h:46</div></div>
 <div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
diff --git a/docs/reference/api/doxygen/trace_8h_source.html b/docs/reference/api/doxygen/trace_8h_source.html
index 207ea2ce75..fd662869b1 100644
--- a/docs/reference/api/doxygen/trace_8h_source.html
+++ b/docs/reference/api/doxygen/trace_8h_source.html
@@ -76,7 +76,7 @@ $(function() {
 <div class="ttc" id="namespacetvm_1_1tir_html_a75918aeef1136f9d6308556902d5bcae"><div class="ttname"><a href="namespacetvm_1_1tir.html#a75918aeef1136f9d6308556902d5bcae">tvm::tir::FTraceDecisionProvider</a></div><div class="ttdeci">runtime::TypedPackedFunc&lt; ObjectRef(const Instruction &amp;inst, const Array&lt; ObjectRef &gt; &amp;inputs, const Array&lt; ObjectRef &gt; &amp;attrs, const Optional&lt; ObjectRef &gt; &amp;decision)&gt; FTraceDecisionProvider</div><div class="ttdoc">A cal [...]
 <div class="ttc" id="instruction_8h_html"><div class="ttname"><a href="instruction_8h.html">instruction.h</a></div></div>
 <div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:745</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:773</div></div>
 <div class="ttc" id="classtvm_1_1tir_1_1TraceNode_html_ad6c859ed32b1e2ae076355eda37df0a2"><div class="ttname"><a href="classtvm_1_1tir_1_1TraceNode.html#ad6c859ed32b1e2ae076355eda37df0a2">tvm::tir::TraceNode::insts</a></div><div class="ttdeci">Array&lt; Instruction &gt; insts</div><div class="ttdoc">The instructions invoked so far in the program execution. </div><div class="ttdef"><b>Definition:</b> trace.h:61</div></div>
 <div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
 <div class="ttc" id="classtvm_1_1tir_1_1TraceNode_html_a764346045e536fa26b56c9e140de8e7b"><div class="ttname"><a href="classtvm_1_1tir_1_1TraceNode.html#a764346045e536fa26b56c9e140de8e7b">tvm::tir::TraceNode::ApplyToSchedule</a></div><div class="ttdeci">void ApplyToSchedule(Schedule sch, bool remove_postproc, FTraceDecisionProvider decision_provider=nullptr) const</div><div class="ttdoc">Apply the trace to a TensorIR schedule. </div></div>
diff --git a/docs/reference/api/python/auto_scheduler.html b/docs/reference/api/python/auto_scheduler.html
index a98a34cbae..1c72e052ec 100644
--- a/docs/reference/api/python/auto_scheduler.html
+++ b/docs/reference/api/python/auto_scheduler.html
@@ -1621,7 +1621,7 @@ 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>
@@ -1905,7 +1905,7 @@ 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/python/tir.html b/docs/reference/api/python/tir.html
index 1bcc1fd293..2f3e8f0151 100644
--- a/docs/reference/api/python/tir.html
+++ b/docs/reference/api/python/tir.html
@@ -5818,6 +5818,12 @@ preserve the semantics of computation. Some example of schedules:
 <tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_write" title="tvm.tir.Schedule.cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_write</span></code></a>(block, write_buffer_index, ...)</p></td>
 <td><p>Create a block that reads a buffer region into a write cache.</p></td>
 </tr>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_read" title="tvm.tir.Schedule.reindex_cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_read</span></code></a>(block, read_buffer_index, ...)</p></td>
+<td><p>Create a block that reads a buffer region into a read cache using customized indices specified by index map.</p></td>
+</tr>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_write" title="tvm.tir.Schedule.reindex_cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_write</span></code></a>(block, ...)</p></td>
+<td><p>Create a block that reads a buffer region into a write cache using customized indices specified by index map.</p></td>
+</tr>
 <tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_inplace" title="tvm.tir.Schedule.cache_inplace"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_inplace</span></code></a>(block, read_buffer_index, ...)</p></td>
 <td><p>Create blocks that reads &amp; write a buffer region into a cache block.</p></td>
 </tr>
@@ -6660,7 +6666,7 @@ all consumers will use the cache.</p></li>
 
 <dl class="py method">
 <dt class="sig sig-object py" id="tvm.tir.Schedule.cache_write">
-<span class="sig-name descname"><span class="pre">cache_write</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">block</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.BlockRV</span><span class="p"><span class="pre">,</span> </span><a class="reference external" href="https://docs.python.o [...]
+<span class="sig-name descname"><span class="pre">cache_write</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">block</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.BlockRV</span><span class="p"><span class="pre">,</span> </span><a class="reference external" href="https://docs.python.o [...]
 <dd><p>Create a block that reads a buffer region into a write cache. It requires:</p>
 <ol class="arabic simple">
 <li><p>There is only one block who write the buffer in the scope.</p></li>
@@ -6722,6 +6728,144 @@ If not specified, all consumers will read from the original buffer.</p></li>
 </div>
 </dd></dl>
 
+<dl class="py method">
+<dt class="sig sig-object py" id="tvm.tir.Schedule.reindex_cache_read">
+<span class="sig-name descname"><span class="pre">reindex_cache_read</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">block</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.BlockRV</span><span class="p"><span class="pre">,</span> </span><a class="reference external" href="https://docs.p [...]
+<dd><p>Create a block that reads a buffer region into a read cache using customized
+indices specified by index map. The read region of the buffer must be a single point.</p>
+<p>The cache stage block follows the original order of loops and block itervars in the block.
+If a block itervar does not appear in the buffer access region, it and its corresponding
+loop variables will be omitted. User can then use <cite>transform_block_layout</cite> primitive to
+reorder the block itervars and surrounding loops of the cache read/write block.</p>
+<p>Unlike <cite>cache_read</cite>, <cite>reindex_cache_read</cite> only supports single consumer, please use
+<cite>cache_read</cite> when there are multiple consumers.</p>
+<dl class="field-list simple">
+<dt class="field-odd">Parameters</dt>
+<dd class="field-odd"><ul class="simple">
+<li><p><strong>block</strong> (<em>BlockRV</em>) – The consumer block of the target buffer.</p></li>
+<li><p><strong>read_buffer_index</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#int" title="(in Python v3.11)"><em>int</em></a>) – The index of the buffer in block’s read region.</p></li>
+<li><p><strong>storage_scope</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.11)"><em>str</em></a>) – The target storage scope.</p></li>
+<li><p><strong>index_map</strong> (<em>Union</em><em>[</em><a class="reference internal" href="#tvm.tir.IndexMap" title="tvm.tir.IndexMap"><em>IndexMap</em></a><em>, </em><em>Callable</em><em>]</em>) – User defined indices to access allocated cache buffer, maps from block iter vars.</p></li>
+</ul>
+</dd>
+<dt class="field-even">Returns</dt>
+<dd class="field-even"><p><strong>cached_block</strong> – The block of the cache stage</p>
+</dd>
+<dt class="field-odd">Return type</dt>
+<dd class="field-odd"><p>BlockRV</p>
+</dd>
+</dl>
+<p class="rubric">Examples</p>
+<p>Before reindex_cache_read, in TensorIR, the IR is:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">before_reindex_cache_read</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
+    <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
+</pre></div>
+</div>
+<p>Create the schedule and reindex_cache_read:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">sch</span> <span class="o">=</span> <span class="n">tir</span><span class="o">.</span><span class="n">Schedule</span><span class="p">(</span><span class="n">before_cache_read</span><span class="p">)</span>
+<span class="n">block_b</span> <span class="o">=</span> <span class="n">sch</span><span class="o">.</span><span class="n">get_block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">)</span>
+<span class="n">sch</span><span class="o">.</span><span class="n">reindex_cache_read</span><span class="p">(</span><span class="n">block_b</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;local&quot;</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">:</span> <span class="p">(</span><span class="n">vj</span><span class="p">,</span> <s [...]
+<span class="nb">print</span><span class="p">(</span><span class="n">sch</span><span class="o">.</span><span class="n">mod</span><span class="p">[</span><span class="s2">&quot;main&quot;</span><span class="p">]</span><span class="o">.</span><span class="n">script</span><span class="p">())</span>
+</pre></div>
+</div>
+<p>After applying reindex_cache_read, the IR becomes:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">after_reindex_cache_read</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
+    <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">A_local</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">alloc_buffer</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="n">scope</span><span class="o">=</span><span class="s2">&quot;local&quot;</span><span class="p">)</span>
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;A_local&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">A_local</span><span class="p">[</span><span class="n">vj</span><span class="p">,</span> <span class="n">vi</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span>
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A_local</span><span class="p">[</span><span class="n">vj</span><span class="p">,</span> <span class="n">vi</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
+</pre></div>
+</div>
+<div class="admonition seealso">
+<p class="admonition-title">See also</p>
+<p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_write" title="tvm.tir.Schedule.reindex_cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_write</span></code></a>, <a class="reference internal" href="#tvm.tir.Schedule.transform_block_layout" title="tvm.tir.Schedule.transform_block_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_block_layout</span></code></a>, <a class=" [...]
+</div>
+</dd></dl>
+
+<dl class="py method">
+<dt class="sig sig-object py" id="tvm.tir.Schedule.reindex_cache_write">
+<span class="sig-name descname"><span class="pre">reindex_cache_write</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">block</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.BlockRV</span><span class="p"><span class="pre">,</span> </span><a class="reference external" href="https://docs. [...]
+<dd><p>Create a block that reads a buffer region into a write cache using customized
+indices specified by index map. The write region of the buffer must be a single point.</p>
+<p>The cache stage block follows the original order of loops and block itervars in the block.
+If a block itervar does not appear in the buffer access region, it and its corresponding
+loop variables will be omitted. User can then use <cite>transform_block_layout</cite> primitive to
+reorder the block itervars and surrounding loops of the cache read/write block.</p>
+<p>Unlike <cite>cache_write</cite>, <cite>reindex_cache_write</cite> only supports single consumer, please use
+<cite>cache_write</cite> when there are multiple consumers.</p>
+<dl class="field-list simple">
+<dt class="field-odd">Parameters</dt>
+<dd class="field-odd"><ul class="simple">
+<li><p><strong>block</strong> (<em>Union</em><em>[</em><em>BlockRV</em><em>, </em><a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.11)"><em>str</em></a><em>]</em>) – The consumer block of the target buffer.</p></li>
+<li><p><strong>write_buffer_index</strong> (<a class="reference external" href="https://docs.python.org/3/library/functions.html#int" title="(in Python v3.11)"><em>int</em></a>) – The index of the buffer in block’s write region.</p></li>
+<li><p><strong>storage_scope</strong> (<a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.11)"><em>str</em></a>) – The target storage scope.</p></li>
+<li><p><strong>index_map</strong> (<em>Union</em><em>[</em><em>Callable</em><em>, </em><a class="reference internal" href="#tvm.tir.IndexMap" title="tvm.tir.IndexMap"><em>IndexMap</em></a><em>]</em>) – User defined indices to access allocated cache buffer, maps from block iter vars.</p></li>
+<li><p><strong>consumer_blocks</strong> (<em>Optional</em><em>[</em><em>List</em><em>[</em><em>Union</em><em>[</em><em>BlockRV</em><em>, </em><a class="reference external" href="https://docs.python.org/3/library/stdtypes.html#str" title="(in Python v3.11)"><em>str</em></a><em>]</em><em>]</em><em>]</em>) – An optional list of consumers that should read directly from the cache.
+If not specified, all consumers will read from the original buffer.</p></li>
+</ul>
+</dd>
+<dt class="field-even">Returns</dt>
+<dd class="field-even"><p><strong>cached_block</strong> – The block of the cache stage</p>
+</dd>
+<dt class="field-odd">Return type</dt>
+<dd class="field-odd"><p>BlockRV</p>
+</dd>
+</dl>
+<p class="rubric">Examples</p>
+<p>Before reindex_cache_write, in TensorIR, the IR is:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">before_reindex_cache_write</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
+    <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
+</pre></div>
+</div>
+<p>Create the schedule and reindex_cache_write:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">sch</span> <span class="o">=</span> <span class="n">tir</span><span class="o">.</span><span class="n">Schedule</span><span class="p">(</span><span class="n">before_cache_write</span><span class="p">)</span>
+<span class="n">block_b</span> <span class="o">=</span> <span class="n">sch</span><span class="o">.</span><span class="n">get_block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">)</span>
+<span class="n">sch</span><span class="o">.</span><span class="n">reindex_cache_write</span><span class="p">(</span><span class="n">block_b</span><span class="p">,</span> <span class="mi">0</span><span class="p">,</span> <span class="s2">&quot;local&quot;</span><span class="p">,</span> <span class="k">lambda</span> <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">:</span> <span class="p">(</span><span class="n">vi</span> <span class="o">//</span> [...]
+<span class="nb">print</span><span class="p">(</span><span class="n">sch</span><span class="o">.</span><span class="n">mod</span><span class="p">[</span><span class="s2">&quot;main&quot;</span><span class="p">]</span><span class="o">.</span><span class="n">script</span><span class="p">())</span>
+</pre></div>
+</div>
+<p>After applying reindex_cache_write, the IR becomes:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">after_cache_write</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">)</span> <span class="o">-&gt;</span> <span class="kc">None</span><span class="p">:</span>
+    <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">64</span><span class="p">,</span> <span class="mi">2</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+    <span class="n">B_local</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">alloc_buffer</span><span class="p">((</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">),</span> <span class="n">scope</span><span class="o">=</span><span class="s2">&quot;local&quot;</span><span class="p">)</span>
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;A_local&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">B_local</span><span class="p">[</span><span class="n">vi</span> <span class="o">%</span> <span class="mi">2</span><span class="p">,</span> <span class="n">vi</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span>  [...]
+    <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+        <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">):</span>
+            <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">&quot;SS&quot;</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+            <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">B_local</span><span class="p">[</span><span class="n">vi</span> <span class="o">%</span> <span class="mi">2</span><span class="p">,</span> <span class="n">vi</span> <span class="o">//</span> <span class="mi">2</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span>
+</pre></div>
+</div>
+<div class="admonition seealso">
+<p class="admonition-title">See also</p>
+<p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_read" title="tvm.tir.Schedule.reindex_cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_read</span></code></a>, <a class="reference internal" href="#tvm.tir.Schedule.transform_block_layout" title="tvm.tir.Schedule.transform_block_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_block_layout</span></code></a>, <a class="ref [...]
+</div>
+</dd></dl>
+
 <dl class="py method">
 <dt class="sig sig-object py" id="tvm.tir.Schedule.cache_inplace">
 <span class="sig-name descname"><span class="pre">cache_inplace</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">block</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">Union</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.BlockRV</span><span class="p"><span class="pre">,</span> </span><a class="reference external" href="https://docs.python [...]
@@ -6905,7 +7049,7 @@ reads/writes of the block.</p>
             <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vj</span><span class="p">,</span> <span class="n">vi</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
 </pre></div>
 </div>
-<p>Create the schedule and do transform_layout:</p>
+<p>Create the schedule and do reindex:</p>
 <div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">sch</span> <span class="o">=</span> <span class="n">tir</span><span class="o">.</span><span class="n">Schedule</span><span class="p">(</span><span class="n">before_reindex</span><span class="p">)</span>
 <span class="n">block</span> <span class="o">=</span> <span class="n">sch</span><span class="o">.</span><span class="n">get_block</span><span class="p">(</span><span class="s2">&quot;B&quot;</span><span class="p">)</span>
 <span class="n">sch</span><span class="o">.</span><span class="n">reindex</span><span class="p">(</span><span class="n">block</span><span class="p">,</span> <span class="p">(</span><span class="s2">&quot;read&quot;</span><span class="p">,</span> <span class="mi">0</span><span class="p">))</span>
diff --git a/docs/reference/api/typedoc/classes/bytestreamreader.html b/docs/reference/api/typedoc/classes/bytestreamreader.html
index bb8a89a8f4..5b63aeeb5f 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/e59d1efc6/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L45">rpc_server.ts:45</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/e59d1efc6/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L45">rpc_server.ts:45</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/e59d1efc6/web/src/rpc_server.ts#L44">rpc_server.ts:44</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L44">rpc_server.ts:44</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/e59d1efc6/web/src/rpc_server.ts#L65">rpc_server.ts:65</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L65">rpc_server.ts:65</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/e59d1efc6/web/src/rpc_server.ts#L51">rpc_server.ts:51</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L51">rpc_server.ts:51</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/e59d1efc6/web/src/rpc_server.ts#L59">rpc_server.ts:59</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/rpc_server.ts#L59">rpc_server.ts:59</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 681baa1557..0b0d6e7178 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/e59d1efc6/web/src/memory.ts#L223">memory.ts:223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L208">memory.ts:208</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L312">memory.ts:312</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L284">memory.ts:284</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L388">memory.ts:388</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L376">memory.ts:376</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L267">memory.ts:267</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L243">memory.ts:243</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L321">memory.ts:321</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L252">memory.ts:252</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L359">memory.ts:359</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L342">memory.ts:342</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L350">memory.ts:350</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L326">memory.ts:326</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L363">memory.ts:363</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L346">memory.ts:346</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L334">memory.ts:334</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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 5abfec7833..9e95ee755f 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/e59d1efc6/web/src/runtime.ts#L359">runtime.ts:359</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L359">runtime.ts:359</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/e59d1efc6/web/src/runtime.ts#L357">runtime.ts:357</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L357">runtime.ts:357</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/e59d1efc6/web/src/runtime.ts#L355">runtime.ts:355</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L355">runtime.ts:355</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/e59d1efc6/web/src/runtime.ts#L359">runtime.ts:359</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L359">runtime.ts:359</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/e59d1efc6/web/src/runtime.ts#L376">runtime.ts:376</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L376">runtime.ts:376</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/e59d1efc6/web/src/runtime.ts#L367">runtime.ts:367</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L367">runtime.ts:367</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 3070f8a4bb..38e6d3fffa 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/e59d1efc6/web/src/runtime.ts#L299">runtime.ts:299</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L299">runtime.ts:299</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/e59d1efc6/web/src/runtime.ts#L297">runtime.ts:297</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L297">runtime.ts:297</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/e59d1efc6/web/src/runtime.ts#L295">runtime.ts:295</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L295">runtime.ts:295</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/e59d1efc6/web/src/runtime.ts#L320">runtime.ts:320</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L320">runtime.ts:320</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/e59d1efc6/web/src/runtime.ts#L327">runtime.ts:327</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L327">runtime.ts:327</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 1ebaab3fa4..5fa7d8141b 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/e59d1efc6/web/src/environment.ts#L86">environment.ts:86</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/environment.ts#L70">environment.ts:70</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/environment.ts#L69">environment.ts:69</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/environment.ts#L78">environment.ts:78</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/environment.ts#L84">environment.ts:84</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/environment.ts#L105">environment.ts:105</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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 db2b9139b4..300b73ba47 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/e59d1efc6/web/src/runtime.ts#L50">runtime.ts:50</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L50">runtime.ts:50</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/e59d1efc6/web/src/runtime.ts#L47">runtime.ts:47</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L47">runtime.ts:47</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/e59d1efc6/web/src/runtime.ts#L46">runtime.ts:46</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L46">runtime.ts:46</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/e59d1efc6/web/src/runtime.ts#L45">runtime.ts:45</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L45">runtime.ts:45</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/e59d1efc6/web/src/runtime.ts#L48">runtime.ts:48</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L48">runtime.ts:48</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/e59d1efc6/web/src/runtime.ts#L77">runtime.ts:77</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L77">runtime.ts:77</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/e59d1efc6/web/src/runtime.ts#L67">runtime.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L67">runtime.ts:67</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/e59d1efc6/web/src/runtime.ts#L85">runtime.ts:85</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L85">runtime.ts:85</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/e59d1efc6/web/src/runtime.ts#L96">runtime.ts:96</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L96">runtime.ts:96</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/e59d1efc6/web/src/runtime.ts#L73">runtime.ts:73</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L73">runtime.ts:73</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/instance.html b/docs/reference/api/typedoc/classes/instance.html
index 07922105af..084bf7915a 100644
--- a/docs/reference/api/typedoc/classes/instance.html
+++ b/docs/reference/api/typedoc/classes/instance.html
@@ -161,7 +161,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L844">runtime.ts:844</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L844">runtime.ts:844</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -224,7 +224,7 @@
 					<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/e59d1efc6/web/src/runtime.ts#L834">runtime.ts:834</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L834">runtime.ts:834</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -234,7 +234,7 @@
 					<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/e59d1efc6/web/src/runtime.ts#L833">runtime.ts:833</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L833">runtime.ts:833</a></li>
 						</ul>
 					</aside>
 				</section>
@@ -251,7 +251,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L973">runtime.ts:973</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L973">runtime.ts:973</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -296,7 +296,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L932">runtime.ts:932</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L932">runtime.ts:932</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -318,7 +318,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L901">runtime.ts:901</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L901">runtime.ts:901</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -381,7 +381,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1215">runtime.ts:1215</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1215">runtime.ts:1215</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -412,7 +412,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1000">runtime.ts:1000</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1000">runtime.ts:1000</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -453,7 +453,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1207">runtime.ts:1207</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1207">runtime.ts:1207</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -491,7 +491,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L922">runtime.ts:922</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L922">runtime.ts:922</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -508,7 +508,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1235">runtime.ts:1235</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1235">runtime.ts:1235</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -552,7 +552,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L943">runtime.ts:943</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L943">runtime.ts:943</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -577,7 +577,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1088">runtime.ts:1088</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1088">runtime.ts:1088</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -609,7 +609,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1363">runtime.ts:1363</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1363">runtime.ts:1363</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -640,7 +640,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1123">runtime.ts:1123</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1123">runtime.ts:1123</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -672,7 +672,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1016">runtime.ts:1016</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1016">runtime.ts:1016</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -695,7 +695,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1281">runtime.ts:1281</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1281">runtime.ts:1281</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -729,7 +729,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L986">runtime.ts:986</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L986">runtime.ts:986</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -769,7 +769,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1341">runtime.ts:1341</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1341">runtime.ts:1341</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -817,7 +817,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1055">runtime.ts:1055</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1055">runtime.ts:1055</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -857,7 +857,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1320">runtime.ts:1320</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1320">runtime.ts:1320</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -900,7 +900,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1197">runtime.ts:1197</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1197">runtime.ts:1197</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -938,7 +938,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1491">runtime.ts:1491</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1491">runtime.ts:1491</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -990,7 +990,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1009">runtime.ts:1009</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1009">runtime.ts:1009</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -1014,7 +1014,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1151">runtime.ts:1151</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1151">runtime.ts:1151</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -1046,7 +1046,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1134">runtime.ts:1134</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1134">runtime.ts:1134</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -1078,7 +1078,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1292">runtime.ts:1292</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1292">runtime.ts:1292</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -1110,7 +1110,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L1223">runtime.ts:1223</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L1223">runtime.ts:1223</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -1141,7 +1141,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L957">runtime.ts:957</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L957">runtime.ts:957</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 06e2182b71..ddfd5adac7 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/e59d1efc6/web/src/memory.ts#L40">memory.ts:40</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L32">memory.ts:32</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L33">memory.ts:33</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L154">memory.ts:154</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L90">memory.ts:90</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L97">memory.ts:97</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L74">memory.ts:74</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L81">memory.ts:81</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L104">memory.ts:104</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L132">memory.ts:132</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L145">memory.ts:145</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L60">memory.ts:60</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L67">memory.ts:67</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L53">memory.ts:53</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L114">memory.ts:114</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L124">memory.ts:124</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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/e59d1efc6/web/src/memory.ts#L175">memory.ts:175</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/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 0b70b41a2f..68b2a9b5e0 100644
--- a/docs/reference/api/typedoc/classes/module.html
+++ b/docs/reference/api/typedoc/classes/module.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/e59d1efc6/web/src/runtime.ts#L614">runtime.ts:614</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L614">runtime.ts:614</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-parameters-title">Parameters</h4>
@@ -169,7 +169,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L626">runtime.ts:626</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L626">runtime.ts:626</a></li>
 								</ul>
 							</aside>
 							<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -186,7 +186,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L653">runtime.ts:653</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L653">runtime.ts:653</a></li>
 								</ul>
 							</aside>
 							<div class="tsd-comment tsd-typography">
@@ -218,7 +218,7 @@
 						<li class="tsd-description">
 							<aside class="tsd-sources">
 								<ul>
-									<li>Defined in <a href="https://github.com/apache/tvm/blob/e59d1efc6/web/src/runtime.ts#L641">runtime.ts:641</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L641">runtime.ts:641</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/e59d1efc6/web/src/runtime.ts#L687">runtime.ts:687</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L687">runtime.ts:687</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 7450ea58ab..540403f5bb 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/e59d1efc6/web/src/runtime.ts#L401">runtime.ts:401</a></li>
+									<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L401">runtime.ts:401</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/e59d1efc6/web/src/runtime.ts#L394">runtime.ts:394</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L394">runtime.ts:394</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/e59d1efc6/web/src/runtime.ts#L390">runtime.ts:390</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L390">runtime.ts:390</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
@@ -188,7 +188,7 @@
 					<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/e59d1efc6/web/src/runtime.ts#L388">runtime.ts:388</a></li>
+							<li>Defined in <a href="https://github.com/apache/tvm/blob/9d732d0fa/web/src/runtime.ts#L388">runtime.ts:388</a></li>
 						</ul>
 					</aside>
 					<div class="tsd-comment tsd-typography">
... 2248 lines suppressed ...