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 2023/01/21 00:10:11 UTC
[tvm-site] branch asf-site updated: deploying docs (apache/tvm@fe372a73096a172ce6e5e8879d2628d84d58cb93)
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 1236dcac0b deploying docs (apache/tvm@fe372a73096a172ce6e5e8879d2628d84d58cb93)
1236dcac0b is described below
commit 1236dcac0b81883ccc3cc6563e05dfb014355a0e
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Sat Jan 21 00:10:05 2023 +0000
deploying docs (apache/tvm@fe372a73096a172ce6e5e8879d2628d84d58cb93)
---
docs/_images/sphx_glr_micro_train_001.png | Bin 293847 -> 337505 bytes
docs/_images/sphx_glr_micro_train_thumb.png | Bin 22348 -> 23675 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 | 4 +-
.../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 | 8 +-
.../how_to/extend_tvm/use_pass_instrument.rst.txt | 16 +-
.../optimize_operators/opt_conv_cuda.rst.txt | 2 +-
.../optimize_operators/opt_conv_tensorcore.rst.txt | 2 +-
.../how_to/optimize_operators/opt_gemm.rst.txt | 16 +-
.../optimize_operators/sg_execution_times.rst.txt | 8 +-
.../sg_execution_times.rst.txt | 14 +-
.../tune_conv2d_layer_cuda.rst.txt | 2358 ++++++++------------
.../tune_network_cuda.rst.txt | 4 +-
.../tune_network_x86.rst.txt | 4 +-
.../tune_sparse_x86.rst.txt | 88 +-
.../tune_with_autotvm/sg_execution_times.rst.txt | 4 +-
.../tune_with_autotvm/tune_conv2d_cuda.rst.txt | 323 +--
.../work_with_microtvm/micro_autotune.rst.txt | 16 +-
.../work_with_microtvm/micro_pytorch.rst.txt | 4 +-
.../how_to/work_with_microtvm/micro_train.rst.txt | 18 +-
.../work_with_microtvm/sg_execution_times.rst.txt | 40 +-
.../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 | 14 +-
.../how_to/work_with_schedules/tensorize.rst.txt | 2 +-
.../tutorials/autotvm/sg_execution_times.rst.txt | 6 +-
.../frontend/deploy_classification.rst.txt | 2 +-
.../tutorials/frontend/deploy_detection.rst.txt | 2 +-
.../tutorials/frontend/sg_execution_times.rst.txt | 6 +-
.../tutorials/optimize/sg_execution_times.rst.txt | 6 +-
.../topic/vta/tutorials/sg_execution_times.rst.txt | 6 +-
.../tutorial/auto_scheduler_matmul_x86.rst.txt | 4 +-
docs/_sources/tutorial/autotvm_matmul_x86.rst.txt | 20 +-
docs/_sources/tutorial/autotvm_relay_x86.rst.txt | 55 +-
.../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 | 49 +-
docs/commit_hash | 2 +-
docs/how_to/compile_models/from_darknet.html | 2 +-
docs/how_to/compile_models/from_keras.html | 2 +-
docs/how_to/compile_models/from_mxnet.html | 2 +-
docs/how_to/compile_models/from_oneflow.html | 11 +-
docs/how_to/compile_models/from_pytorch.html | 9 +-
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 | 4 +-
.../deploy_models/deploy_model_on_android.html | 2 +-
.../deploy_object_detection_pytorch.html | 40 +-
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 | 37 +-
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 | 8 +-
docs/how_to/extend_tvm/use_pass_instrument.html | 16 +-
docs/how_to/optimize_operators/opt_conv_cuda.html | 2 +-
.../optimize_operators/opt_conv_tensorcore.html | 2 +-
docs/how_to/optimize_operators/opt_gemm.html | 16 +-
.../optimize_operators/sg_execution_times.html | 8 +-
.../sg_execution_times.html | 14 +-
.../tune_conv2d_layer_cuda.html | 2358 ++++++++------------
.../tune_with_autoscheduler/tune_network_cuda.html | 4 +-
.../tune_with_autoscheduler/tune_network_x86.html | 4 +-
.../tune_with_autoscheduler/tune_sparse_x86.html | 88 +-
.../tune_with_autotvm/sg_execution_times.html | 4 +-
.../how_to/tune_with_autotvm/tune_conv2d_cuda.html | 323 +--
docs/how_to/work_with_microtvm/micro_autotune.html | 16 +-
docs/how_to/work_with_microtvm/micro_pytorch.html | 4 +-
docs/how_to/work_with_microtvm/micro_train.html | 16 +-
.../work_with_microtvm/sg_execution_times.html | 16 +-
.../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 | 14 +-
docs/how_to/work_with_schedules/tensorize.html | 2 +-
docs/reference/api/python/auto_scheduler.html | 4 +-
.../api/typedoc/classes/bytestreamreader.html | 12 +-
.../api/typedoc/classes/cachedcallstack.html | 34 +-
docs/reference/api/typedoc/classes/dldatatype.html | 12 +-
docs/reference/api/typedoc/classes/dldevice.html | 10 +-
.../reference/api/typedoc/classes/environment.html | 12 +-
docs/reference/api/typedoc/classes/ffilibrary.html | 20 +-
.../api/typedoc/classes/graphexecutor.html | 16 +-
docs/reference/api/typedoc/classes/instance.html | 40 +-
docs/reference/api/typedoc/classes/memory.html | 34 +-
docs/reference/api/typedoc/classes/module.html | 10 +-
docs/reference/api/typedoc/classes/ndarray.html | 22 +-
.../api/typedoc/classes/packedfunccell.html | 6 +-
docs/reference/api/typedoc/classes/rpcserver.html | 14 +-
docs/reference/api/typedoc/classes/scalar.html | 6 +-
.../api/typedoc/classes/webgpucontext.html | 12 +-
docs/reference/api/typedoc/enums/argtypecode.html | 30 +-
.../api/typedoc/enums/aynccallbackcode.html | 4 +-
.../api/typedoc/enums/dldatatypecode.html | 8 +-
.../api/typedoc/enums/rpcserverstate.html | 12 +-
docs/reference/api/typedoc/enums/sizeof.html | 18 +-
docs/reference/api/typedoc/index.html | 112 +-
.../api/typedoc/interfaces/disposable.html | 2 +-
.../api/typedoc/interfaces/functioninfo.html | 6 +-
.../api/typedoc/interfaces/libraryprovider.html | 4 +-
docs/searchindex.js | 2 +-
.../vta/tutorials/autotvm/sg_execution_times.html | 6 +-
.../tutorials/frontend/deploy_classification.html | 2 +-
.../vta/tutorials/frontend/deploy_detection.html | 2 +-
.../vta/tutorials/frontend/sg_execution_times.html | 6 +-
.../vta/tutorials/optimize/sg_execution_times.html | 6 +-
docs/topic/vta/tutorials/sg_execution_times.html | 6 +-
docs/tutorial/auto_scheduler_matmul_x86.html | 4 +-
docs/tutorial/autotvm_matmul_x86.html | 20 +-
docs/tutorial/autotvm_relay_x86.html | 265 +--
docs/tutorial/cross_compilation_and_rpc.html | 2 +-
docs/tutorial/intro_topi.html | 2 +-
docs/tutorial/sg_execution_times.html | 22 +-
docs/tutorial/tensor_expr_get_started.html | 45 +-
129 files changed, 2749 insertions(+), 4412 deletions(-)
diff --git a/docs/_images/sphx_glr_micro_train_001.png b/docs/_images/sphx_glr_micro_train_001.png
index 71e241b4ac..cdece017f1 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 20e0aa607a..3ea3b2a601 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 81d6057e10..ac68efa882 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 13.657 seconds)
+ **Total running time of the script:** ( 1 minutes 13.704 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 e7f0075bee..8a5b476e81 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 900ms/step
+
1/1 [==============================] - ETA: 0s
1/1 [==============================] - 1s 926ms/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 f942130293..6f7d727108 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.zip13fd4e9a-1499-4f6e-8b30-76b6c5e5da10 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+ Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip3f917333-f21e-4db2-976c-eb635d245168 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 206a7a8683..82f51c0483 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]
19%|#9 | 7.99M/41.5M [00:00<00:00, 47.2MB/s]
47%|####6 | 19.4M/41.5M [00:00<00:00, 79.1MB/s]
67%|######7 | 27.9M/41.5M [00:00<00:00, 66.7MB/s]
84%|########4 | 34.9M/41.5M [00:00<00:00, 65.3MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 67.9MB/s]
+
0%| | 0.00/41.5M [00:00<?, ?B/s]
19%|#9 | 7.99M/41.5M [00:00<00:00, 53.2MB/s]
39%|###8 | 16.0M/41.5M [00:00<00:00, 67.2MB/s]
57%|#####7 | 23.7M/41.5M [00:00<00:00, 68.2MB/s]
73%|#######3 | 30.4M/41.5M [00:00<00:00, 61.8MB/s]
88%|########7 | 36.4M/41.5M [00:00<00:00, 61.3MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 65.0MB/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 f1e8188fd6..8819fab5d7 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]
18%|#7 | 8.02M/44.7M [00:00<00:00, 84.1MB/s]
36%|###5 | 16.0M/44.7M [00:00<00:00, 75.3MB/s]
62%|######1 | 27.6M/44.7M [00:00<00:00, 95.0MB/s]
82%|########2 | 36.8M/44.7M [00:00<00:00, 88.1MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 77.9MB/s]
+
0%| | 0.00/44.7M [00:00<?, ?B/s]
27%|##7 | 12.1M/44.7M [00:00<00:00, 127MB/s]
54%|#####4 | 24.2M/44.7M [00:00<00:00, 108MB/s]
78%|#######7 | 34.7M/44.7M [00:00<00:00, 105MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 102MB/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 efce8f6700..28ae700bb8 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 17.037 seconds)
+ **Total running time of the script:** ( 1 minutes 18.916 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 70048349cc..a4ead9c855 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
=================
-**06:04.501** total execution time for **how_to_compile_models** files:
+**06:07.485** total execution time for **how_to_compile_models** files:
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:17.037 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:18.916 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:13.657 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:13.704 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:49.361 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:50.021 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:33.583 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:34.358 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:29.796 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:29.718 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:29.253 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:29.571 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:26.384 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:25.936 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:23.531 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:23.165 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:19.329 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:19.579 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.571 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.516 | 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 ac0f40909c..39a5f38520 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
@@ -727,7 +727,7 @@ 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)
- 3340.0471 3339.4969 3343.5940 3338.1399 1.8071
+ 3290.8978 3289.1562 3299.1580 3288.0557 3.3897
@@ -736,7 +736,7 @@ well as provides information about the model's performance
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 2.299 seconds)
+ **Total running time of the script:** ( 1 minutes 1.871 seconds)
.. _sphx_glr_download_how_to_deploy_models_deploy_model_on_adreno.py:
diff --git a/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt b/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt
index 948f87132f..6fa623a026 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)
- 15.5180 15.5055 15.6018 15.4754 0.0393
+ 15.1210 14.9484 15.9747 14.8685 0.3481
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 f82da32d89..7cc69129c0 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%|4 | 7.99M/170M [00:00<00:02, 59.9MB/s]
9%|9 | 16.0M/170M [00:00<00:02, 70.7MB/s]
14%|#4 | 24.0M/170M [00:00<00:02, 64.7MB/s]
19%|#8 | 32.0M/170M [00:00<00:02, 66.7MB/s]
24%|##3 | 40.0M/170M [00:00<00:01, 69.1MB/s]
28%|##8 | 48.0M/170M [00:00<00:01, 64.8MB/s]
33%|###2 | 56.0M/170M [00:00<00:01, 69.8MB/s]
38%|###7 | 64.0M/170M [00:00<00:01, 73.6MB/s]
42%|####2 | 72.0M/170M [00:01<00:01, 66.8MB/s]
48%|####8 | 82.3M/170M [00:01<00:01, 77.7MB/s]
53%|#####3 | 90.0M/170M [00:01<00:01, 66.2MB/s]
57%|#####7 | 96.8M/170M [00:01<00:01, 67.0MB/s]
61%|######1 | 104M/170M [00:01<00:01, 64.8MB/s]
66%|######5 | 112M/170M [00:01<00:00, 67.2MB/s]
71%|####### | 120M/170M [00:01<00:00, 70.7MB/s]
75%|#######5 | 128M/170M [00:01<00:00, 73.0MB/s]
80%|######## | 136M/170M [00:02<00:00, 64.9MB/s]
88%|########7 | 149M/170M [00:02<00:00, 83.1MB/s]
93%|#########2| 158M/170M [00:02<00:00, 74.1MB/s]
99%|#########8| 168M/170M [00:02<00:00, 82.5MB/s]
100%|##########| 170M/170M [00:02<00:00, 71.6MB/s]
+
0%| | 0.00/170M [00:00<?, ?B/s]
5%|5 | 8.84M/170M [00:00<00:01, 92.7MB/s]
12%|#2 | 20.6M/170M [00:00<00:01, 111MB/s]
18%|#8 | 31.2M/170M [00:00<00:01, 101MB/s]
25%|##5 | 42.5M/170M [00:00<00:01, 107MB/s]
31%|###1 | 52.8M/170M [00:00<00:01, 105MB/s]
37%|###7 | 62.9M/170M [00:00<00:01, 103MB/s]
43%|####2 | 72.8M/170M [00:00<00:01, 101MB/s]
49%|####8 | 82.5M/170M [00:00<00:00, 101MB/s]
54%|#####4 | 92.2M/170M [00:00<00:00, 95.9MB/s]
61%|###### | 103M/170M [00:01<00:00, 102MB/s]
67%|######6 | 113M/170M [00:01<00:00, 102MB/s]
72%|#######2 | 123M/170M [00:01<00:00, 101MB/s]
78%|#######8 | 133M/170M [00:01<00:00, 86.2MB/s]
87%|########7 | 148M/170M [00:01<00:00, 105MB/s]
93%|#########3| 158M/170M [00:01<00:00, 104MB/s]
99%|#########9| 169M/170M [00:01<00:00, 92.0MB/s]
100%|##########| 170M/170M [00:01<00:00, 99.3MB/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 20.745 seconds)
+ **Total running time of the script:** ( 3 minutes 19.889 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 1df78589fe..92aa0d396e 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]
59%|#####8 | 7.99M/13.6M [00:00<00:00, 78.7MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 100MB/s]
+
0%| | 0.00/13.6M [00:00<?, ?B/s]
91%|######### | 12.3M/13.6M [00:00<00:00, 129MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 125MB/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)
- 90.2470 90.1751 91.1447 89.9100 0.2500
+ 88.4134 88.2537 92.0269 88.0521 0.5086
@@ -458,7 +458,7 @@ TODO
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 11.472 seconds)
+ **Total running time of the script:** ( 1 minutes 11.225 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 8f27d3ca57..229e230b1d 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)
- 119.3888 119.3236 126.3505 118.3648 0.8178
+ 118.3385 118.1645 130.6115 117.2539 1.3747
@@ -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 25.849 seconds)
+ **Total running time of the script:** ( 2 minutes 22.767 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 9c46ed9c5d..9af26833b5 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 27.242 seconds)
+ **Total running time of the script:** ( 1 minutes 26.919 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 9a4b8b8163..e767d41568 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%|4 | 5432/132723 [00:00<00:02, 54315.91KB/s]
10%|9 | 12935/132723 [00:00<00:01, 66495.76KB/s]
16%|#5 | 20870/132723 [00:00<00:01, 72361.84KB/s]
21%|##1 | 28456/132723 [00:00<00:01, 73733.94KB/s]
27%|##7 | 35953/132723 [00:00<00:01, 74176.58KB/s]
33%|###2 | 43702/132723 [00:00<00:01, 75225.80KB/s]
39%|###8 | 51406/132723 [00:00<00:01, 75811.69KB/s]
44%|####4 | 59001/132723 [00:00<00:00, 75854.05KB/s]
50%|##### | 66699/132723 [00:00<00:00, 76204.27KB/s]
56%|#####5 | 74320/132723 [00:01<00:00, 76092.06KB/s]
62%|######1 | 81942/132723 [00:01<00:00, 76128.53KB/s]
68%|######7 | 89654/132723 [00:01<00:00, 76424.97KB/s]
73%|#######3 | 97297/132723 [00:01<00:00, 75956.41KB/s]
79%|#######9 | 104929/132723 [00:01<00:00, 76063.66KB/s]
85%|########4 | 112575/132723 [00:01<00:00, 76181.49KB/s]
91%|#########
| 120194/132723 [00:01<00:00, 76109.69KB/s]
96%|#########6| 127847/132723 [00:01<00:00, 76232.26KB/s]
100%|##########| 132723/132723 [00:01<00:00, 75327.12KB/s]
+
0%| | 0/132723 [00:00<?, ?KB/s]
3%|3 | 4205/132723 [00:00<00:03, 42042.02KB/s]
9%|9 | 12397/132723 [00:00<00:01, 65492.72KB/s]
16%|#5 | 20610/132723 [00:00<00:01, 73084.90KB/s]
22%|##1 | 28851/132723 [00:00<00:01, 76764.61KB/s]
28%|##7 | 37019/132723 [00:00<00:01, 78521.79KB/s]
34%|###4 | 45221/132723 [00:00<00:01, 79706.87KB/s]
40%|#### | 53417/132723 [00:00<00:00, 80440.38KB/s]
46%|####6 | 61675/132723 [00:00<00:00, 81112.85KB/s]
53%|#####2 | 69933/132723 [00:00<00:00, 81569.75KB/s]
59%|#####8 | 78160/132723 [00:01<00:00, 81783.34KB/s]
65%|######5 | 86468/132723 [00:01<00:00, 82178.13KB/s]
71%|#######1 | 94742/132723 [00:01<00:00, 82347.15KB/s]
78%|#######7 | 103035/132723 [00:01<00:00, 82518.19KB/s]
84%|########3 | 111310/132723 [00:01<00:00, 82586.67KB/s]
90%|######### | 119570/132723 [00:01<00:00, 82587.98KB/s]
96%|########
#6| 127885/132723 [00:01<00:00, 82755.38KB/s]
100%|##########| 132723/132723 [00:01<00:00, 80010.08KB/s]
@@ -246,7 +246,7 @@ Display result
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 3 minutes 26.172 seconds)
+ **Total running time of the script:** ( 3 minutes 28.286 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 aa5b8eeba8..1642783505 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
=================
-**14:24.707** total execution time for **how_to_deploy_models** files:
+**14:22.549** total execution time for **how_to_deploy_models** files:
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:26.172 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:28.286 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:20.745 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:19.889 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:25.849 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:22.767 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:27.242 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:26.919 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:11.472 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:11.225 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 01:02.299 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 01:01.871 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:38.780 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:39.316 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:26.264 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:26.340 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:25.878 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:25.931 | 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 a49934bef2..98cf07e413 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.zipaafa5250-ff11-4c64-9fe6-402dccc856f2 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+ Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip13e7efad-d6aa-4e86-a528-4a394bd59da2 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 d046190bd9..f7f98b5a8e 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
=================
-**00:51.840** total execution time for **how_to_extend_tvm** files:
+**00:53.524** 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``) | 00:48.105 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:49.713 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.665 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.708 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.064 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.096 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``) | 00:00.007 | 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 406afd3cf5..27edaa84a5 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: 17944us [17944us] (48.40%; 48.40%)
- FoldScaleAxis: 19129us [8us] (51.60%; 51.60%)
- FoldConstant: 19121us [1694us] (51.58%; 99.96%)
- InferType: 17427us [17427us] (47.01%; 91.14%)
+ InferType: 18786us [18786us] (48.49%; 48.49%)
+ FoldScaleAxis: 19960us [9us] (51.51%; 51.51%)
+ FoldConstant: 19951us [1796us] (51.49%; 99.96%)
+ InferType: 18155us [18155us] (46.86%; 91.00%)
@@ -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: 18260us [18260us] (48.89%; 48.89%)
- FoldScaleAxis: 19088us [6us] (51.11%; 51.11%)
- FoldConstant: 19082us [1798us] (51.09%; 99.97%)
- InferType: 17285us [17285us] (46.28%; 90.58%)
+ InferType: 18397us [18397us] (48.12%; 48.12%)
+ FoldScaleAxis: 19835us [8us] (51.88%; 51.88%)
+ FoldConstant: 19827us [1821us] (51.86%; 99.96%)
+ InferType: 18006us [18006us] (47.10%; 90.82%)
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 a82b28d89a..5706469d9c 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: 54.228958 ms
+ Convolution: 33.413215 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 9d1c93bd37..927d52f817 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
@@ -602,7 +602,7 @@ be able to run on our build server
.. code-block:: none
- conv2d with tensor core: 6.580464 ms
+ conv2d with tensor core: 13.337607 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 cf12ef1e8e..cdce7dc49b 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: 0.018212
- Baseline: 3.443270
+ Numpy running time: 0.019248
+ Baseline: 3.213059
@@ -224,7 +224,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
.. code-block:: none
- Opt1: 0.298191
+ Opt1: 0.307405
@@ -312,7 +312,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
.. code-block:: none
- Opt2: 0.336805
+ Opt2: 0.335485
@@ -397,7 +397,7 @@ the access pattern for A matrix is more cache friendly.
.. code-block:: none
- Opt3: 0.114357
+ Opt3: 0.115603
@@ -511,7 +511,7 @@ flattening.
.. code-block:: none
- Opt4: 0.109719
+ Opt4: 0.109914
@@ -620,7 +620,7 @@ write to C when all the block results are ready.
.. code-block:: none
- Opt5: 0.111613
+ Opt5: 0.115596
@@ -730,7 +730,7 @@ Furthermore, we can also utilize multi-core processors to do the thread-level pa
.. code-block:: none
- Opt6: 0.147156
+ Opt6: 0.152508
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 dfe4ac51a4..4f76138e45 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
=================
-**00:34.886** total execution time for **how_to_optimize_operators** files:
+**00:34.921** total execution time for **how_to_optimize_operators** files:
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:32.353 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:32.193 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.469 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.586 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.064 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.142 | 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 97400d9589..a632d1dc68 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
=================
-**09:11.432** total execution time for **how_to_tune_with_autoscheduler** files:
+**09:32.550** total execution time for **how_to_tune_with_autoscheduler** files:
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 05:36.422 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 05:45.586 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:36.756 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:39.667 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:04.945 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:07.544 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:27.406 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:31.542 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:13.524 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:14.666 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:12.380 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:13.545 | 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 d8d33ab385..9783b2a303 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
@@ -241,768 +241,481 @@ 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.env_thread("blockIdx.x")
- T.launch_thread(blockIdx_x, 8)
- conv2d_nchw = T.allocate([28], "float32", "local")
- pad_temp_shared = T.allocate([504], "float32", "shared")
- kernel_shared = T.allocate([1536], "float32", "shared")
+ T.launch_thread(blockIdx_x, 28)
+ 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.env_thread("threadIdx.x")
- T.launch_thread(threadIdx_x, 112)
- conv2d_nchw_1 = T.buffer_decl((28,), data=conv2d_nchw, scope="local")
+ T.launch_thread(threadIdx_x, 64)
+ conv2d_nchw_1 = T.buffer_decl((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
- conv2d_nchw_1[7] = T.float32(0)
- conv2d_nchw_1[14] = T.float32(0)
- conv2d_nchw_1[21] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
- conv2d_nchw_1[8] = T.float32(0)
- conv2d_nchw_1[15] = T.float32(0)
- conv2d_nchw_1[22] = T.float32(0)
conv2d_nchw_1[2] = T.float32(0)
- conv2d_nchw_1[9] = T.float32(0)
- conv2d_nchw_1[16] = T.float32(0)
- conv2d_nchw_1[23] = T.float32(0)
conv2d_nchw_1[3] = T.float32(0)
- conv2d_nchw_1[10] = T.float32(0)
- conv2d_nchw_1[17] = T.float32(0)
- conv2d_nchw_1[24] = T.float32(0)
conv2d_nchw_1[4] = T.float32(0)
- conv2d_nchw_1[11] = T.float32(0)
- conv2d_nchw_1[18] = T.float32(0)
- conv2d_nchw_1[25] = T.float32(0)
conv2d_nchw_1[5] = T.float32(0)
- conv2d_nchw_1[12] = T.float32(0)
- conv2d_nchw_1[19] = T.float32(0)
- conv2d_nchw_1[26] = T.float32(0)
conv2d_nchw_1[6] = T.float32(0)
+ 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)
- conv2d_nchw_1[20] = T.float32(0)
- conv2d_nchw_1[27] = T.float32(0)
for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
- cse_var_4: T.int32 = rc_outer_outer * 392
- cse_var_3: T.int32 = ry_outer_outer * 7
cse_var_2: T.int32 = rc_outer_outer * 72
cse_var_1: T.int32 = ry_outer_outer * 3
threadIdx_x_1 = T.env_thread("threadIdx.x")
- pad_temp_shared_1 = T.buffer_decl((504,), data=pad_temp_shared, scope="shared")
- data_1 = T.buffer_decl((25088,), data=data.data)
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1] = T.if_then_else(1 <= threadIdx_x_1 % 63 // 9 + ry_outer_outer and threadIdx_x_1 % 63 // 9 + ry_outer_outer < 8 and 1 <= threadIdx_x_1 % 9 and threadIdx_x_1 % 9 < 8, data_1[cse_var_4 + threadIdx_x_1 // 9 * 7 + cse_var_3 + threadIdx_x_1 % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 112] = T.if_then_else(1 <= (threadIdx_x_1 + 49) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 49) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 4) % 9 and (threadIdx_x_1 + 4) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 112) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 4) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= (threadIdx_x_1 + 35) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 35) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 8) % 9 and (threadIdx_x_1 + 8) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 224) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 8) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= (threadIdx_x_1 + 21) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 21) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 3) % 9 and (threadIdx_x_1 + 3) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 336) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 3) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- if T.likely(threadIdx_x_1 < 56):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= (threadIdx_x_1 + 7) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 7) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 7) % 9 and (threadIdx_x_1 + 7) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 448) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 7) % 9 - 8], T.float32(0))
+ pad_temp_shared_1 = T.buffer_decl((72,), data=pad_temp_shared, scope="shared")
+ with T.launch_thread(threadIdx_x_1, 64):
+ data_1 = T.buffer_decl((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_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.buffer_decl((1536,), data=kernel_shared, scope="shared")
+ kernel_shared_1 = T.buffer_decl((3072,), data=kernel_shared, scope="shared")
kernel_1 = T.buffer_decl((2359296,), data=kernel.data)
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 112] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 112) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 224] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 224) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 560) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 672] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 784) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 193536]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1120] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1120) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1232] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1232) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 258048]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 80):
- kernel_shared_1[threadIdx_x_2 + 1456] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1456) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- for i1_inner, i3_inner in T.grid(4, 7):
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 64] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 128] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 256] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 320] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 512] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 640] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 704] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 832] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1024] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1088] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1216] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1280] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1408] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1472] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1600] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1664] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1792] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1856] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1984] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2048] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2176] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2240] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2368] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2432] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2560] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2624] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2752] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2816] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2944] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 3008] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 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_decl((25088,), data=compute.data)
bias_1 = T.buffer_decl((512,), data=bias.data)
- compute_1[blockIdx_x * 3136 + threadIdx_x // 7 * 196 + i1_inner * 49 + threadIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x * 64 + threadIdx_x // 7 * 4 + i1_inner], 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))
@@ -1052,7 +765,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 0.418 ms
+ Execution time of this operator: 0.351 ms
@@ -1100,20 +813,20 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
conv2d_nchw_nn_o_o_i, conv2d_nchw_nn_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_i, factor=1)
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=4)
- 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=16)
+ 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=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=1)
- conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=7)
+ conv2d_nchw_yy_o_o_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=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=1)
- conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=8)
+ 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=1)
@@ -1122,11 +835,11 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
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=4)
- compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
+ 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=1)
- compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=7)
+ compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=7)
compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=1)
@@ -1149,14 +862,14 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
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=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=112)
+ 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=112)
+ 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:
@@ -1174,742 +887,429 @@ 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__(112) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
- float conv2d_nchw[28];
- __shared__ float pad_temp_shared[504];
- __shared__ float kernel_shared[1536];
+ 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[7] = 0.000000e+00f;
- conv2d_nchw[14] = 0.000000e+00f;
- conv2d_nchw[21] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
- conv2d_nchw[8] = 0.000000e+00f;
- conv2d_nchw[15] = 0.000000e+00f;
- conv2d_nchw[22] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
- conv2d_nchw[9] = 0.000000e+00f;
- conv2d_nchw[16] = 0.000000e+00f;
- conv2d_nchw[23] = 0.000000e+00f;
conv2d_nchw[3] = 0.000000e+00f;
- conv2d_nchw[10] = 0.000000e+00f;
- conv2d_nchw[17] = 0.000000e+00f;
- conv2d_nchw[24] = 0.000000e+00f;
conv2d_nchw[4] = 0.000000e+00f;
- conv2d_nchw[11] = 0.000000e+00f;
- conv2d_nchw[18] = 0.000000e+00f;
- conv2d_nchw[25] = 0.000000e+00f;
conv2d_nchw[5] = 0.000000e+00f;
- conv2d_nchw[12] = 0.000000e+00f;
- conv2d_nchw[19] = 0.000000e+00f;
- conv2d_nchw[26] = 0.000000e+00f;
conv2d_nchw[6] = 0.000000e+00f;
+ conv2d_nchw[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;
- conv2d_nchw[20] = 0.000000e+00f;
- conv2d_nchw[27] = 0.000000e+00f;
for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
__syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 63) / 9) + ry_outer_outer)) && ((((((int)threadIdx.x) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 112) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= ((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 224) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= ((((((int)threadIdx.x) + 21) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 21) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 3) % 9))) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 336) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
- if (((int)threadIdx.x) < 56) {
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (((((int)threadIdx.x) + 7) / 9) + ry_outer_outer)) && ((((((int)threadIdx.x) + 7) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 448) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+ 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);
}
- kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((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) + 112)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 112) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 224) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((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) + 560)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 560) / 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) + 672)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 129024)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 784) / 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)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((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) + 1008)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 193536)];
- kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1120) / 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) + 1232)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1232) / 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) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
- if (((int)threadIdx.x) < 80) {
- kernel_shared[(((int)threadIdx.x) + 1456)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1456) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
+ 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)]));
}
}
- for (int i1_inner = 0; i1_inner < 4; ++i1_inner) {
+ for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[(((((((int)blockIdx.x) * 3136) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + ((((int)threadIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[(((((int)blockIdx.x) * 64) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+ 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);
}
}
}
@@ -1972,7 +1372,7 @@ In the example below we resume the status and do more 5 trials.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 5 minutes 36.422 seconds)
+ **Total running time of the script:** ( 5 minutes 45.586 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 e8844721f5..12a410e1e1 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)
- 7.8718 7.8728 7.8788 7.8637 0.0062
+ 7.8863 7.8847 7.8956 7.8785 0.0071
@@ -675,7 +675,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 4.945 seconds)
+ **Total running time of the script:** ( 1 minutes 7.544 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 3a14dc492a..a9c40b139d 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)
- 748.3044 749.7266 749.7805 745.4060 2.0496
+ 770.1284 770.5184 771.6733 768.1936 1.4471
@@ -694,7 +694,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 36.756 seconds)
+ **Total running time of the script:** ( 1 minutes 39.667 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 b07bfa86ba..438bf77e03 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
@@ -386,12 +386,12 @@ 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(64):
- compute_1 = T.allocate([1024], "float32", "global")
- compute_2 = T.buffer_decl((1024,), data=compute_1)
- for i_outer_inner, nb_j_inner in T.grid(8, 2):
- for i_inner_init in range(4):
- cse_var_1: T.int32 = i_outer_inner * 128 + i_inner_init * 32 + nb_j_inner * 16
+ for i0_outer_i1_outer_fused in T.parallel(16):
+ compute_1 = T.allocate([4096], "float32", "global")
+ compute_2 = T.buffer_decl((4096,), data=compute_1)
+ for i_outer_inner, nb_j_inner in T.grid(2, 2):
+ for i_inner_init in range(64):
+ cse_var_1: T.int32 = i_outer_inner * 2048 + 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)
@@ -408,49 +408,49 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
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(cse_var_2, i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner, placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]), 4):
+ for elem_idx, i_inner in T.grid(T.let(cse_var_2, i0_outer_i1_outer_fused * 2 + nb_j_inner, placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]), 64):
cse_var_2 = T.var("int32")
placeholder_5 = T.buffer_decl((33,), "int32", data=placeholder_3.data)
cse_var_21: T.int32 = elem_idx * 16
- cse_var_20: T.int32 = i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner
- cse_var_19: T.int32 = i_outer_inner * 128 + i_inner * 32 + nb_j_inner * 16
- cse_var_18: T.int32 = i0_outer_i1_outer_fused // 16 * 8192 + i_outer_inner * 1024 + i_inner * 256
- cse_var_17: T.int32 = cse_var_19 + 9
- cse_var_16: T.int32 = cse_var_19 + 8
- cse_var_15: T.int32 = cse_var_19 + 7
- cse_var_14: T.int32 = cse_var_19 + 6
- cse_var_13: T.int32 = cse_var_19 + 5
- cse_var_12: T.int32 = cse_var_19 + 4
- cse_var_11: T.int32 = cse_var_19 + 3
- cse_var_10: T.int32 = cse_var_19 + 2
- cse_var_9: T.int32 = cse_var_19 + 15
- cse_var_8: T.int32 = cse_var_19 + 14
- cse_var_7: T.int32 = cse_var_19 + 13
- cse_var_6: T.int32 = cse_var_19 + 12
- cse_var_5: T.int32 = cse_var_19 + 11
- cse_var_4: T.int32 = cse_var_19 + 10
- cse_var_3: T.int32 = cse_var_19 + 1
+ cse_var_20: T.int32 = i0_outer_i1_outer_fused * 2 + nb_j_inner
+ cse_var_19: T.int32 = i_outer_inner * 16384 + i_inner * 256
+ cse_var_18: T.int32 = i_outer_inner * 2048 + 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_decl((78656,), data=placeholder_1.data)
placeholder_7 = T.buffer_decl((32768,), data=placeholder.data)
placeholder_8 = T.buffer_decl((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_19] = compute_2[cse_var_19] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(32, 32):
- cse_var_22: T.int32 = i0_outer_i1_outer_fused // 16 * 16384 + i0_inner * 512 + i0_outer_i1_outer_fused % 16 * 32 + i1_inner
+ 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, i1_inner in T.grid(128, 32):
+ cse_var_22: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 32 + i1_inner
compute_3 = T.buffer_decl((65536,), data=compute.data)
placeholder_5 = T.buffer_decl((65536,), data=placeholder_4.data)
compute_3[cse_var_22] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_22], T.float32(0))
@@ -503,7 +503,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 2.162 ms
+ Execution time of this operator: 1.871 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 6a143afa49..b77c8ac231 100644
--- a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
Computation times
=================
-**00:37.419** total execution time for **how_to_tune_with_autotvm** files:
+**00:43.516** 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``) | 00:37.387 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``) | 00:43.483 | 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 |
+--------------------------------------------------------------------------------------------------+-----------+--------+
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 3c95ad301c..28f0a479a6 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
@@ -390,7 +390,7 @@ 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, 8, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6120432
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 32, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1521344
No: 2 GFLOPS: 0.00/0.00 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)
@@ -513,7 +513,7 @@ 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, 16, 4, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1615316
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3305364
No: 3 GFLOPS: 0.00/0.00 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)
@@ -636,9 +636,8 @@ 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, 8, 1, 64]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3006723
- No: 4 GFLOPS: 5.81/5.81 result: MeasureResult(costs=(0.03987096,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.805912494659424, timestamp=1674235925.9045806) [('tile_f', [-1, 1, 4, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,462395
- No: 5 GFLOPS: 0.00/5.81 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, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9665629
+ No: 4 GFLOPS: 0.00/0.00 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
@@ -760,10 +759,11 @@ 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, 8, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1128340
- No: 6 GFLOPS: 140.68/140.68 result: MeasureResult(costs=(0.001645544046875,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.708493709564209, timestamp=1674235929.6998758) [('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7652764
- No: 7 GFLOPS: 19.51/140.68 result: MeasureResult(costs=(0.011865912444444444,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2616629600524902, timestamp=1674235930.496339) [('tile_f', [-1, 2, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3175404
- No: 8 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 1, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4624679
+ No: 5 GFLOPS: 33.86/33.86 result: MeasureResult(costs=(0.0068367294,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.5034289360046387, timestamp=1674258295.3208296) [('tile_f', [-1, 1, 1, 16]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3721464
+ No: 6 GFLOPS: 5.70/33.86 result: MeasureResult(costs=(0.0406320075,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.8984193801879883, timestamp=1674258297.2149966) [('tile_f', [-1, 16, 16, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,196769
+ No: 7 GFLOPS: 863.54/863.54 result: MeasureResult(costs=(0.0002680832871794872,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4097778797149658, timestamp=1674258297.9234798) [('tile_f', [-1, 2, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2561305
+ No: 8 GFLOPS: 0.00/863.54 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
@@ -885,8 +885,8 @@ 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, 8, 16, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9845037
- No: 9 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 4, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9481192
+ No: 9 GFLOPS: 0.00/863.54 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
@@ -1008,8 +1008,8 @@ 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, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10214490
- No: 10 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7869550
+ No: 10 GFLOPS: 0.00/863.54 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
@@ -1131,8 +1131,8 @@ 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, 1, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2817901
- No: 11 GFLOPS: 0.00/140.68 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, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6129419
+ No: 11 GFLOPS: 0.00/863.54 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
@@ -1254,8 +1254,10 @@ 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, 2, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,5064326
- No: 12 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8404785
+ No: 12 GFLOPS: 15.10/863.54 result: MeasureResult(costs=(0.015329305,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.648505449295044, timestamp=1674258301.7926817) [('tile_f', [-1, 1, 8, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8281927
+ No: 13 GFLOPS: 8.30/863.54 result: MeasureResult(costs=(0.027881015749999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=7.8216047286987305, timestamp=1674258312.876017) [('tile_f', [-1, 16, 4, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4187776
+ No: 14 GFLOPS: 0.00/863.54 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
@@ -1377,8 +1379,8 @@ 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, 64, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1655443
- No: 13 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 8, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 512, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3519299
+ No: 15 GFLOPS: 0.00/863.54 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
@@ -1500,8 +1502,26 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 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, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8054940
- No: 14 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 16]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9843634
+ No: 16 GFLOPS: 0.00/863.54 result: Traceback (most recent call last):
+ File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 142, in build
+ res = future.result()
+ File "/usr/lib/python3.7/concurrent/futures/_base.py", line 435, in result
+ return self.__get_result()
+ File "/usr/lib/python3.7/concurrent/futures/_base.py", line 384, in __get_result
+ raise self._exception
+ File "/usr/lib/python3.7/concurrent/futures/thread.py", line 57, in run
+ result = self.fn(*self.args, **self.kwargs)
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 432, in <lambda>
+ worker = lambda *args: self._worker_run(*args)
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 401, in _worker_run
+ return proc.recv()
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 309, in recv
+ raise TimeoutError()
+ TimeoutError
+
+ [('tile_f', [-1, 16, 1, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9981239
+ No: 17 GFLOPS: 0.00/863.54 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
@@ -1623,8 +1643,10 @@ 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, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3081024
- No: 15 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 8, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3405315
+ No: 18 GFLOPS: 5.67/863.54 result: MeasureResult(costs=(0.04086507025,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.664371490478516, timestamp=1674258317.8645864) [('tile_f', [-1, 8, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2210407
+ No: 19 GFLOPS: 262.79/863.54 result: MeasureResult(costs=(0.0008809437967032968,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7555351257324219, timestamp=1674258318.8751545) [('tile_f', [-1, 1, 32, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2139540
+ No: 20 GFLOPS: 0.00/863.54 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
@@ -1746,256 +1768,7 @@ 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, 4, 32, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7239012
- No: 16 GFLOPS: 0.00/140.68 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_host=task.target_host, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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, 32, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9854529
- No: 17 GFLOPS: 6.77/140.68 result: MeasureResult(costs=(0.034183689,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.969929218292236, timestamp=1674235940.0712135) [('tile_f', [-1, 32, 1, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9364580
- No: 18 GFLOPS: 1.88/140.68 result: MeasureResult(costs=(0.1230732345,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.94360613822937, timestamp=1674235942.1101472) [('tile_f', [-1, 1, 1, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1990525
- No: 19 GFLOPS: 0.00/140.68 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_host=task.target_host, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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, 512, 1, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1219909
- No: 20 GFLOPS: 25.41/140.68 result: MeasureResult(costs=(0.009110510363636365,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.347748041152954, timestamp=1674235942.8749223) [('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2381511
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6594745
@@ -2050,9 +1823,9 @@ and measure running time.
Finish loading 20 records
Best config:
- [('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7652764
+ [('tile_f', [-1, 2, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2561305
Finish loading 20 records
- Time cost of this operator: 0.001894
+ Time cost of this operator: 0.000642
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 d3eb88daa3..fceca5b57c 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
@@ -363,10 +363,10 @@ 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 312.9 98.734 (1, 2, 10, 10, 3) 2 1 [312.9]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.021 0.953 (1, 6, 10, 10) 1 1 [3.021]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.993 0.313 (1, 1, 10, 10, 3) 1 1 [0.993]
- Total_time - 316.913 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 312.5 98.716 (1, 2, 10, 10, 3) 2 1 [312.5]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.041 0.961 (1, 6, 10, 10) 1 1 [3.041]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 1.023 0.323 (1, 1, 10, 10, 3) 1 1 [1.023]
+ Total_time - 316.563 - - - - -
@@ -431,10 +431,10 @@ 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 174.0 97.875 (1, 1, 10, 10, 6) 2 1 [174.0]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 2.932 1.649 (1, 6, 10, 10) 1 1 [2.932]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.847 0.476 (1, 3, 10, 10, 1) 1 1 [0.847]
- Total_time - 177.779 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 103.3 97.478 (1, 6, 10, 10, 1) 2 1 [103.3]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.805 1.704 (1, 6, 10, 10) 1 1 [1.805]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.867 0.818 (1, 3, 10, 10, 1) 1 1 [0.867]
+ Total_time - 105.973 - - - - -
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 b48bc2964d..d77e9a72f5 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
@@ -117,7 +117,7 @@ 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, 65.9MB/s]
+
0%| | 0.00/3.42M [00:00<?, ?B/s]
100%|##########| 3.42M/3.42M [00:00<00:00, 53.1MB/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.
@@ -322,7 +322,7 @@ Look up prediction top 1 index in 1000 class synset.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 7.743 seconds)
+ **Total running time of the script:** ( 1 minutes 14.000 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 66e2fcca56..56d62ff214 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/tmplu1ennu6/images/random'
+ '/tmp/tmpld9y9_kq/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], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0]
+ :alt: [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]
: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/tmplu1ennu6/images/target contains 8144 images
- /tmp/tmplu1ennu6/images/random contains 5000 images
+ /tmp/tmpld9y9_kq/images/target contains 8144 images
+ /tmp/tmpld9y9_kq/images/random contains 5000 images
@@ -494,13 +494,13 @@ the time on our validation set).
.. code-block:: none
Epoch 1/3
- 328/328 - 46s - loss: 0.2033 - accuracy: 0.9324 - val_loss: 0.1294 - val_accuracy: 0.9619 - 46s/epoch - 141ms/step
+ 328/328 - 49s - loss: 0.2060 - accuracy: 0.9244 - val_loss: 0.1419 - val_accuracy: 0.9445 - 49s/epoch - 148ms/step
Epoch 2/3
- 328/328 - 43s - loss: 0.0927 - accuracy: 0.9649 - val_loss: 0.1044 - val_accuracy: 0.9634 - 43s/epoch - 131ms/step
+ 328/328 - 45s - loss: 0.0996 - accuracy: 0.9644 - val_loss: 0.1011 - val_accuracy: 0.9626 - 45s/epoch - 137ms/step
Epoch 3/3
- 328/328 - 43s - loss: 0.0637 - accuracy: 0.9760 - val_loss: 0.1855 - val_accuracy: 0.9471 - 43s/epoch - 130ms/step
+ 328/328 - 44s - loss: 0.0699 - accuracy: 0.9732 - val_loss: 0.0904 - val_accuracy: 0.9683 - 44s/epoch - 136ms/step
- <keras.callbacks.History object at 0x7fdf35d99790>
+ <keras.callbacks.History object at 0x7f450fb7dd90>
@@ -857,7 +857,7 @@ 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 15.356 seconds)
+ **Total running time of the script:** ( 4 minutes 24.153 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 aeb07cd1c7..509e16cf49 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
@@ -5,24 +5,24 @@
Computation times
=================
-**06:25.642** total execution time for **how_to_work_with_microtvm** files:
+**06:48.403** total execution time for **how_to_work_with_microtvm** files:
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:15.356 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:07.743 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 00:50.179 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:08.657 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:03.706 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``) | 00:00.000 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``) | 00:00.000 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``) | 00:00.000 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_mlperftiny.py` (``micro_mlperftiny.py``) | 00:00.000 | 0.0 MB |
-+---------------------------------------------------------------------------------------------+-----------+--------+
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:24.153 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:13.1000 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 00:56.625 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:09.434 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:04.192 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``) | 00:00.000 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``) | 00:00.000 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``) | 00:00.000 | 0.0 MB |
++---------------------------------------------------------------------------------------------+------------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_mlperftiny.py` (``micro_mlperftiny.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 292c65951f..b2a24a6a4e 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
=================
-**00:43.083** total execution time for **how_to_work_with_relay** files:
+**00:44.304** 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``) | 00:31.505 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:32.574 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:10.043 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:10.224 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:01.529 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:01.501 | 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 51c4447df9..83f4830954 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 0x7fdf35a45b90>
+ <function my_cuda_math_rule at 0x7f442e5ca830>
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 15881eb77b..bec102f2e7 100644
--- a/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
@@ -5,20 +5,20 @@
Computation times
=================
-**00:04.512** total execution time for **how_to_work_with_schedules** files:
+**00:06.069** total execution time for **how_to_work_with_schedules** files:
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:02.191 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:03.604 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.054 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.094 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.531 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.580 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``) | 00:00.516 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``) | 00:00.569 | 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_extern_op.py` (``extern_op.py``) | 00:00.118 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.049 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.050 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``) | 00:00.031 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
index 0298510a41..e6a293353a 100644
--- a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
@@ -328,7 +328,7 @@ The importing needs to happen before the tensorized GEMV being executed.
def main(A: T.Buffer((1024, 64), "float32"), B: T.Buffer((512, 64), "float32"), C: T.Buffer((1024, 512), "float32")):
T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
i = T.var("int32")
- T.attr(T.iter_var(i, None, "DataPar", ""), "pragma_import_llvm", "; ModuleID = '/tmp/tmphrpach3t/input0.cc'\nsource_filename = \"/tmp/tmphrpach3t/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca float*, align 8\n %8 = alloca float*, align 8\n %9 = alloca float*, [...]
+ T.attr(T.iter_var(i, None, "DataPar", ""), "pragma_import_llvm", "; ModuleID = '/tmp/tmp5ienotj_/input0.cc'\nsource_filename = \"/tmp/tmp5ienotj_/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca float*, align 8\n %8 = alloca float*, align 8\n %9 = alloca float*, [...]
for i, j_outer in T.grid(1024, 32):
T.call_extern("int32", "gemv_update", T.tvm_access_ptr(T.type_annotation("float32"), C.data, i * 512 + j_outer * 16, 16, 2), T.tvm_access_ptr(T.type_annotation("float32"), A.data, i * 64, 64, 1), T.tvm_access_ptr(T.type_annotation("float32"), B.data, j_outer * 1024, 1024, 1), 16, 64, 64)
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 66f76f28b6..0c5cdf8527 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
=================
-**00:29.527** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:31.945** total execution time for **topic_vta_tutorials_autotvm** files:
+---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:29.521 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.939 | 0.0 MB |
+---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``) | 00:00.006 | 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 2fbb8ff37e..aeb52fbe97 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 31.80s!
+ resnet18_v1 inference graph built in 33.96s!
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 b1289fc5f6..0f34b5e299 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 21.91s!
+ yolov3-tiny inference graph built in 23.00s!
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 4ed1f48cda..dbceca0dee 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
=================
-**01:36.923** total execution time for **topic_vta_tutorials_frontend** files:
+**01:41.049** total execution time for **topic_vta_tutorials_frontend** files:
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:48.557 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:51.014 | 0.0 MB |
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:48.366 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:50.035 | 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 faaa39e0ea..49f4af469b 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
=================
-**00:03.113** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.264** total execution time for **topic_vta_tutorials_optimize** files:
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.686 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.765 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.427 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.499 | 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 e72d7482b9..854edaa8be 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
=================
-**00:00.723** total execution time for **topic_vta_tutorials** files:
+**00:00.910** total execution time for **topic_vta_tutorials** files:
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.382 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.486 | 0.0 MB |
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.341 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.425 | 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 a1482956b5..2b4c65f4e6 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -322,7 +322,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 95.260 ms
+ Execution time of this operator: 93.671 ms
@@ -440,7 +440,7 @@ operations.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 29.862 seconds)
+ **Total running time of the script:** ( 1 minutes 33.752 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 02c838f7b3..34e9e90e51 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: 1.31/1.31 result: MeasureResult(costs=(0.2056417466,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.517169237136841, timestamp=1674234445.4414015) [('tile_y', [-1, 2]), ('tile_x', [-1, 1])],None,1
- No: 2 GFLOPS: 10.87/10.87 result: MeasureResult(costs=(0.0246956128,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6956515312194824, timestamp=1674234446.8534567) [('tile_y', [-1, 4]), ('tile_x', [-1, 128])],None,72
- No: 3 GFLOPS: 10.93/10.93 result: MeasureResult(costs=(0.024570536999999996,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.729093074798584, timestamp=1674234447.499262) [('tile_y', [-1, 1]), ('tile_x', [-1, 64])],None,60
- No: 4 GFLOPS: 2.18/10.93 result: MeasureResult(costs=(0.12336555900000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.182971715927124, timestamp=1674234450.4712882) [('tile_y', [-1, 512]), ('tile_x', [-1, 8])],None,39
- No: 5 GFLOPS: 3.71/10.93 result: MeasureResult(costs=(0.0723802372,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4075696468353271, timestamp=1674234452.0531094) [('tile_y', [-1, 128]), ('tile_x', [-1, 16])],None,47
- No: 6 GFLOPS: 4.14/10.93 result: MeasureResult(costs=(0.06476187800000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2888550758361816, timestamp=1674234453.3373234) [('tile_y', [-1, 16]), ('tile_x', [-1, 16])],None,44
- No: 7 GFLOPS: 14.43/14.43 result: MeasureResult(costs=(0.0185965448,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6065623760223389, timestamp=1674234454.6364162) [('tile_y', [-1, 32]), ('tile_x', [-1, 64])],None,65
- No: 8 GFLOPS: 12.94/14.43 result: MeasureResult(costs=(0.0207464668,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5913591384887695, timestamp=1674234455.2273161) [('tile_y', [-1, 4]), ('tile_x', [-1, 256])],None,82
- No: 9 GFLOPS: 0.51/14.43 result: MeasureResult(costs=(0.5277840008,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.6808443069458, timestamp=1674234464.0203645) [('tile_y', [-1, 32]), ('tile_x', [-1, 1])],None,5
- No: 10 GFLOPS: 1.57/14.43 result: MeasureResult(costs=(0.17134171839999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.9349889755249023, timestamp=1674234467.0087938) [('tile_y', [-1, 8]), ('tile_x', [-1, 1])],None,3
+ No: 1 GFLOPS: 10.12/10.12 result: MeasureResult(costs=(0.026523252200000003,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6626763343811035, timestamp=1674256809.3798146) [('tile_y', [-1, 1]), ('tile_x', [-1, 256])],None,80
+ No: 2 GFLOPS: 1.16/10.12 result: MeasureResult(costs=(0.2318167774,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.937443494796753, timestamp=1674256814.10344) [('tile_y', [-1, 1]), ('tile_x', [-1, 2])],None,10
+ No: 3 GFLOPS: 1.51/10.12 result: MeasureResult(costs=(0.17731797,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.0626556873321533, timestamp=1674256817.1931603) [('tile_y', [-1, 512]), ('tile_x', [-1, 4])],None,29
+ No: 4 GFLOPS: 13.66/13.66 result: MeasureResult(costs=(0.0196530548,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5495791435241699, timestamp=1674256818.5291126) [('tile_y', [-1, 4]), ('tile_x', [-1, 256])],None,82
+ No: 5 GFLOPS: 11.83/13.66 result: MeasureResult(costs=(0.0226819934,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6207194328308105, timestamp=1674256819.2879243) [('tile_y', [-1, 32]), ('tile_x', [-1, 32])],None,55
+ No: 6 GFLOPS: 12.13/13.66 result: MeasureResult(costs=(0.0221262474,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6035187244415283, timestamp=1674256820.6680129) [('tile_y', [-1, 256]), ('tile_x', [-1, 256])],None,88
+ No: 7 GFLOPS: 1.96/13.66 result: MeasureResult(costs=(0.1367607588,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.421938419342041, timestamp=1674256823.1067498) [('tile_y', [-1, 4]), ('tile_x', [-1, 2])],None,12
+ No: 8 GFLOPS: 12.62/13.66 result: MeasureResult(costs=(0.0212703284,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.621758222579956, timestamp=1674256823.706723) [('tile_y', [-1, 16]), ('tile_x', [-1, 256])],None,84
+ No: 9 GFLOPS: 1.62/13.66 result: MeasureResult(costs=(0.1660671058,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.849247694015503, timestamp=1674256826.6714988) [('tile_y', [-1, 4]), ('tile_x', [-1, 1])],None,2
+ No: 10 GFLOPS: 13.28/13.66 result: MeasureResult(costs=(0.020206584,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5243816375732422, timestamp=1674256827.2492478) [('tile_y', [-1, 2]), ('tile_x', [-1, 512])],None,91
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index 3194a8605c..81c9d3386e 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': 515.8801153599779, 'median': 516.1550652997903, 'std': 2.0908542727619}
+ {'mean': 476.6050386599976, 'median': 476.417146350002, 'std': 1.4322166401243506}
@@ -545,29 +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: 14.35/ 14.35 GFLOPS | Progress: (4/20) | 7.68 s
[Task 1/25] Current/Best: 12.49/ 16.24 GFLOPS | Progress: (8/20) | 12.01 s
[Task 1/25] Current/Best: 17.90/ 17.90 GFLOPS | Progress: (12/20) | 16.46 s
[Task 1/25] Current/Best: 5.82/ 17.90 GFLOPS | Progress: (16/20) | 18.94 s
[Task 1/25] Current/Best: 3.20/ 17.90 GFLOPS | Progress: (20/20) | 22.98 s Done.
-
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 5.92/ 19.10 GFLOPS | Progress: (4/20) | 3.39 s
[Task 2/25] Current/Best: 6.26/ 19.10 GFLOPS | Progress: (8/20) | 4.96 s
[Task 2/25] Current/Best: 21.74/ 21.74 GFLOPS | Progress: (12/20) | 6.93 s
[Task 2/25] Current/Best: 12.28/ 21.74 GFLOPS | Progress: (16/20) | 8.38 s
[Task 2/25] Current/Best: 14.66/ 21.74 GFLOPS | Progress: (20/20) | 9.96 s Done.
-
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 3.15/ 17.04 GFLOPS | Progress: (4/20) | 5.13 s
[Task 3/25] Current/Best: 21.16/ 21.16 GFLOPS | Progress: (8/20) | 7.36 s
[Task 3/25] Current/Best: 9.57/ 21.16 GFLOPS | Progress: (12/20) | 9.83 s
[Task 3/25] Current/Best: 21.01/ 21.16 GFLOPS | Progress: (16/20) | 11.58 s
[Task 3/25] Current/Best: 15.20/ 21.16 GFLOPS | Progress: (20/20) | 14.23 s Done.
-
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 10.89/ 13.23 GFLOPS | Progress: (4/20) | 4.93 s
[Task 4/25] Current/Best: 7.49/ 13.41 GFLOPS | Progress: (8/20) | 7.07 s
[Task 4/25] Current/Best: 16.99/ 16.99 GFLOPS | Progress: (12/20) | 9.26 s
[Task 4/25] Current/Best: 11.72/ 16.99 GFLOPS | Progress: (16/20) | 15.20 s
[Task 4/25] Current/Best: 7.92/ 20.58 GFLOPS | Progress: (20/20) | 17.00 s Done.
-
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 5.60/ 14.61 GFLOPS | Progress: (4/20) | 4.45 s
[Task 5/25] Current/Best: 5.55/ 15.86 GFLOPS | Progress: (8/20) | 6.66 s
[Task 5/25] Current/Best: 8.82/ 17.27 GFLOPS | Progress: (12/20) | 8.45 s
[Task 5/25] Current/Best: 5.84/ 19.66 GFLOPS | Progress: (16/20) | 11.03 s
[Task 5/25] Current/Best: 15.49/ 19.66 GFLOPS | Progress: (20/20) | 13.20 s Done.
-
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 8.75/ 13.82 GFLOPS | Progress: (4/20) | 5.52 s
[Task 6/25] Current/Best: 13.96/ 21.72 GFLOPS | Progress: (8/20) | 8.04 s
[Task 6/25] Current/Best: 17.46/ 21.72 GFLOPS | Progress: (12/20) | 11.86 s
[Task 6/25] Current/Best: 8.58/ 21.72 GFLOPS | Progress: (16/20) | 14.90 s
[Task 6/25] Current/Best: 14.18/ 21.72 GFLOPS | Progress: (20/20) | 16.85 s Done.
-
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 6.05/ 12.07 GFLOPS | Progress: (4/20) | 4.59 s
[Task 7/25] Current/Best: 7.41/ 19.32 GFLOPS | Progress: (8/20) | 6.80 s
[Task 7/25] Current/Best: 6.65/ 19.32 GFLOPS | Progress: (12/20) | 9.47 s
[Task 7/25] Current/Best: 18.19/ 19.32 GFLOPS | Progress: (16/20) | 11.98 s
[Task 7/25] Current/Best: 14.89/ 19.32 GFLOPS | Progress: (20/20) | 14.15 s Done.
-
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 4.54/ 4.54 GFLOPS | Progress: (4/20) | 6.18 s
[Task 8/25] Current/Best: 11.81/ 18.85 GFLOPS | Progress: (8/20) | 8.48 s
[Task 8/25] Current/Best: 9.40/ 18.85 GFLOPS | Progress: (12/20) | 13.18 s
[Task 8/25] Current/Best: 10.84/ 18.85 GFLOPS | Progress: (16/20) | 24.59 s
[Task 8/25] Current/Best: 14.47/ 18.85 GFLOPS | Progress: (20/20) | 26.91 s Done.
-
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 14.93/ 16.68 GFLOPS | Progress: (4/20) | 6.02 s
[Task 9/25] Current/Best: 9.20/ 17.98 GFLOPS | Progress: (8/20) | 17.09 s
[Task 9/25] Current/Best: 20.25/ 20.25 GFLOPS | Progress: (12/20) | 21.14 s
[Task 9/25] Current/Best: 11.38/ 21.05 GFLOPS | Progress: (16/20) | 23.77 s
[Task 9/25] Current/Best: 15.50/ 21.05 GFLOPS | Progress: (20/20) | 26.76 s
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 6.77/ 13.04 GFLOPS | Progress: (4/20) | 3.81 s
[Task 10/25] Current/Best: 3.00/ 18.61 GFLOPS | Progress: (8/20) | 5.86 s
[Task 10/25] Current/Best: 17.93/ 18.61 GFLOPS | Progress: (12/20) | 7.49 s
[Task 10/25] Current/Best: 12.91/ 18.61 GFLOPS | Progress: (16/20) | 10.00 s
[Task 10/25] Current/Best: 14.01/ 18.61 GFLOPS | Progress: (20/20
) | 11.82 s Done.
-
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 10.94/ 11.63 GFLOPS | Progress: (4/20) | 4.69 s
[Task 11/25] Current/Best: 19.13/ 19.13 GFLOPS | Progress: (8/20) | 7.26 s
[Task 11/25] Current/Best: 16.38/ 21.48 GFLOPS | Progress: (12/20) | 9.46 s
[Task 11/25] Current/Best: 17.11/ 21.48 GFLOPS | Progress: (16/20) | 11.74 s
[Task 11/25] Current/Best: 12.03/ 21.48 GFLOPS | Progress: (20/20) | 14.46 s Done.
-
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 12.67/ 12.67 GFLOPS | Progress: (4/20) | 7.47 s
[Task 12/25] Current/Best: 6.57/ 15.58 GFLOPS | Progress: (8/20) | 10.17 s
[Task 12/25] Current/Best: 5.27/ 15.58 GFLOPS | Progress: (12/20) | 13.82 s
[Task 12/25] Current/Best: 16.36/ 16.36 GFLOPS | Progress: (16/20) | 16.23 s
[Task 12/25] Current/Best: 14.61/ 16.36 GFLOPS | Progress: (20/20) | 18.30 s Done.
-
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 8.41/ 18.02 GFLOPS | Progress: (4/20) | 5.05 s
[Task 13/25] Current/Best: 14.95/ 18.02 GFLOPS | Progress: (8/20) | 7.13 s
[Task 13/25] Current/Best: 6.21/ 22.11 GFLOPS | Progress: (12/20) | 10.24 s
[Task 13/25] Current/Best: 3.09/ 22.11 GFLOPS | Progress: (16/20) | 13.88 s
[Task 13/25] Current/Best: 11.62/ 22.11 GFLOPS | Progress: (20/20) | 18.40 s Done.
-
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 8.33/ 14.82 GFLOPS | Progress: (4/20) | 4.10 s
[Task 14/25] Current/Best: 4.95/ 15.12 GFLOPS | Progress: (8/20) | 7.88 s
[Task 14/25] Current/Best: 15.88/ 21.61 GFLOPS | Progress: (12/20) | 10.42 s
[Task 14/25] Current/Best: 16.14/ 21.61 GFLOPS | Progress: (16/20) | 14.25 s
[Task 14/25] Current/Best: 10.03/ 21.61 GFLOPS | Progress: (20/20) | 16.97 s Done.
-
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 16.51/ 21.07 GFLOPS | Progress: (4/20) | 5.32 s
[Task 15/25] Current/Best: 17.74/ 21.07 GFLOPS | Progress: (8/20) | 7.33 s
[Task 15/25] Current/Best: 8.57/ 21.07 GFLOPS | Progress: (12/20) | 10.61 s
[Task 15/25] Current/Best: 19.07/ 21.07 GFLOPS | Progress: (16/20) | 12.34 s
[Task 15/25] Current/Best: 16.39/ 21.08 GFLOPS | Progress: (20/20) | 15.55 s Done.
-
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 5.84/ 18.44 GFLOPS | Progress: (4/20) | 3.58 s
[Task 16/25] Current/Best: 5.97/ 18.44 GFLOPS | Progress: (8/20) | 5.49 s
[Task 16/25] Current/Best: 13.63/ 19.06 GFLOPS | Progress: (12/20) | 7.79 s
[Task 16/25] Current/Best: 6.89/ 19.06 GFLOPS | Progress: (16/20) | 10.11 s
[Task 16/25] Current/Best: 17.70/ 19.06 GFLOPS | Progress: (20/20) | 11.87 s Done.
-
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 18.23/ 18.23 GFLOPS | Progress: (4/20) | 4.14 s
[Task 17/25] Current/Best: 3.08/ 18.23 GFLOPS | Progress: (8/20) | 7.25 s
[Task 17/25] Current/Best: 19.23/ 21.55 GFLOPS | Progress: (12/20) | 9.49 s
[Task 17/25] Current/Best: 13.35/ 21.55 GFLOPS | Progress: (16/20) | 12.45 s
[Task 17/25] Current/Best: 16.97/ 21.55 GFLOPS | Progress: (20/20) | 16.16 s Done.
-
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 14.84/ 17.55 GFLOPS | Progress: (4/20) | 3.98 s
[Task 18/25] Current/Best: 9.26/ 17.55 GFLOPS | Progress: (8/20) | 8.19 s
[Task 18/25] Current/Best: 16.42/ 17.55 GFLOPS | Progress: (12/20) | 10.65 s
[Task 18/25] Current/Best: 19.23/ 19.23 GFLOPS | Progress: (16/20) | 13.84 s
[Task 18/25] Current/Best: 11.87/ 19.23 GFLOPS | Progress: (20/20) | 16.81 s Done.
-
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 13.78/ 21.16 GFLOPS | Progress: (4/20) | 5.86 s
[Task 19/25] Current/Best: 19.14/ 21.16 GFLOPS | Progress: (8/20) | 8.86 s
[Task 19/25] Current/Best: 6.18/ 21.16 GFLOPS | Progress: (12/20) | 12.64 s
[Task 19/25] Current/Best: 19.71/ 21.16 GFLOPS | Progress: (16/20) | 15.97 s
[Task 19/25] Current/Best: 9.56/ 21.16 GFLOPS | Progress: (20/20) | 19.90 s Done.
-
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 9.58/ 17.69 GFLOPS | Progress: (4/20) | 4.08 s
[Task 20/25] Current/Best: 11.77/ 17.69 GFLOPS | Progress: (8/20) | 5.98 s
[Task 20/25] Current/Best: 15.91/ 17.69 GFLOPS | Progress: (12/20) | 8.49 s Done.
-
[Task 20/25] Current/Best: 8.40/ 17.69 GFLOPS | Progress: (16/20) | 11.69 s
[Task 20/25] Current/Best: 17.20/ 17.69 GFLOPS | Progress: (20/20) | 14.98 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 14.58/ 15.92 GFLOPS | Progress: (4/20) | 4.28 s
[Task 21/25] Current/Best: 15.49/ 18.63 GFLOPS | Progress: (8/20) | 6.40 s
[Task 21/25] Current/Best: 5.29/ 18.63 GFLOPS | Progress: (12/20) | 8.57 s
[Task 21/25] Current/Best: 9.01/ 18.63 GFLOPS | Progress: (16/20) | 11.46 s
[Task 21/25] Current/Best: 11.73/ 18.63 GFLOPS | Progress: (20/20) | 13.80 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 18.05/ 18.05 GFLOPS | Progress: (4/20) | 4.41 s Done.
-
[Task 22/25] Current/Best: 9.99/ 18.05 GFLOPS | Progress: (8/20) | 7.97 s
[Task 22/25] Current/Best: 19.84/ 19.84 GFLOPS | Progress: (12/20) | 9.92 s
[Task 22/25] Current/Best: 7.96/ 19.84 GFLOPS | Progress: (16/20) | 13.12 s
[Task 22/25] Current/Best: 9.64/ 19.84 GFLOPS | Progress: (20/20) | 15.58 s Done.
-
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 6.47/ 10.72 GFLOPS | Progress: (4/20) | 4.69 s
[Task 23/25] Current/Best: 11.06/ 19.93 GFLOPS | Progress: (8/20) | 8.80 s
[Task 23/25] Current/Best: 12.13/ 20.33 GFLOPS | Progress: (12/20) | 11.86 s
[Task 23/25] Current/Best: 11.13/ 20.33 GFLOPS | Progress: (16/20) | 15.78 s
[Task 23/25] Current/Best: 8.27/ 20.33 GFLOPS | Progress: (20/20) | 18.06 s Done.
-
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 7.74/ 7.74 GFLOPS | Progress: (4/20) | 12.43 s
[Task 24/25] Current/Best: 1.55/ 7.74 GFLOPS | Progress: (8/20) | 18.38 s
[Task 24/25] Current/Best: 1.89/ 10.34 GFLOPS | Progress: (12/20) | 19.98 s
[Task 24/25] Current/Best: 3.03/ 10.34 GFLOPS | Progress: (16/20) | 27.47 s
[Task 24/25] Current/Best: 4.61/ 10.34 GFLOPS | Progress: (20/20) | 38.26 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 25/25] Current/Best: 7.45/ 7.45 GFLOPS | Progress: (4/20) | 4.25 s
[Task 25/25] Current/Best: 6.01/ 8.35 GFLOPS | Progress: (8/20) | 6.17 s
[Task 25/25] Current/Best: 2.80/ 8.35 GFLOPS | Progress: (12/20) | 8.75 s
[Task 25/25] Current/Best: 6.60/ 9.12 GFLOPS | Progress: (16/20) | 11.30 s
[Task 25/25] Current/Best: 8.20/ 9.12 GFLOPS | Progress: (20/2
0) | 21.94 s
+
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 1/25] Current/Best: 19.54/ 19.54 GFLOPS | Progress: (4/20) | 8.47 s
[Task 1/25] Current/Best: 7.37/ 24.45 GFLOPS | Progress: (8/20) | 11.59 s
[Task 1/25] Current/Best: 15.12/ 24.45 GFLOPS | Progress: (12/20) | 14.23 s
[Task 1/25] Current/Best: 12.55/ 24.45 GFLOPS | Progress: (16/20) | 16.85 s
[Task 1/25] Current/Best: 11.22/ 24.45 GFLOPS | Progress: (20/20) | 20.90 s Done.
+
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 8.47/ 23.41 GFLOPS | Progress: (4/20) | 3.16 s
[Task 2/25] Current/Best: 16.33/ 23.41 GFLOPS | Progress: (8/20) | 5.22 s
[Task 2/25] Current/Best: 16.55/ 23.41 GFLOPS | Progress: (12/20) | 6.73 s
[Task 2/25] Current/Best: 12.40/ 23.41 GFLOPS | Progress: (16/20) | 8.63 s
[Task 2/25] Current/Best: 13.30/ 23.41 GFLOPS | Progress: (20/20) | 10.24 s Done.
+
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 15.22/ 19.83 GFLOPS | Progress: (4/20) | 3.64 s
[Task 3/25] Current/Best: 12.38/ 21.67 GFLOPS | Progress: (8/20) | 5.53 s
[Task 3/25] Current/Best: 10.32/ 21.67 GFLOPS | Progress: (12/20) | 8.82 s
[Task 3/25] Current/Best: 7.12/ 23.54 GFLOPS | Progress: (16/20) | 11.09 s
[Task 3/25] Current/Best: 23.51/ 23.54 GFLOPS | Progress: (20/20) | 13.13 s Done.
+
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 11.71/ 17.76 GFLOPS | Progress: (4/20) | 3.70 s
[Task 4/25] Current/Best: 20.13/ 20.13 GFLOPS | Progress: (8/20) | 5.35 s
[Task 4/25] Current/Best: 16.84/ 20.13 GFLOPS | Progress: (12/20) | 7.53 s
[Task 4/25] Current/Best: 13.68/ 20.13 GFLOPS | Progress: (16/20) | 12.24 s
[Task 4/25] Current/Best: 12.10/ 20.13 GFLOPS | Progress: (20/20) | 17.46 s Done.
+
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 15.38/ 22.29 GFLOPS | Progress: (4/20) | 3.87 s
[Task 5/25] Current/Best: 6.18/ 22.29 GFLOPS | Progress: (8/20) | 5.93 s
[Task 5/25] Current/Best: 14.67/ 22.99 GFLOPS | Progress: (12/20) | 7.54 s
[Task 5/25] Current/Best: 18.78/ 22.99 GFLOPS | Progress: (16/20) | 9.48 s
[Task 5/25] Current/Best: 13.05/ 22.99 GFLOPS | Progress: (20/20) | 12.09 s Done.
+
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 5.36/ 15.39 GFLOPS | Progress: (4/20) | 5.02 s
[Task 6/25] Current/Best: 5.80/ 16.48 GFLOPS | Progress: (8/20) | 7.50 s
[Task 6/25] Current/Best: 11.31/ 16.48 GFLOPS | Progress: (12/20) | 11.08 s
[Task 6/25] Current/Best: 5.43/ 16.48 GFLOPS | Progress: (16/20) | 13.94 s
[Task 6/25] Current/Best: 12.60/ 16.48 GFLOPS | Progress: (20/20) | 17.44 s Done.
+
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 17.66/ 17.66 GFLOPS | Progress: (4/20) | 4.45 s
[Task 7/25] Current/Best: 19.75/ 19.75 GFLOPS | Progress: (8/20) | 6.66 s
[Task 7/25] Current/Best: 15.95/ 19.75 GFLOPS | Progress: (12/20) | 8.90 s
[Task 7/25] Current/Best: 7.39/ 19.75 GFLOPS | Progress: (16/20) | 11.22 s
[Task 7/25] Current/Best: 18.21/ 19.75 GFLOPS | Progress: (20/20) | 13.89 s Done.
+
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 16.81/ 16.81 GFLOPS | Progress: (4/20) | 7.27 s
[Task 8/25] Current/Best: 10.80/ 17.75 GFLOPS | Progress: (8/20) | 19.24 s
[Task 8/25] Current/Best: 3.61/ 17.75 GFLOPS | Progress: (12/20) | 22.85 s
[Task 8/25] Current/Best: 9.32/ 17.75 GFLOPS | Progress: (16/20) | 25.19 s
[Task 8/25] Current/Best: 2.83/ 17.75 GFLOPS | Progress: (20/20) | 30.46 s Done.
+
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 20.30/ 21.92 GFLOPS | Progress: (4/20) | 4.89 s
[Task 9/25] Current/Best: 21.47/ 21.92 GFLOPS | Progress: (8/20) | 16.00 s
[Task 9/25] Current/Best: 9.01/ 21.92 GFLOPS | Progress: (12/20) | 19.22 s
[Task 9/25] Current/Best: 16.72/ 21.92 GFLOPS | Progress: (16/20) | 21.00 s
[Task 9/25] Current/Best: 16.67/ 21.92 GFLOPS | Progress: (20/20) | 23.00 s
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 15.41/ 15.54 GFLOPS | Progress: (4/20) | 4.20 s
[Task 10/25] Current/Best: 18.40/ 22.58 GFLOPS | Progress: (8/20) | 5.95 s
[Task 10/25] Current/Best: 10.15/ 22.58 GFLOPS | Progress: (12/20) | 7.73 s
[Task 10/25] Current/Best: 11.73/ 22.58 GFLOPS | Progress: (16/20) | 10.23 s
[Task 10/25] Current/Best: 10.08/ 22.58 GFLOPS | Progress: (20/20
) | 12.76 s Done.
+
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 19.58/ 20.30 GFLOPS | Progress: (4/20) | 3.66 s
[Task 11/25] Current/Best: 10.41/ 20.30 GFLOPS | Progress: (8/20) | 8.02 s
[Task 11/25] Current/Best: 20.09/ 21.50 GFLOPS | Progress: (12/20) | 10.21 s
[Task 11/25] Current/Best: 19.58/ 21.50 GFLOPS | Progress: (16/20) | 12.70 s
[Task 11/25] Current/Best: 10.21/ 21.50 GFLOPS | Progress: (20/20) | 15.92 s Done.
+
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 17.30/ 19.35 GFLOPS | Progress: (4/20) | 4.63 s
[Task 12/25] Current/Best: 19.39/ 19.39 GFLOPS | Progress: (8/20) | 7.53 s
[Task 12/25] Current/Best: 13.18/ 19.39 GFLOPS | Progress: (12/20) | 10.69 s
[Task 12/25] Current/Best: 5.79/ 19.39 GFLOPS | Progress: (16/20) | 13.22 s
[Task 12/25] Current/Best: 14.15/ 19.39 GFLOPS | Progress: (20/20) | 16.44 s Done.
+
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 16.96/ 20.43 GFLOPS | Progress: (4/20) | 4.67 s
[Task 13/25] Current/Best: 12.89/ 20.69 GFLOPS | Progress: (8/20) | 8.01 s
[Task 13/25] Current/Best: 19.32/ 21.77 GFLOPS | Progress: (12/20) | 10.81 s
[Task 13/25] Current/Best: 13.39/ 21.77 GFLOPS | Progress: (16/20) | 14.71 s
[Task 13/25] Current/Best: 12.98/ 21.77 GFLOPS | Progress: (20/20) | 17.92 s Done.
+
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 4.21/ 16.23 GFLOPS | Progress: (4/20) | 5.77 s
[Task 14/25] Current/Best: 5.97/ 16.23 GFLOPS | Progress: (8/20) | 8.38 s
[Task 14/25] Current/Best: 15.11/ 16.23 GFLOPS | Progress: (12/20) | 12.18 s Done.
+
[Task 14/25] Current/Best: 15.91/ 16.23 GFLOPS | Progress: (16/20) | 14.86 s
[Task 14/25] Current/Best: 9.41/ 16.23 GFLOPS | Progress: (20/20) | 21.43 s Done.
+
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 5.18/ 5.25 GFLOPS | Progress: (4/20) | 7.82 s
[Task 15/25] Current/Best: 13.69/ 17.17 GFLOPS | Progress: (8/20) | 14.83 s
[Task 15/25] Current/Best: 24.09/ 24.09 GFLOPS | Progress: (12/20) | 17.32 s
[Task 15/25] Current/Best: 10.06/ 24.09 GFLOPS | Progress: (16/20) | 20.85 s
[Task 15/25] Current/Best: 4.20/ 24.09 GFLOPS | Progress: (20/20) | 24.26 s
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 11.75/ 11.75 GFLOPS | Progress: (4/20) | 5.10 s
[Task 16/25] Current/Best: 16.46/ 16.46 GFLOPS | Progress: (8/20) | 7.70 s
[Task 16/25] Current/Best: 6.52/ 16.46 GFLOPS | Progress: (12/20) | 10.90 s
[Task 16/25] Current/Best: 18.41/ 18.41 GFLOPS | Progress: (16/20) | 13.23 s
[Task 16/25] Current/Best: 4.20/ 18.41 GFLOPS | Progress: (20/2
0) | 15.14 s Done.
+
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 8.75/ 17.61 GFLOPS | Progress: (4/20) | 4.43 s
[Task 17/25] Current/Best: 17.90/ 21.85 GFLOPS | Progress: (8/20) | 6.69 s
[Task 17/25] Current/Best: 7.81/ 21.85 GFLOPS | Progress: (12/20) | 9.31 s
[Task 17/25] Current/Best: 7.91/ 22.22 GFLOPS | Progress: (16/20) | 12.17 s
[Task 17/25] Current/Best: 18.57/ 22.22 GFLOPS | Progress: (20/20) | 14.81 s Done.
+
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 5.97/ 8.13 GFLOPS | Progress: (4/20) | 4.34 s
[Task 18/25] Current/Best: 6.04/ 14.41 GFLOPS | Progress: (8/20) | 10.15 s
[Task 18/25] Current/Best: 10.58/ 16.33 GFLOPS | Progress: (12/20) | 13.73 s
[Task 18/25] Current/Best: 13.20/ 16.33 GFLOPS | Progress: (16/20) | 19.42 s
[Task 18/25] Current/Best: 6.69/ 16.33 GFLOPS | Progress: (20/20) | 24.26 s Done.
+
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 12.22/ 18.45 GFLOPS | Progress: (4/20) | 5.10 s
[Task 19/25] Current/Best: 7.39/ 18.53 GFLOPS | Progress: (8/20) | 10.41 s
[Task 19/25] Current/Best: 11.97/ 18.53 GFLOPS | Progress: (12/20) | 14.87 s
[Task 19/25] Current/Best: 13.89/ 23.33 GFLOPS | Progress: (16/20) | 17.41 s
[Task 19/25] Current/Best: 1.58/ 23.33 GFLOPS | Progress: (20/20) | 22.93 s Done.
+
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 22.33/ 22.33 GFLOPS | Progress: (4/20) | 4.12 s
[Task 20/25] Current/Best: 11.07/ 22.33 GFLOPS | Progress: (8/20) | 6.66 s
[Task 20/25] Current/Best: 5.27/ 22.33 GFLOPS | Progress: (12/20) | 9.18 s
[Task 20/25] Current/Best: 19.75/ 22.33 GFLOPS | Progress: (16/20) | 11.84 s Done.
+
[Task 20/25] Current/Best: 14.46/ 22.33 GFLOPS | Progress: (20/20) | 14.45 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 9.04/ 23.01 GFLOPS | Progress: (4/20) | 7.02 s
[Task 21/25] Current/Best: 15.99/ 23.01 GFLOPS | Progress: (8/20) | 9.07 s
[Task 21/25] Current/Best: 19.04/ 23.01 GFLOPS | Progress: (12/20) | 11.51 s
[Task 21/25] Current/Best: 9.60/ 23.01 GFLOPS | Progress: (16/20) | 13.45 s
[Task 21/25] Current/Best: 4.89/ 23.01 GFLOPS | Progress: (20/20) | 15.54 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 10.50/ 10.78 GFLOPS | Progress: (4/20) | 6.31 s
[Task 22/25] Current/Best: 2.67/ 10.79 GFLOPS | Progress: (8/20) | 8.91 s
[Task 22/25] Current/Best: 13.51/ 13.51 GFLOPS | Progress: (12/20) | 10.76 s
[Task 22/25] Current/Best: 7.05/ 14.11 GFLOPS | Progress: (16/20
) | 15.58 s
[Task 22/25] Current/Best: 14.74/ 17.59 GFLOPS | Progress: (20/20) | 17.33 s Done.
+
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 24.35/ 24.35 GFLOPS | Progress: (4/20) | 4.48 s
[Task 23/25] Current/Best: 17.19/ 24.35 GFLOPS | Progress: (8/20) | 7.39 s
[Task 23/25] Current/Best: 13.10/ 24.35 GFLOPS | Progress: (12/20) | 10.18 s
[Task 23/25] Current/Best: 20.51/ 24.35 GFLOPS | Progress: (16/20) | 13.99 s
[Task 23/25] Current/Best: 2.73/ 24.35 GFLOPS | Progress: (20/20) | 18.57 s Done.
+
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 1.67/ 3.83 GFLOPS | Progress: (4/20) | 10.76 s
[Task 24/25] Current/Best: 1.79/ 4.96 GFLOPS | Progress: (8/20) | 13.00 s
[Task 24/25] Current/Best: 2.21/ 10.09 GFLOPS | Progress: (12/20) | 23.93 s
[Task 24/25] Current/Best: 6.04/ 10.35 GFLOPS | Progress: (16/20) | 35.36 s
[Task 24/25] Current/Best: 1.75/ 10.35 GFLOPS | Progress: (20/20) | 41.88 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
+ Done.
+
[Task 25/25] Current/Best: 3.07/ 8.78 GFLOPS | Progress: (4/20) | 3.15 s
[Task 25/25] Current/Best: 7.87/ 9.61 GFLOPS | Progress: (8/20) | 14.08 s
[Task 25/25] Current/Best: 2.83/ 9.61 GFLOPS | Progress: (12/20) | 16.37 s
[Task 25/25] Current/Best: 9.12/ 9.61 GFLOPS | Progress: (16/20) | 27.01 s
[Task 25/25] Current/Best: 8.63/ 9.61 GFLOPS | Progress: (20/20) | 29.95 s
@@ -721,8 +722,8 @@ improvement in comparing the optimized model to the unoptimized model.
.. code-block:: none
- optimized: {'mean': 434.1846287700173, 'median': 432.1034665501429, 'std': 4.478247307073096}
- unoptimized: {'mean': 515.8801153599779, 'median': 516.1550652997903, 'std': 2.0908542727619}
+ optimized: {'mean': 402.5487477599995, 'median': 402.2745003499949, 'std': 2.6479013247398746}
+ unoptimized: {'mean': 476.6050386599976, 'median': 476.417146350002, 'std': 1.4322166401243506}
@@ -745,7 +746,7 @@ profiling/benchmarking.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 11 minutes 8.430 seconds)
+ **Total running time of the script:** ( 11 minutes 31.656 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 09d68d8e06..2c7c46f61a 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
- 1.247e-07 secs/op
+ 1.195e-07 secs/op
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index 78cf773a54..a28c78350f 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -268,7 +268,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
.. code-block:: none
- [stage(a, placeholder(a, 0x19939f90)), stage(b, placeholder(b, 0x915d560)), 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. [...]
+ [stage(a, placeholder(a, 0x112f4ea0)), stage(b, placeholder(b, 0x210f7340)), 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 c4e87fb1e8..cf931db2d6 100644
--- a/docs/_sources/tutorial/sg_execution_times.rst.txt
+++ b/docs/_sources/tutorial/sg_execution_times.rst.txt
@@ -5,32 +5,32 @@
Computation times
=================
-**14:45.861** total execution time for **tutorial** files:
+**15:03.438** total execution time for **tutorial** files:
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 11:08.430 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 11:31.656 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:29.862 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:33.752 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 01:01.396 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 00:58.515 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:34.907 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:34.634 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:29.416 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:22.768 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:00.871 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:01.135 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``) | 00:00.818 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``) | 00:00.809 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.160 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.169 | 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_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_tvmc_command_line_driver.py` (``tvmc_command_line_driver.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 5eef2efb22..d4e111f45a 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -286,7 +286,7 @@ helper function to run a profile of the TVM generated code.
.. code-block:: none
Numpy running time: 0.000007
- naive: 0.000007
+ naive: 0.000006
@@ -389,7 +389,7 @@ compile and run this new schedule with the parallel operation applied:
.. code-block:: none
- parallel: 0.000007
+ parallel: 0.000006
@@ -444,7 +444,7 @@ factor to be the number of threads on your CPU.
.. code-block:: none
- vector: 0.000024
+ vector: 0.000025
@I.ir_module
class Module:
@T.prim_func
@@ -498,10 +498,10 @@ We can now compare the different schedules
.. code-block:: none
Operator Timing Performance
- numpy 6.9473799885599875e-06 1.0
- naive 6.6833e-06 0.961988549784978
- parallel 6.990599999999999e-06 1.0062210518945531
- vector 2.4442700000000002e-05 3.518261566266558
+ numpy 6.961720000617788e-06 1.0
+ naive 5.8359e-06 0.8382842170443681
+ parallel 6.1383e-06 0.881721758337779
+ vector 2.53837e-05 3.646182264978688
@@ -922,7 +922,7 @@ matrix multiplication.
.. code-block:: none
- Numpy running time: 0.017477
+ Numpy running time: 0.015945
@@ -980,7 +980,7 @@ optimizations.
.. code-block:: none
- none: 3.454114
+ none: 3.266834
@@ -1077,7 +1077,7 @@ schedule.
.. code-block:: none
- blocking: 0.302482
+ blocking: 0.300061
@@ -1158,7 +1158,7 @@ already cache friendly from our previous optimizations.
.. code-block:: none
- vectorization: 0.336674
+ vectorization: 0.335525
@I.ir_module
class Module:
@T.prim_func
@@ -1221,7 +1221,7 @@ more cache friendly.
.. code-block:: none
- loop permutation: 0.114511
+ loop permutation: 0.116691
@I.ir_module
class Module:
@T.prim_func
@@ -1309,7 +1309,7 @@ optimized schedule.
.. code-block:: none
- array packing: 0.109034
+ array packing: 0.106637
@I.ir_module
class Module:
@T.prim_func
@@ -1389,7 +1389,7 @@ to `C` when all the block results are ready.
.. code-block:: none
- block caching: 0.110991
+ block caching: 0.102687
@I.ir_module
class Module:
@T.prim_func
@@ -1460,7 +1460,7 @@ of thread-level parallelization.
.. code-block:: none
- parallelization: 0.145964
+ parallelization: 0.134827
@I.ir_module
class Module:
@T.prim_func
@@ -1527,13 +1527,13 @@ working, we can compare the results.
.. code-block:: none
Operator Timing Performance
- none 3.4541139238000005 1.0
- blocking 0.3024817235 0.08757143805124659
- vectorization 0.33667420459999997 0.09747049808641287
- loop permutation 0.11451105589999999 0.0331520784855938
- array packing 0.1090344682 0.03156655240833721
- block caching 0.1109909353 0.032132968902743864
- parallelization 0.1459641247 0.04225805167984135
+ none 3.2668338954 1.0
+ blocking 0.3000612884 0.09185079438000006
+ vectorization 0.33552476750000004 0.10270640572587712
+ loop permutation 0.1166908167 0.03571984999430529
+ array packing 0.10663652380000002 0.032642162783407494
+ block caching 0.1026865111 0.031433037120311494
+ parallelization 0.1348268558 0.041271414500090906
@@ -1573,11 +1573,6 @@ operations with tunable parameters that allows you to automatically optimize
the computation for specific platforms.
-.. rst-class:: sphx-glr-timing
-
- **Total running time of the script:** ( 1 minutes 1.396 seconds)
-
-
.. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
.. only:: html
diff --git a/docs/commit_hash b/docs/commit_hash
index 60a22524f3..6b16028cc6 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-c8973a550790babcf2d1860c6d7fac889bb9fe15
+fe372a73096a172ce6e5e8879d2628d84d58cb93
diff --git a/docs/how_to/compile_models/from_darknet.html b/docs/how_to/compile_models/from_darknet.html
index 7b0e465092..3ef827fb3f 100644
--- a/docs/how_to/compile_models/from_darknet.html
+++ b/docs/how_to/compile_models/from_darknet.html
@@ -585,7 +585,7 @@ class:['truck 0.9266'] left:471 top:83 right:689 bottom:169
class:['bicycle 0.9984'] 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 13.657 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 13.704 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 cbb9deddee..485364a160 100644
--- a/docs/how_to/compile_models/from_keras.html
+++ b/docs/how_to/compile_models/from_keras.html
@@ -506,7 +506,7 @@ 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 900ms/step
+1/1 [==============================] - 1s 926ms/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 9cae045919..18bcdde7f1 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -439,7 +439,7 @@
<span class="nb">print</span><span class="p">(</span><span class="s2">"x"</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.zip13fd4e9a-1499-4f6e-8b30-76b6c5e5da10 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.zip3f917333-f21e-4db2-976c-eb635d245168 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 51fad9cf51..81681cabe4 100644
--- a/docs/how_to/compile_models/from_oneflow.html
+++ b/docs/how_to/compile_models/from_oneflow.html
@@ -449,11 +449,12 @@ 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: "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]
- 19%|#9 | 7.99M/41.5M [00:00<00:00, 47.2MB/s]
- 47%|####6 | 19.4M/41.5M [00:00<00:00, 79.1MB/s]
- 67%|######7 | 27.9M/41.5M [00:00<00:00, 66.7MB/s]
- 84%|########4 | 34.9M/41.5M [00:00<00:00, 65.3MB/s]
-100%|##########| 41.5M/41.5M [00:00<00:00, 67.9MB/s]
+ 19%|#9 | 7.99M/41.5M [00:00<00:00, 53.2MB/s]
+ 39%|###8 | 16.0M/41.5M [00:00<00:00, 67.2MB/s]
+ 57%|#####7 | 23.7M/41.5M [00:00<00:00, 68.2MB/s]
+ 73%|#######3 | 30.4M/41.5M [00:00<00:00, 61.8MB/s]
+ 88%|########7 | 36.4M/41.5M [00:00<00:00, 61.3MB/s]
+100%|##########| 41.5M/41.5M [00:00<00:00, 65.0MB/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 55eee18ab5..8b97a5d3fc 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -432,11 +432,10 @@ be unstable.</p>
Downloading: "https://download.pytorch.org/models/resnet18-f37072fd.pth" to /workspace/.cache/torch/hub/checkpoints/resnet18-f37072fd.pth
0%| | 0.00/44.7M [00:00<?, ?B/s]
- 18%|#7 | 8.02M/44.7M [00:00<00:00, 84.1MB/s]
- 36%|###5 | 16.0M/44.7M [00:00<00:00, 75.3MB/s]
- 62%|######1 | 27.6M/44.7M [00:00<00:00, 95.0MB/s]
- 82%|########2 | 36.8M/44.7M [00:00<00:00, 88.1MB/s]
-100%|##########| 44.7M/44.7M [00:00<00:00, 77.9MB/s]
+ 27%|##7 | 12.1M/44.7M [00:00<00:00, 127MB/s]
+ 54%|#####4 | 24.2M/44.7M [00:00<00:00, 108MB/s]
+ 78%|#######7 | 34.7M/44.7M [00:00<00:00, 105MB/s]
+100%|##########| 44.7M/44.7M [00:00<00:00, 102MB/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 06dddd1103..c9dd56bd01 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -649,7 +649,7 @@ 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 17.037 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 18.916 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 fef0b75dc1..fb9af36b4a 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-compile-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>06:04.501</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>06:07.485</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 81%" />
@@ -349,43 +349,43 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="from_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>01:17.037</p></td>
+<td><p>01:18.916</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>01:13.657</p></td>
+<td><p>01:13.704</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>00:49.361</p></td>
+<td><p>00:50.021</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_oneflow.html#sphx-glr-how-to-compile-models-from-oneflow-py"><span class="std std-ref">Compile OneFlow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_oneflow.py</span></code>)</p></td>
-<td><p>00:33.583</p></td>
+<td><p>00:34.358</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="from_mxnet.html#sphx-glr-how-to-compile-models-from-mxnet-py"><span class="std std-ref">Compile MXNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_mxnet.py</span></code>)</p></td>
-<td><p>00:29.796</p></td>
+<td><p>00:29.718</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_coreml.html#sphx-glr-how-to-compile-models-from-coreml-py"><span class="std std-ref">Compile CoreML Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_coreml.py</span></code>)</p></td>
-<td><p>00:29.253</p></td>
+<td><p>00:29.571</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="from_tflite.html#sphx-glr-how-to-compile-models-from-tflite-py"><span class="std std-ref">Compile TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tflite.py</span></code>)</p></td>
-<td><p>00:26.384</p></td>
+<td><p>00:25.936</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_pytorch.html#sphx-glr-how-to-compile-models-from-pytorch-py"><span class="std std-ref">Compile PyTorch Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_pytorch.py</span></code>)</p></td>
-<td><p>00:23.531</p></td>
+<td><p>00:23.165</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>00:19.329</p></td>
+<td><p>00:19.579</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>00:02.571</p></td>
+<td><p>00:02.516</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 80f8f0b0d4..1a7a025c0f 100644
--- a/docs/how_to/deploy_models/deploy_model_on_adreno.html
+++ b/docs/how_to/deploy_models/deploy_model_on_adreno.html
@@ -920,10 +920,10 @@ Top5 predictions:
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 3340.0471 3339.4969 3343.5940 3338.1399 1.8071
+ 3290.8978 3289.1562 3299.1580 3288.0557 3.3897
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 2.299 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 1.871 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 53493dd9c8..b43356b3e7 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -662,7 +662,7 @@ to the remote android device.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 15.5180 15.5055 15.6018 15.4754 0.0393
+ 15.1210 14.9484 15.9747 14.8685 0.3481
</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 79621cd20f..448c5e212d 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -454,27 +454,23 @@ be unstable.</p>
Downloading: "https://download.pytorch.org/models/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth" to /workspace/.cache/torch/hub/checkpoints/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth
0%| | 0.00/170M [00:00<?, ?B/s]
- 5%|4 | 7.99M/170M [00:00<00:02, 59.9MB/s]
- 9%|9 | 16.0M/170M [00:00<00:02, 70.7MB/s]
- 14%|#4 | 24.0M/170M [00:00<00:02, 64.7MB/s]
- 19%|#8 | 32.0M/170M [00:00<00:02, 66.7MB/s]
- 24%|##3 | 40.0M/170M [00:00<00:01, 69.1MB/s]
- 28%|##8 | 48.0M/170M [00:00<00:01, 64.8MB/s]
- 33%|###2 | 56.0M/170M [00:00<00:01, 69.8MB/s]
- 38%|###7 | 64.0M/170M [00:00<00:01, 73.6MB/s]
- 42%|####2 | 72.0M/170M [00:01<00:01, 66.8MB/s]
- 48%|####8 | 82.3M/170M [00:01<00:01, 77.7MB/s]
- 53%|#####3 | 90.0M/170M [00:01<00:01, 66.2MB/s]
- 57%|#####7 | 96.8M/170M [00:01<00:01, 67.0MB/s]
- 61%|######1 | 104M/170M [00:01<00:01, 64.8MB/s]
- 66%|######5 | 112M/170M [00:01<00:00, 67.2MB/s]
- 71%|####### | 120M/170M [00:01<00:00, 70.7MB/s]
- 75%|#######5 | 128M/170M [00:01<00:00, 73.0MB/s]
- 80%|######## | 136M/170M [00:02<00:00, 64.9MB/s]
- 88%|########7 | 149M/170M [00:02<00:00, 83.1MB/s]
- 93%|#########2| 158M/170M [00:02<00:00, 74.1MB/s]
- 99%|#########8| 168M/170M [00:02<00:00, 82.5MB/s]
-100%|##########| 170M/170M [00:02<00:00, 71.6MB/s]
+ 5%|5 | 8.84M/170M [00:00<00:01, 92.7MB/s]
+ 12%|#2 | 20.6M/170M [00:00<00:01, 111MB/s]
+ 18%|#8 | 31.2M/170M [00:00<00:01, 101MB/s]
+ 25%|##5 | 42.5M/170M [00:00<00:01, 107MB/s]
+ 31%|###1 | 52.8M/170M [00:00<00:01, 105MB/s]
+ 37%|###7 | 62.9M/170M [00:00<00:01, 103MB/s]
+ 43%|####2 | 72.8M/170M [00:00<00:01, 101MB/s]
+ 49%|####8 | 82.5M/170M [00:00<00:00, 101MB/s]
+ 54%|#####4 | 92.2M/170M [00:00<00:00, 95.9MB/s]
+ 61%|###### | 103M/170M [00:01<00:00, 102MB/s]
+ 67%|######6 | 113M/170M [00:01<00:00, 102MB/s]
+ 72%|#######2 | 123M/170M [00:01<00:00, 101MB/s]
+ 78%|#######8 | 133M/170M [00:01<00:00, 86.2MB/s]
+ 87%|########7 | 148M/170M [00:01<00:00, 105MB/s]
+ 93%|#########3| 158M/170M [00:01<00:00, 104MB/s]
+ 99%|#########9| 169M/170M [00:01<00:00, 92.0MB/s]
+100%|##########| 170M/170M [00:01<00:00, 99.3MB/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=& [...]
@@ -572,7 +568,7 @@ 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 20.745 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes 19.889 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 285392f9d4..ffc33c3a86 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -495,8 +495,8 @@ training. Other models require a full post training calibration.</p>
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]
- 59%|#####8 | 7.99M/13.6M [00:00<00:00, 78.7MB/s]
-100%|##########| 13.6M/13.6M [00:00<00:00, 100MB/s]
+ 91%|######### | 12.3M/13.6M [00:00<00:00, 129MB/s]
+100%|##########| 13.6M/13.6M [00:00<00:00, 125MB/s]
</pre></div>
</div>
</div>
@@ -587,7 +587,7 @@ 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)
- 90.2470 90.1751 91.1447 89.9100 0.2500
+ 88.4134 88.2537 92.0269 88.0521 0.5086
</pre></div>
</div>
<div class="admonition note">
@@ -626,7 +626,7 @@ 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 11.472 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 11.225 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 b569177ccb..7e8791272e 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -580,7 +580,7 @@ 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)
- 119.3888 119.3236 126.3505 118.3648 0.8178
+ 118.3385 118.1645 130.6115 117.2539 1.3747
</pre></div>
</div>
<div class="admonition note">
@@ -608,7 +608,7 @@ 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 25.849 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 22.767 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 281be00a75..4e1150960c 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -521,7 +521,7 @@ 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 27.242 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 26.919 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 ea95d479eb..d3c05a3b01 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -463,24 +463,23 @@ to your device.</p>
Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
0%| | 0/132723 [00:00<?, ?KB/s]
- 4%|4 | 5432/132723 [00:00<00:02, 54315.91KB/s]
- 10%|9 | 12935/132723 [00:00<00:01, 66495.76KB/s]
- 16%|#5 | 20870/132723 [00:00<00:01, 72361.84KB/s]
- 21%|##1 | 28456/132723 [00:00<00:01, 73733.94KB/s]
- 27%|##7 | 35953/132723 [00:00<00:01, 74176.58KB/s]
- 33%|###2 | 43702/132723 [00:00<00:01, 75225.80KB/s]
- 39%|###8 | 51406/132723 [00:00<00:01, 75811.69KB/s]
- 44%|####4 | 59001/132723 [00:00<00:00, 75854.05KB/s]
- 50%|##### | 66699/132723 [00:00<00:00, 76204.27KB/s]
- 56%|#####5 | 74320/132723 [00:01<00:00, 76092.06KB/s]
- 62%|######1 | 81942/132723 [00:01<00:00, 76128.53KB/s]
- 68%|######7 | 89654/132723 [00:01<00:00, 76424.97KB/s]
- 73%|#######3 | 97297/132723 [00:01<00:00, 75956.41KB/s]
- 79%|#######9 | 104929/132723 [00:01<00:00, 76063.66KB/s]
- 85%|########4 | 112575/132723 [00:01<00:00, 76181.49KB/s]
- 91%|######### | 120194/132723 [00:01<00:00, 76109.69KB/s]
- 96%|#########6| 127847/132723 [00:01<00:00, 76232.26KB/s]
-100%|##########| 132723/132723 [00:01<00:00, 75327.12KB/s]
+ 3%|3 | 4205/132723 [00:00<00:03, 42042.02KB/s]
+ 9%|9 | 12397/132723 [00:00<00:01, 65492.72KB/s]
+ 16%|#5 | 20610/132723 [00:00<00:01, 73084.90KB/s]
+ 22%|##1 | 28851/132723 [00:00<00:01, 76764.61KB/s]
+ 28%|##7 | 37019/132723 [00:00<00:01, 78521.79KB/s]
+ 34%|###4 | 45221/132723 [00:00<00:01, 79706.87KB/s]
+ 40%|#### | 53417/132723 [00:00<00:00, 80440.38KB/s]
+ 46%|####6 | 61675/132723 [00:00<00:00, 81112.85KB/s]
+ 53%|#####2 | 69933/132723 [00:00<00:00, 81569.75KB/s]
+ 59%|#####8 | 78160/132723 [00:01<00:00, 81783.34KB/s]
+ 65%|######5 | 86468/132723 [00:01<00:00, 82178.13KB/s]
+ 71%|#######1 | 94742/132723 [00:01<00:00, 82347.15KB/s]
+ 78%|#######7 | 103035/132723 [00:01<00:00, 82518.19KB/s]
+ 84%|########3 | 111310/132723 [00:01<00:00, 82586.67KB/s]
+ 90%|######### | 119570/132723 [00:01<00:00, 82587.98KB/s]
+ 96%|#########6| 127885/132723 [00:01<00:00, 82755.38KB/s]
+100%|##########| 132723/132723 [00:01<00:00, 80010.08KB/s]
</pre></div>
</div>
<p>Create TVM runtime and do inference
@@ -519,7 +518,7 @@ 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 26.172 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 28.286 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 4c8bce41a1..5928667a4c 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-deploy-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>14:24.707</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>14:22.549</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 86%" />
@@ -349,39 +349,39 @@
</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>03:26.172</p></td>
+<td><p>03:28.286</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>03:20.745</p></td>
+<td><p>03:19.889</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>02:25.849</p></td>
+<td><p>02:22.767</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>01:27.242</p></td>
+<td><p>01:26.919</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>01:11.472</p></td>
+<td><p>01:11.225</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>01:02.299</p></td>
+<td><p>01:01.871</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>00:38.780</p></td>
+<td><p>00:39.316</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>00:26.264</p></td>
+<td><p>00:26.340</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>00:25.878</p></td>
+<td><p>00:25.931</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 11aaed4e91..50ec7628e1 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -619,7 +619,7 @@ 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.zipaafa5250-ff11-4c64-9fe6-402dccc856f2 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.zip13e7efad-d6aa-4e86-a528-4a394bd59da2 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 834bbfc323..8871a5e7dd 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-extend-tvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:51.840</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:53.524</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -349,15 +349,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="bring_your_own_datatypes.html#sphx-glr-how-to-extend-tvm-bring-your-own-datatypes-py"><span class="std std-ref">Bring Your Own Datatypes to TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">bring_your_own_datatypes.py</span></code>)</p></td>
-<td><p>00:48.105</p></td>
+<td><p>00:49.713</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>00:02.665</p></td>
+<td><p>00:02.708</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>00:01.064</p></td>
+<td><p>00:01.096</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="low_level_custom_pass.html#sphx-glr-how-to-extend-tvm-low-level-custom-pass-py"><span class="std std-ref">Writing a Customized Pass</span></a> (<code class="docutils literal notranslate"><span class="pre">low_level_custom_pass.py</span></code>)</p></td>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index 07cac9fa1c..17aa24526a 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -526,10 +526,10 @@ 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: 17944us [17944us] (48.40%; 48.40%)
-FoldScaleAxis: 19129us [8us] (51.60%; 51.60%)
- FoldConstant: 19121us [1694us] (51.58%; 99.96%)
- InferType: 17427us [17427us] (47.01%; 91.14%)
+InferType: 18786us [18786us] (48.49%; 48.49%)
+FoldScaleAxis: 19960us [9us] (51.51%; 51.51%)
+ FoldConstant: 19951us [1796us] (51.49%; 99.96%)
+ InferType: 18155us [18155us] (46.86%; 91.00%)
</pre></div>
</div>
</div>
@@ -551,10 +551,10 @@ 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: 18260us [18260us] (48.89%; 48.89%)
-FoldScaleAxis: 19088us [6us] (51.11%; 51.11%)
- FoldConstant: 19082us [1798us] (51.09%; 99.97%)
- InferType: 17285us [17285us] (46.28%; 90.58%)
+InferType: 18397us [18397us] (48.12%; 48.12%)
+FoldScaleAxis: 19835us [8us] (51.88%; 51.88%)
+ FoldConstant: 19827us [1821us] (51.86%; 99.96%)
+ InferType: 18006us [18006us] (47.10%; 90.82%)
</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 a5c753c1a2..78edcf3bbb 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -575,7 +575,7 @@ latency of convolution.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Convolution: </span><span class="si">%f</span><span class="s2"> ms"</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: 54.228958 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 33.413215 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 99d8b97981..5600e94e3e 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -861,7 +861,7 @@ be able to run on our build server</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"conv2d with tensor core: </span><span class="si">%f</span><span class="s2"> ms"</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: 6.580464 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 13.337607 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 91ea5227e7..290082c82e 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -472,8 +472,8 @@ 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">"Baseline: </span><span class="si">%f</span><span class="s2">"</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: 0.018212
-Baseline: 3.443270
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019248
+Baseline: 3.213059
</pre></div>
</div>
<p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -529,7 +529,7 @@ 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">"Opt1: </span><span class="si">%f</span><span class="s2">"</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: 0.298191
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.307405
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -583,7 +583,7 @@ vastly.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Opt2: </span><span class="si">%f</span><span class="s2">"</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: 0.336805
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.335485
</pre></div>
</div>
<p>Here is the generated IR after vectorization.</p>
@@ -635,7 +635,7 @@ the access pattern for A matrix is more cache friendly.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Opt3: </span><span class="si">%f</span><span class="s2">"</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: 0.114357
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.115603
</pre></div>
</div>
<p>Here is the generated IR after loop permutation.</p>
@@ -709,7 +709,7 @@ flattening.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Opt4: </span><span class="si">%f</span><span class="s2">"</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: 0.109719
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.109914
</pre></div>
</div>
<p>Here is the generated IR after array packing.</p>
@@ -784,7 +784,7 @@ write to C when all the block results are ready.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"Opt5: </span><span class="si">%f</span><span class="s2">"</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: 0.111613
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.115596
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -861,7 +861,7 @@ class Module:
<span class="nb">print</span><span class="p">(</span><span class="s2">"Opt6: </span><span class="si">%f</span><span class="s2">"</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: 0.147156
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.152508
</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 726c1c00d3..f770f52360 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-optimize-operators-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:34.886</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:34.921</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -349,15 +349,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="opt_gemm.html#sphx-glr-how-to-optimize-operators-opt-gemm-py"><span class="std std-ref">How to optimize GEMM on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_gemm.py</span></code>)</p></td>
-<td><p>00:32.353</p></td>
+<td><p>00:32.193</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>00:01.469</p></td>
+<td><p>00:01.586</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>00:01.064</p></td>
+<td><p>00:01.142</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 7fb0b90ab7..b0c055e247 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-tune-with-autoscheduler-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>09:11.432</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>09:32.550</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 85%" />
@@ -349,27 +349,27 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_conv2d_layer_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-conv2d-layer-cuda-py"><span class="std std-ref">Auto-scheduling a Convolution Layer for GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_layer_cuda.py</span></code>)</p></td>
-<td><p>05:36.422</p></td>
+<td><p>05:45.586</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>01:36.756</p></td>
+<td><p>01:39.667</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_network_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-cuda-py"><span class="std std-ref">Auto-scheduling a Neural Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_cuda.py</span></code>)</p></td>
-<td><p>01:04.945</p></td>
+<td><p>01:07.544</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>00:27.406</p></td>
+<td><p>00:31.542</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>00:13.524</p></td>
+<td><p>00:14.666</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>00:12.380</p></td>
+<td><p>00:13.545</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 dfb3b74a30..bc12b9b73a 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
@@ -503,768 +503,481 @@ class Module:
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.env_thread("blockIdx.x")
- T.launch_thread(blockIdx_x, 8)
- conv2d_nchw = T.allocate([28], "float32", "local")
- pad_temp_shared = T.allocate([504], "float32", "shared")
- kernel_shared = T.allocate([1536], "float32", "shared")
+ T.launch_thread(blockIdx_x, 28)
+ 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.env_thread("threadIdx.x")
- T.launch_thread(threadIdx_x, 112)
- conv2d_nchw_1 = T.buffer_decl((28,), data=conv2d_nchw, scope="local")
+ T.launch_thread(threadIdx_x, 64)
+ conv2d_nchw_1 = T.buffer_decl((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
- conv2d_nchw_1[7] = T.float32(0)
- conv2d_nchw_1[14] = T.float32(0)
- conv2d_nchw_1[21] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
- conv2d_nchw_1[8] = T.float32(0)
- conv2d_nchw_1[15] = T.float32(0)
- conv2d_nchw_1[22] = T.float32(0)
conv2d_nchw_1[2] = T.float32(0)
- conv2d_nchw_1[9] = T.float32(0)
- conv2d_nchw_1[16] = T.float32(0)
- conv2d_nchw_1[23] = T.float32(0)
conv2d_nchw_1[3] = T.float32(0)
- conv2d_nchw_1[10] = T.float32(0)
- conv2d_nchw_1[17] = T.float32(0)
- conv2d_nchw_1[24] = T.float32(0)
conv2d_nchw_1[4] = T.float32(0)
- conv2d_nchw_1[11] = T.float32(0)
- conv2d_nchw_1[18] = T.float32(0)
- conv2d_nchw_1[25] = T.float32(0)
conv2d_nchw_1[5] = T.float32(0)
- conv2d_nchw_1[12] = T.float32(0)
- conv2d_nchw_1[19] = T.float32(0)
- conv2d_nchw_1[26] = T.float32(0)
conv2d_nchw_1[6] = T.float32(0)
+ 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)
- conv2d_nchw_1[20] = T.float32(0)
- conv2d_nchw_1[27] = T.float32(0)
for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
- cse_var_4: T.int32 = rc_outer_outer * 392
- cse_var_3: T.int32 = ry_outer_outer * 7
cse_var_2: T.int32 = rc_outer_outer * 72
cse_var_1: T.int32 = ry_outer_outer * 3
threadIdx_x_1 = T.env_thread("threadIdx.x")
- pad_temp_shared_1 = T.buffer_decl((504,), data=pad_temp_shared, scope="shared")
- data_1 = T.buffer_decl((25088,), data=data.data)
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1] = T.if_then_else(1 <= threadIdx_x_1 % 63 // 9 + ry_outer_outer and threadIdx_x_1 % 63 // 9 + ry_outer_outer < 8 and 1 <= threadIdx_x_1 % 9 and threadIdx_x_1 % 9 < 8, data_1[cse_var_4 + threadIdx_x_1 // 9 * 7 + cse_var_3 + threadIdx_x_1 % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 112] = T.if_then_else(1 <= (threadIdx_x_1 + 49) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 49) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 4) % 9 and (threadIdx_x_1 + 4) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 112) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 4) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= (threadIdx_x_1 + 35) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 35) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 8) % 9 and (threadIdx_x_1 + 8) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 224) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 8) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= (threadIdx_x_1 + 21) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 21) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 3) % 9 and (threadIdx_x_1 + 3) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 336) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 3) % 9 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- if T.likely(threadIdx_x_1 < 56):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= (threadIdx_x_1 + 7) % 63 // 9 + ry_outer_outer and (threadIdx_x_1 + 7) % 63 // 9 + ry_outer_outer < 8 and 1 <= (threadIdx_x_1 + 7) % 9 and (threadIdx_x_1 + 7) % 9 < 8, data_1[cse_var_4 + (threadIdx_x_1 + 448) // 9 * 7 + cse_var_3 + (threadIdx_x_1 + 7) % 9 - 8], T.float32(0))
+ pad_temp_shared_1 = T.buffer_decl((72,), data=pad_temp_shared, scope="shared")
+ with T.launch_thread(threadIdx_x_1, 64):
+ data_1 = T.buffer_decl((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_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.buffer_decl((1536,), data=kernel_shared, scope="shared")
+ kernel_shared_1 = T.buffer_decl((3072,), data=kernel_shared, scope="shared")
kernel_1 = T.buffer_decl((2359296,), data=kernel.data)
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 112] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 112) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 224] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 224) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 560) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 672] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 784) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 193536]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1120] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1120) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1232] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1232) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x * 294912 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 258048]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 80):
- kernel_shared_1[threadIdx_x_2 + 1456] = kernel_1[blockIdx_x * 294912 + (threadIdx_x_2 + 1456) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 24]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 48]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 72]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 1] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 1]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 25]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 49]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 73]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 2] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 3] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 4] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 5] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 6] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 7] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 2]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 26]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 50]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 8] * kernel_shared_1[threadIdx_x // 7 * 96 + 74]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 63] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 3]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 27]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 51]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 75]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 64] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 4]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 28]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 52]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 76]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 65] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 66] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 67] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 68] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 69] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 70] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 5]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 29]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 53]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 71] * kernel_shared_1[threadIdx_x // 7 * 96 + 77]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 126] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 6]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 30]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 54]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 78]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 127] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 7]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 31]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 55]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 79]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 128] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 129] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 130] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 131] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 132] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 133] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 8]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 32]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 56]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 134] * kernel_shared_1[threadIdx_x // 7 * 96 + 80]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 189] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 9]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 33]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 57]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 81]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 190] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 10]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 34]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 58]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 82]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 191] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 192] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 193] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 194] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 195] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 196] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 11]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 35]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 59]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 197] * kernel_shared_1[threadIdx_x // 7 * 96 + 83]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 252] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 12]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 36]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 60]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 84]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 253] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 13]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 37]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 61]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 85]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 254] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 255] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 256] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 257] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 258] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 259] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 14]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 38]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 62]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 260] * kernel_shared_1[threadIdx_x // 7 * 96 + 86]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 315] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 15]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 39]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 63]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 87]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 316] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 16]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 40]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 64]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 88]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 317] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 318] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 319] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 320] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 321] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 322] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 17]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 41]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 65]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 323] * kernel_shared_1[threadIdx_x // 7 * 96 + 89]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 378] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 18]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 42]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 66]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 90]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 379] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 19]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 43]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 67]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 91]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 380] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 381] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 382] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 383] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 384] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 385] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 20]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 44]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 68]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 386] * kernel_shared_1[threadIdx_x // 7 * 96 + 92]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 441] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 21]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 45]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 69]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 93]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 442] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 22]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 46]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 70]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 94]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[14] = conv2d_nchw_1[14] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[21] = conv2d_nchw_1[21] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 443] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[15] = conv2d_nchw_1[15] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[22] = conv2d_nchw_1[22] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 444] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[16] = conv2d_nchw_1[16] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[23] = conv2d_nchw_1[23] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 445] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[17] = conv2d_nchw_1[17] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[24] = conv2d_nchw_1[24] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 446] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[18] = conv2d_nchw_1[18] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[25] = conv2d_nchw_1[25] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 447] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[19] = conv2d_nchw_1[19] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[26] = conv2d_nchw_1[26] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 448] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 23]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 47]
- conv2d_nchw_1[20] = conv2d_nchw_1[20] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 71]
- conv2d_nchw_1[27] = conv2d_nchw_1[27] + pad_temp_shared_1[threadIdx_x % 7 * 9 + 449] * kernel_shared_1[threadIdx_x // 7 * 96 + 95]
- for i1_inner, i3_inner in T.grid(4, 7):
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 64] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 128] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 256] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 320] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 512] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 640] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 704] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 832] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1024] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1088] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1216] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1280] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1408] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1472] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1600] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1664] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1792] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1856] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 1984] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2048] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2176] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2240] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2368] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2432] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2560] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2624] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2752] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2816] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 2) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_2 // 24 * 4608 + cse_var_2 + threadIdx_x_2 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_2 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 2944] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 1) % 3]
+ with T.launch_thread(threadIdx_x_2, 64):
+ kernel_shared_1[threadIdx_x_2 + 3008] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_2 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_2 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_2 + 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_decl((25088,), data=compute.data)
bias_1 = T.buffer_decl((512,), data=bias.data)
- compute_1[blockIdx_x * 3136 + threadIdx_x // 7 * 196 + i1_inner * 49 + threadIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x * 64 + threadIdx_x // 7 * 4 + i1_inner], 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>
@@ -1298,7 +1011,7 @@ 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: 0.418 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.351 ms
</pre></div>
</div>
</div>
@@ -1327,20 +1040,20 @@ conv2d_nchw_nn_o_i, conv2d_nchw_nn_i = s[conv2d_nchw].split(conv2d_nchw_nn, fact
conv2d_nchw_nn_o_o_i, conv2d_nchw_nn_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_i, factor=1)
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=4)
-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=16)
+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=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=1)
-conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=7)
+conv2d_nchw_yy_o_o_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=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=1)
-conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=8)
+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=1)
@@ -1349,11 +1062,11 @@ s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, 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=4)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
+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=1)
-compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=7)
+compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=7)
compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=1)
@@ -1376,14 +1089,14 @@ s[compute].bind(compute_i0_o_i_i1_o_i_fused_i2_o_i_fused_i3_o_i_fused, te.thread
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=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=112)
+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=112)
+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:
@@ -1401,742 +1114,429 @@ CUDA source code:
#define int64_t long long
#define uint64_t unsigned long long
#endif
-extern "C" __global__ void __launch_bounds__(112) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
- float conv2d_nchw[28];
- __shared__ float pad_temp_shared[504];
- __shared__ float kernel_shared[1536];
+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[7] = 0.000000e+00f;
- conv2d_nchw[14] = 0.000000e+00f;
- conv2d_nchw[21] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
- conv2d_nchw[8] = 0.000000e+00f;
- conv2d_nchw[15] = 0.000000e+00f;
- conv2d_nchw[22] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
- conv2d_nchw[9] = 0.000000e+00f;
- conv2d_nchw[16] = 0.000000e+00f;
- conv2d_nchw[23] = 0.000000e+00f;
conv2d_nchw[3] = 0.000000e+00f;
- conv2d_nchw[10] = 0.000000e+00f;
- conv2d_nchw[17] = 0.000000e+00f;
- conv2d_nchw[24] = 0.000000e+00f;
conv2d_nchw[4] = 0.000000e+00f;
- conv2d_nchw[11] = 0.000000e+00f;
- conv2d_nchw[18] = 0.000000e+00f;
- conv2d_nchw[25] = 0.000000e+00f;
conv2d_nchw[5] = 0.000000e+00f;
- conv2d_nchw[12] = 0.000000e+00f;
- conv2d_nchw[19] = 0.000000e+00f;
- conv2d_nchw[26] = 0.000000e+00f;
conv2d_nchw[6] = 0.000000e+00f;
+ conv2d_nchw[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;
- conv2d_nchw[20] = 0.000000e+00f;
- conv2d_nchw[27] = 0.000000e+00f;
for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
__syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 63) / 9) + ry_outer_outer)) && ((((((int)threadIdx.x) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + ((((int)threadIdx.x) / 9) * 7)) + (ry_outer_outer * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= ((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 49) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 112) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= ((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 35) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 224) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= ((((((int)threadIdx.x) + 21) % 63) / 9) + ry_outer_outer)) && (((((((int)threadIdx.x) + 21) % 63) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 3) % 9))) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 336) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0.000000e+00f);
- if (((int)threadIdx.x) < 56) {
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (((((int)threadIdx.x) + 7) / 9) + ry_outer_outer)) && ((((((int)threadIdx.x) + 7) / 9) + ry_outer_outer) < 8)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[(((((rc_outer_outer * 392) + (((((int)threadIdx.x) + 448) / 9) * 7)) + (ry_outer_outer * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] : 0.000000e+00f);
+ 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);
}
- kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 294912) + ((((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) + 112)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 112) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 224) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((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) + 560)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 560) / 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) + 672)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 129024)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 784) / 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)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((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) + 1008)] = kernel[(((((((((int)blockIdx.x) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 193536)];
- kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1120) / 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) + 1232)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1232) / 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) * 294912) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
- if (((int)threadIdx.x) < 80) {
- kernel_shared[(((int)threadIdx.x) + 1456)] = kernel[((((((((int)blockIdx.x) * 294912) + (((((int)threadIdx.x) + 1456) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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)] = 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[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 9)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[((((int)threadIdx.x) / 7) * 96)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 24)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 48)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 72)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 1)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 25)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 49)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 73)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 2)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 26)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 50)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 8)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 74)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 63)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 3)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 27)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 51)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 75)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 64)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 4)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 28)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 52)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 76)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 65)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 66)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 67)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 68)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 69)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 70)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 5)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 29)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 53)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 71)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 77)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 126)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 6)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 30)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 54)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 78)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 7)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 31)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 55)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 79)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 8)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 32)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 56)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 134)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 80)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 189)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 9)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 33)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 57)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 81)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 190)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 10)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 34)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 58)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 82)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 191)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 192)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 193)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 194)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 195)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 11)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 35)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 59)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 83)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 252)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 12)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 36)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 60)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 84)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 253)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 13)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 37)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 61)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 85)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 254)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 255)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 256)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 257)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 258)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 259)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 14)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 38)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 62)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 260)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 86)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 315)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 15)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 39)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 63)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 87)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 316)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 16)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 40)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 64)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 88)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 317)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 318)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 319)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 320)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 321)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 322)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 17)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 41)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 65)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 89)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 18)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 42)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 66)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 90)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 379)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 19)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 43)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 67)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 91)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 380)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 381)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 382)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 383)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 384)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 385)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 20)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 44)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 68)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 386)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 92)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 21)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 45)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 69)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 93)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 22)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 46)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 70)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 94)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[14] = (conv2d_nchw[14] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[21] = (conv2d_nchw[21] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[15] = (conv2d_nchw[15] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[22] = (conv2d_nchw[22] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[16] = (conv2d_nchw[16] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[23] = (conv2d_nchw[23] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[17] = (conv2d_nchw[17] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[24] = (conv2d_nchw[24] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[18] = (conv2d_nchw[18] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[25] = (conv2d_nchw[25] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[19] = (conv2d_nchw[19] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[26] = (conv2d_nchw[26] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 448)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 23)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 47)]));
- conv2d_nchw[20] = (conv2d_nchw[20] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 71)]));
- conv2d_nchw[27] = (conv2d_nchw[27] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 9) + 449)] * kernel_shared[(((((int)threadIdx.x) / 7) * 96) + 95)]));
+ 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)]));
}
}
- for (int i1_inner = 0; i1_inner < 4; ++i1_inner) {
+ for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[(((((((int)blockIdx.x) * 3136) + ((((int)threadIdx.x) / 7) * 196)) + (i1_inner * 49)) + ((((int)threadIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[(((((int)blockIdx.x) * 64) + ((((int)threadIdx.x) / 7) * 4)) + i1_inner)]), 0.000000e+00f);
+ 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);
}
}
}
@@ -2174,7 +1574,7 @@ In the example below we resume the status and do more 5 trials.</p>
Get devices for measurement successfully!
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 5 minutes 36.422 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 5 minutes 45.586 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 82cc1270a7..d16ec5391c 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -916,7 +916,7 @@ 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)
- 7.8718 7.8728 7.8788 7.8637 0.0062
+ 7.8863 7.8847 7.8956 7.8785 0.0071
</pre></div>
</div>
</div>
@@ -938,7 +938,7 @@ to learn how to use the RPC Tracker and RPC Server.
To use the RPC Tracker in auto-scheduler, replace the runner in <code class="code docutils literal notranslate"><span class="pre">TuningOptions</span></code>
with <a class="reference internal" href="../../reference/api/python/auto_scheduler.html#tvm.auto_scheduler.RPCRunner" title="tvm.auto_scheduler.RPCRunner"><code class="xref any py py-class docutils literal notranslate"><span class="pre">auto_scheduler.RPCRunner</span></code></a>.</p></li>
</ol>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 4.945 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 7.544 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 be16f3ab3d..e06fb96bb9 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -935,7 +935,7 @@ 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)
- 748.3044 749.7266 749.7805 745.4060 2.0496
+ 770.1284 770.5184 771.6733 768.1936 1.4471
</pre></div>
</div>
</div>
@@ -957,7 +957,7 @@ 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 36.756 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 39.667 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 b1f7738ea4..88cc986180 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -629,12 +629,12 @@ class Module:
@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(64):
- compute_1 = T.allocate([1024], "float32", "global")
- compute_2 = T.buffer_decl((1024,), data=compute_1)
- for i_outer_inner, nb_j_inner in T.grid(8, 2):
- for i_inner_init in range(4):
- cse_var_1: T.int32 = i_outer_inner * 128 + i_inner_init * 32 + nb_j_inner * 16
+ for i0_outer_i1_outer_fused in T.parallel(16):
+ compute_1 = T.allocate([4096], "float32", "global")
+ compute_2 = T.buffer_decl((4096,), data=compute_1)
+ for i_outer_inner, nb_j_inner in T.grid(2, 2):
+ for i_inner_init in range(64):
+ cse_var_1: T.int32 = i_outer_inner * 2048 + 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)
@@ -651,49 +651,49 @@ class Module:
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(cse_var_2, i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner, placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]), 4):
+ for elem_idx, i_inner in T.grid(T.let(cse_var_2, i0_outer_i1_outer_fused * 2 + nb_j_inner, placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]), 64):
cse_var_2 = T.var("int32")
placeholder_5 = T.buffer_decl((33,), "int32", data=placeholder_3.data)
cse_var_21: T.int32 = elem_idx * 16
- cse_var_20: T.int32 = i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner
- cse_var_19: T.int32 = i_outer_inner * 128 + i_inner * 32 + nb_j_inner * 16
- cse_var_18: T.int32 = i0_outer_i1_outer_fused // 16 * 8192 + i_outer_inner * 1024 + i_inner * 256
- cse_var_17: T.int32 = cse_var_19 + 9
- cse_var_16: T.int32 = cse_var_19 + 8
- cse_var_15: T.int32 = cse_var_19 + 7
- cse_var_14: T.int32 = cse_var_19 + 6
- cse_var_13: T.int32 = cse_var_19 + 5
- cse_var_12: T.int32 = cse_var_19 + 4
- cse_var_11: T.int32 = cse_var_19 + 3
- cse_var_10: T.int32 = cse_var_19 + 2
- cse_var_9: T.int32 = cse_var_19 + 15
- cse_var_8: T.int32 = cse_var_19 + 14
- cse_var_7: T.int32 = cse_var_19 + 13
- cse_var_6: T.int32 = cse_var_19 + 12
- cse_var_5: T.int32 = cse_var_19 + 11
- cse_var_4: T.int32 = cse_var_19 + 10
- cse_var_3: T.int32 = cse_var_19 + 1
+ cse_var_20: T.int32 = i0_outer_i1_outer_fused * 2 + nb_j_inner
+ cse_var_19: T.int32 = i_outer_inner * 16384 + i_inner * 256
+ cse_var_18: T.int32 = i_outer_inner * 2048 + 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_decl((78656,), data=placeholder_1.data)
placeholder_7 = T.buffer_decl((32768,), data=placeholder.data)
placeholder_8 = T.buffer_decl((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_19] = compute_2[cse_var_19] + placeholder_6[placeholder_5[cse_var_20] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + 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_18 + placeholder_8[placeholder_5[cse_var_20] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(32, 32):
- cse_var_22: T.int32 = i0_outer_i1_outer_fused // 16 * 16384 + i0_inner * 512 + i0_outer_i1_outer_fused % 16 * 32 + i1_inner
+ 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, i1_inner in T.grid(128, 32):
+ cse_var_22: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 32 + i1_inner
compute_3 = T.buffer_decl((65536,), data=compute.data)
placeholder_5 = T.buffer_decl((65536,), data=placeholder_4.data)
compute_3[cse_var_22] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_22], T.float32(0))
@@ -730,7 +730,7 @@ 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: 2.162 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.871 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 638a9c6062..d68d2c3855 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-tune-with-autotvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:37.419</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:43.516</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -349,7 +349,7 @@
</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>00:37.387</p></td>
+<td><p>00:43.483</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>
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 25e69f85b8..86703a98a7 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -690,7 +690,7 @@ Traceback (most recent call last):
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, 8, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6120432
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 32, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1521344
No: 2 GFLOPS: 0.00/0.00 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)
@@ -813,7 +813,7 @@ Traceback (most recent call last):
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, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1615316
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3305364
No: 3 GFLOPS: 0.00/0.00 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)
@@ -936,9 +936,8 @@ Traceback (most recent call last):
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, 1, 64]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3006723
-No: 4 GFLOPS: 5.81/5.81 result: MeasureResult(costs=(0.03987096,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.805912494659424, timestamp=1674235925.9045806) [('tile_f', [-1, 1, 4, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,462395
-No: 5 GFLOPS: 0.00/5.81 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, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9665629
+No: 4 GFLOPS: 0.00/0.00 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
@@ -1060,10 +1059,11 @@ Traceback (most recent call last):
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, 8, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1128340
-No: 6 GFLOPS: 140.68/140.68 result: MeasureResult(costs=(0.001645544046875,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.708493709564209, timestamp=1674235929.6998758) [('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7652764
-No: 7 GFLOPS: 19.51/140.68 result: MeasureResult(costs=(0.011865912444444444,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2616629600524902, timestamp=1674235930.496339) [('tile_f', [-1, 2, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3175404
-No: 8 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 1, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4624679
+No: 5 GFLOPS: 33.86/33.86 result: MeasureResult(costs=(0.0068367294,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.5034289360046387, timestamp=1674258295.3208296) [('tile_f', [-1, 1, 1, 16]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3721464
+No: 6 GFLOPS: 5.70/33.86 result: MeasureResult(costs=(0.0406320075,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.8984193801879883, timestamp=1674258297.2149966) [('tile_f', [-1, 16, 16, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,196769
+No: 7 GFLOPS: 863.54/863.54 result: MeasureResult(costs=(0.0002680832871794872,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4097778797149658, timestamp=1674258297.9234798) [('tile_f', [-1, 2, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2561305
+No: 8 GFLOPS: 0.00/863.54 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
@@ -1185,8 +1185,8 @@ Traceback (most recent call last):
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, 16, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9845037
-No: 9 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 4, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9481192
+No: 9 GFLOPS: 0.00/863.54 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
@@ -1308,8 +1308,8 @@ Traceback (most recent call last):
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, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10214490
-No: 10 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7869550
+No: 10 GFLOPS: 0.00/863.54 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
@@ -1431,8 +1431,8 @@ Traceback (most recent call last):
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, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2817901
-No: 11 GFLOPS: 0.00/140.68 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, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6129419
+No: 11 GFLOPS: 0.00/863.54 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
@@ -1554,8 +1554,10 @@ Traceback (most recent call last):
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, 2, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,5064326
-No: 12 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8404785
+No: 12 GFLOPS: 15.10/863.54 result: MeasureResult(costs=(0.015329305,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.648505449295044, timestamp=1674258301.7926817) [('tile_f', [-1, 1, 8, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8281927
+No: 13 GFLOPS: 8.30/863.54 result: MeasureResult(costs=(0.027881015749999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=7.8216047286987305, timestamp=1674258312.876017) [('tile_f', [-1, 16, 4, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4187776
+No: 14 GFLOPS: 0.00/863.54 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
@@ -1677,8 +1679,8 @@ Traceback (most recent call last):
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, 64, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1655443
-No: 13 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 8, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 512, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3519299
+No: 15 GFLOPS: 0.00/863.54 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
@@ -1800,8 +1802,26 @@ Traceback (most recent call last):
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, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8054940
-No: 14 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 16]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9843634
+No: 16 GFLOPS: 0.00/863.54 result: Traceback (most recent call last):
+ File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 142, in build
+ res = future.result()
+ File "/usr/lib/python3.7/concurrent/futures/_base.py", line 435, in result
+ return self.__get_result()
+ File "/usr/lib/python3.7/concurrent/futures/_base.py", line 384, in __get_result
+ raise self._exception
+ File "/usr/lib/python3.7/concurrent/futures/thread.py", line 57, in run
+ result = self.fn(*self.args, **self.kwargs)
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 432, in <lambda>
+ worker = lambda *args: self._worker_run(*args)
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 401, in _worker_run
+ return proc.recv()
+ File "/workspace/python/tvm/contrib/popen_pool.py", line 309, in recv
+ raise TimeoutError()
+TimeoutError
+
+ [('tile_f', [-1, 16, 1, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9981239
+No: 17 GFLOPS: 0.00/863.54 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
@@ -1923,8 +1943,10 @@ Traceback (most recent call last):
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, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3081024
-No: 15 GFLOPS: 0.00/140.68 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 8, 8]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3405315
+No: 18 GFLOPS: 5.67/863.54 result: MeasureResult(costs=(0.04086507025,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.664371490478516, timestamp=1674258317.8645864) [('tile_f', [-1, 8, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 8, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2210407
+No: 19 GFLOPS: 262.79/863.54 result: MeasureResult(costs=(0.0008809437967032968,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7555351257324219, timestamp=1674258318.8751545) [('tile_f', [-1, 1, 32, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2139540
+No: 20 GFLOPS: 0.00/863.54 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
@@ -2046,256 +2068,7 @@ Traceback (most recent call last):
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, 32, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7239012
-No: 16 GFLOPS: 0.00/140.68 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_host=task.target_host, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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, 32, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9854529
-No: 17 GFLOPS: 6.77/140.68 result: MeasureResult(costs=(0.034183689,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.969929218292236, timestamp=1674235940.0712135) [('tile_f', [-1, 32, 1, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9364580
-No: 18 GFLOPS: 1.88/140.68 result: MeasureResult(costs=(0.1230732345,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.94360613822937, timestamp=1674235942.1101472) [('tile_f', [-1, 1, 1, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1990525
-No: 19 GFLOPS: 0.00/140.68 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_host=task.target_host, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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:1730
- 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:1670
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1630
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1630
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1645
- 13: operator()
- at ../src/driver/driver_api.cc:395
- 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:381
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:276
- 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:1749
- 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:1693
- 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:1617
- 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, 512, 1, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1219909
-No: 20 GFLOPS: 25.41/140.68 result: MeasureResult(costs=(0.009110510363636365,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.347748041152954, timestamp=1674235942.8749223) [('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2381511
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6594745
</pre></div>
</div>
<p>Finally we can inspect the best config from log file, check correctness,
@@ -2334,9 +2107,9 @@ 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:
-[('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7652764
+[('tile_f', [-1, 2, 2, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2561305
Finish loading 20 records
-Time cost of this operator: 0.001894
+Time cost of this operator: 0.000642
</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 2341097274..03c5b992ff 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -647,10 +647,10 @@ 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 312.9 98.734 (1, 2, 10, 10, 3) 2 1 [312.9]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.021 0.953 (1, 6, 10, 10) 1 1 [3.021]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.993 0.313 (1, 1, 10, 10, 3) 1 1 [0.993]
-Total_time - 316.913 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 312.5 98.716 (1, 2, 10, 10, 3) 2 1 [312.5]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.041 0.961 (1, 6, 10, 10) 1 1 [3.041]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 1.023 0.323 (1, 1, 10, 10, 3) 1 1 [1.023]
+Total_time - 316.563 - - - - -
</pre></div>
</div>
</div>
@@ -702,10 +702,10 @@ Total_time -
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 174.0 97.875 (1, 1, 10, 10, 6) 2 1 [174.0]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 2.932 1.649 (1, 6, 10, 10) 1 1 [2.932]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.847 0.476 (1, 3, 10, 10, 1) 1 1 [0.847]
-Total_time - 177.779 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 103.3 97.478 (1, 6, 10, 10, 1) 2 1 [103.3]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.805 1.704 (1, 6, 10, 10) 1 1 [1.805]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.867 0.818 (1, 3, 10, 10, 1) 1 1 [0.867]
+Total_time - 105.973 - - - - -
</pre></div>
</div>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-autotune-py">
diff --git a/docs/how_to/work_with_microtvm/micro_pytorch.html b/docs/how_to/work_with_microtvm/micro_pytorch.html
index 18c59d20b5..3a71ca60a5 100644
--- a/docs/how_to/work_with_microtvm/micro_pytorch.html
+++ b/docs/how_to/work_with_microtvm/micro_pytorch.html
@@ -454,7 +454,7 @@ download a cat image and preprocess it to use as the model input.</p>
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, 65.9MB/s]
+100%|##########| 3.42M/3.42M [00:00<00:00, 53.1MB/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.
@@ -578,7 +578,7 @@ 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 7.743 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 14.000 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 351f5816da..ff3703fb32 100644
--- a/docs/how_to/work_with_microtvm/micro_train.html
+++ b/docs/how_to/work_with_microtvm/micro_train.html
@@ -524,7 +524,7 @@ 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">"</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>'/tmp/tmplu1ennu6/images/random'
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>'/tmp/tmpld9y9_kq/images/random'
</pre></div>
</div>
</div>
@@ -584,8 +584,8 @@ 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">"off"</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], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [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/tmplu1ennu6/images/target contains 8144 images
-/tmp/tmplu1ennu6/images/random contains 5000 images
+<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmpld9y9_kq/images/target contains 8144 images
+/tmp/tmpld9y9_kq/images/random contains 5000 images
</pre></div>
</div>
</div>
@@ -697,13 +697,13 @@ 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 - 46s - loss: 0.2033 - accuracy: 0.9324 - val_loss: 0.1294 - val_accuracy: 0.9619 - 46s/epoch - 141ms/step
+328/328 - 49s - loss: 0.2060 - accuracy: 0.9244 - val_loss: 0.1419 - val_accuracy: 0.9445 - 49s/epoch - 148ms/step
Epoch 2/3
-328/328 - 43s - loss: 0.0927 - accuracy: 0.9649 - val_loss: 0.1044 - val_accuracy: 0.9634 - 43s/epoch - 131ms/step
+328/328 - 45s - loss: 0.0996 - accuracy: 0.9644 - val_loss: 0.1011 - val_accuracy: 0.9626 - 45s/epoch - 137ms/step
Epoch 3/3
-328/328 - 43s - loss: 0.0637 - accuracy: 0.9760 - val_loss: 0.1855 - val_accuracy: 0.9471 - 43s/epoch - 130ms/step
+328/328 - 44s - loss: 0.0699 - accuracy: 0.9732 - val_loss: 0.0904 - val_accuracy: 0.9683 - 44s/epoch - 136ms/step
-<keras.callbacks.History object at 0x7fdf35d99790>
+<keras.callbacks.History object at 0x7f450fb7dd90>
</pre></div>
</div>
</div>
@@ -963,7 +963,7 @@ 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 15.356 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes 24.153 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 0c6967b829..e6bc05ddd9 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -340,32 +340,32 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-microtvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>06:25.642</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>06:48.403</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
-<col style="width: 83%" />
-<col style="width: 10%" />
+<col style="width: 82%" />
+<col style="width: 11%" />
<col style="width: 7%" />
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_train.html#sphx-glr-how-to-work-with-microtvm-micro-train-py"><span class="std std-ref">Training Vision Models for microTVM on Arduino</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_train.py</span></code>)</p></td>
-<td><p>04:15.356</p></td>
+<td><p>04:24.153</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="micro_pytorch.html#sphx-glr-how-to-work-with-microtvm-micro-pytorch-py"><span class="std std-ref">microTVM PyTorch Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_pytorch.py</span></code>)</p></td>
-<td><p>01:07.743</p></td>
+<td><p>01:13.1000</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">Autotuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></td>
-<td><p>00:50.179</p></td>
+<td><p>00:56.625</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="micro_aot.html#sphx-glr-how-to-work-with-microtvm-micro-aot-py"><span class="std std-ref">microTVM Host-Driven AoT</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_aot.py</span></code>)</p></td>
-<td><p>00:08.657</p></td>
+<td><p>00:09.434</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">microTVM with TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></td>
-<td><p>00:03.706</p></td>
+<td><p>00:04.192</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">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 0a0150fedd..196da359b0 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-relay-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:43.083</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:44.304</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -349,15 +349,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="using_pipeline_executor.html#sphx-glr-how-to-work-with-relay-using-pipeline-executor-py"><span class="std std-ref">Using Pipeline Executor in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_pipeline_executor.py</span></code>)</p></td>
-<td><p>00:31.505</p></td>
+<td><p>00:32.574</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>00:10.043</p></td>
+<td><p>00:10.224</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>00:01.529</p></td>
+<td><p>00:01.501</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 2d17dc4a05..a51db71494 100644
--- a/docs/how_to/work_with_schedules/intrin_math.html
+++ b/docs/how_to/work_with_schedules/intrin_math.html
@@ -535,7 +535,7 @@ The following example customizes CUDA lowering rule for <code class="code docuti
<a href="../../reference/api/python/ir.html#tvm.ir.register_intrin_lowering" title="tvm.ir.register_intrin_lowering" class="sphx-glr-backref-module-tvm-ir sphx-glr-backref-type-py-function"><span class="n">register_intrin_lowering</span></a><span class="p">(</span><span class="s2">"tir.exp"</span><span class="p">,</span> <span class="n">target</span><span class="o">=</span><span class="s2">"cuda"</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><function my_cuda_math_rule at 0x7fdf35a45b90>
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span><function my_cuda_math_rule at 0x7f442e5ca830>
</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 5588c76c23..176f5b9f4c 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -340,7 +340,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-schedules-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:04.512</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:06.069</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -349,27 +349,27 @@
</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>00:02.191</p></td>
+<td><p>00:03.604</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>00:01.054</p></td>
+<td><p>00:01.094</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>00:00.531</p></td>
+<td><p>00:00.580</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>00:00.516</p></td>
+<td><p>00:00.569</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>00:00.116</p></td>
+<td><p>00:00.118</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>
-<td><p>00:00.049</p></td>
+<td><p>00:00.050</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tedd.html#sphx-glr-how-to-work-with-schedules-tedd-py"><span class="std std-ref">Use Tensor Expression Debug Display (TEDD) for Visualization</span></a> (<code class="docutils literal notranslate"><span class="pre">tedd.py</span></code>)</p></td>
diff --git a/docs/how_to/work_with_schedules/tensorize.html b/docs/how_to/work_with_schedules/tensorize.html
index be8c32a12d..d7bb1fe615 100644
--- a/docs/how_to/work_with_schedules/tensorize.html
+++ b/docs/how_to/work_with_schedules/tensorize.html
@@ -574,7 +574,7 @@ class Module:
def main(A: T.Buffer((1024, 64), "float32"), B: T.Buffer((512, 64), "float32"), C: T.Buffer((1024, 512), "float32")):
T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
i = T.var("int32")
- T.attr(T.iter_var(i, None, "DataPar", ""), "pragma_import_llvm", "; ModuleID = '/tmp/tmphrpach3t/input0.cc'\nsource_filename = \"/tmp/tmphrpach3t/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca [...]
+ T.attr(T.iter_var(i, None, "DataPar", ""), "pragma_import_llvm", "; ModuleID = '/tmp/tmp5ienotj_/input0.cc'\nsource_filename = \"/tmp/tmp5ienotj_/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca [...]
for i, j_outer in T.grid(1024, 32):
T.call_extern("int32", "gemv_update", T.tvm_access_ptr(T.type_annotation("float32"), C.data, i * 512 + j_outer * 16, 16, 2), T.tvm_access_ptr(T.type_annotation("float32"), A.data, i * 64, 64, 1), T.tvm_access_ptr(T.type_annotation("float32"), B.data, j_outer * 1024, 1024, 1), 16, 64, 64)
</pre></div>
diff --git a/docs/reference/api/python/auto_scheduler.html b/docs/reference/api/python/auto_scheduler.html
index b47aca41f4..a44b439810 100644
--- a/docs/reference/api/python/auto_scheduler.html
+++ b/docs/reference/api/python/auto_scheduler.html
@@ -1615,7 +1615,7 @@ history states as starting point to perform Evolutionary Search).</p></li>
<dl class="py class">
<dt class="sig sig-object py" id="tvm.auto_scheduler.SketchPolicy">
-<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
+<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
<dd><p>The search policy that searches in a hierarchical search space defined by sketches.
The policy randomly samples programs from the space defined by sketches and use evolutionary
search to fine-tune them.</p>
@@ -1899,7 +1899,7 @@ Candidates:
<dl class="py function">
<dt class="sig sig-object py" id="tvm.auto_scheduler.auto_schedule">
-<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
+<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
<dd><p>THIS API IS DEPRECATED.</p>
<p>Run auto scheduling search for a task.</p>
<dl class="field-list simple">
diff --git a/docs/reference/api/typedoc/classes/bytestreamreader.html b/docs/reference/api/typedoc/classes/bytestreamreader.html
index e15e19fc61..bd0de210c8 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/c8973a550/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -141,7 +141,7 @@
<div class="tsd-signature tsd-kind-icon">bytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Uint8Array</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L43">rpc_server.ts:43</a></li>
</ul>
</aside>
</section>
@@ -151,7 +151,7 @@
<div class="tsd-signature tsd-kind-icon">offset<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span><span class="tsd-signature-symbol"> = 0</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L42">rpc_server.ts:42</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L42">rpc_server.ts:42</a></li>
</ul>
</aside>
</section>
@@ -168,7 +168,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L63">rpc_server.ts:63</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L63">rpc_server.ts:63</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">Uint8Array</span></h4>
@@ -185,7 +185,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L49">rpc_server.ts:49</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L49">rpc_server.ts:49</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -202,7 +202,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L57">rpc_server.ts:57</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L57">rpc_server.ts:57</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
diff --git a/docs/reference/api/typedoc/classes/cachedcallstack.html b/docs/reference/api/typedoc/classes/cachedcallstack.html
index 9d747c3faf..eb399cfc47 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/c8973a550/web/src/memory.ts#L223">memory.ts:223</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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"><</span><a href="../interfaces/disposable.html" class="tsd-signature-type">Disposable</a><span class="tsd-signature-symbol">></span><span class="tsd-signature-symbol"> = []</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/memory.ts#L208">memory.ts:208</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L312">memory.ts:312</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L284">memory.ts:284</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L388">memory.ts:388</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L376">memory.ts:376</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L267">memory.ts:267</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L243">memory.ts:243</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L321">memory.ts:321</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L252">memory.ts:252</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L359">memory.ts:359</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L342">memory.ts:342</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L350">memory.ts:350</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L326">memory.ts:326</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L363">memory.ts:363</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L346">memory.ts:346</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L334">memory.ts:334</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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 599eb6f5e0..230d3d1d28 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/c8973a550/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L262">runtime.ts:262</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -147,7 +147,7 @@
<div class="tsd-signature tsd-kind-icon">bits<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L260">runtime.ts:260</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L260">runtime.ts:260</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -162,7 +162,7 @@
<div class="tsd-signature tsd-kind-icon">code<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L258">runtime.ts:258</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L258">runtime.ts:258</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -177,7 +177,7 @@
<div class="tsd-signature tsd-kind-icon">lanes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L262">runtime.ts:262</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L262">runtime.ts:262</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -199,7 +199,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L279">runtime.ts:279</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L279">runtime.ts:279</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -216,7 +216,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L270">runtime.ts:270</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L270">runtime.ts:270</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/dldevice.html b/docs/reference/api/typedoc/classes/dldevice.html
index 3a9e26a330..b99f8faab9 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/c8973a550/web/src/runtime.ts#L202">runtime.ts:202</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L202">runtime.ts:202</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -146,7 +146,7 @@
<div class="tsd-signature tsd-kind-icon">device<wbr>Id<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L200">runtime.ts:200</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L200">runtime.ts:200</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -161,7 +161,7 @@
<div class="tsd-signature tsd-kind-icon">device<wbr>Type<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L198">runtime.ts:198</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L198">runtime.ts:198</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -183,7 +183,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L223">runtime.ts:223</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L223">runtime.ts:223</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -205,7 +205,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L230">runtime.ts:230</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L230">runtime.ts:230</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/environment.html b/docs/reference/api/typedoc/classes/environment.html
index 82f938b080..9cba413592 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/c8973a550/web/src/environment.ts#L86">environment.ts:86</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/environment.ts#L70">environment.ts:70</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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"> => </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/c8973a550/web/src/environment.ts#L69">environment.ts:69</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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"><</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">></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/c8973a550/web/src/environment.ts#L78">environment.ts:78</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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"><</span><span class="tsd-signature-type">number</span><span class="tsd-signature-symbol">></span><span class="tsd-signature-symbol"> = []</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/environment.ts#L84">environment.ts:84</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/environment.ts#L105">environment.ts:105</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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 4a5055be0f..6c88f0a20b 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/c8973a550/web/src/runtime.ts#L49">runtime.ts:49</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L49">runtime.ts:49</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -156,7 +156,7 @@
<div class="tsd-signature tsd-kind-icon">exports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol"><</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">></span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L46">runtime.ts:46</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L46">runtime.ts:46</a></li>
</ul>
</aside>
</section>
@@ -166,7 +166,7 @@
<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <a href="memory.html" class="tsd-signature-type">Memory</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L45">runtime.ts:45</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L45">runtime.ts:45</a></li>
</ul>
</aside>
</section>
@@ -176,7 +176,7 @@
<div class="tsd-signature tsd-kind-icon">wasm32<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">boolean</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L44">runtime.ts:44</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L44">runtime.ts:44</a></li>
</ul>
</aside>
</section>
@@ -186,7 +186,7 @@
<div class="tsd-signature tsd-kind-icon">webGPUContext<span class="tsd-signature-symbol">:</span> <a href="webgpucontext.html" class="tsd-signature-type">WebGPUContext</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L47">runtime.ts:47</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L47">runtime.ts:47</a></li>
</ul>
</aside>
</section>
@@ -203,7 +203,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L76">runtime.ts:76</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L76">runtime.ts:76</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -226,7 +226,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L66">runtime.ts:66</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L66">runtime.ts:66</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -243,7 +243,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L84">runtime.ts:84</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L84">runtime.ts:84</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <a href="cachedcallstack.html" class="tsd-signature-type">CachedCallStack</a></h4>
@@ -260,7 +260,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L95">runtime.ts:95</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L95">runtime.ts:95</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -283,7 +283,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L72">runtime.ts:72</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L72">runtime.ts:72</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
diff --git a/docs/reference/api/typedoc/classes/graphexecutor.html b/docs/reference/api/typedoc/classes/graphexecutor.html
index a2b04dedee..ac6ebb39cb 100644
--- a/docs/reference/api/typedoc/classes/graphexecutor.html
+++ b/docs/reference/api/typedoc/classes/graphexecutor.html
@@ -130,7 +130,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L583">runtime.ts:583</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L583">runtime.ts:583</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -162,7 +162,7 @@
<div class="tsd-signature tsd-kind-icon">module<span class="tsd-signature-symbol">:</span> <a href="module.html" class="tsd-signature-type">Module</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L579">runtime.ts:579</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L579">runtime.ts:579</a></li>
</ul>
</aside>
</section>
@@ -179,7 +179,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L654">runtime.ts:654</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L654">runtime.ts:654</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -224,7 +224,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L597">runtime.ts:597</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L597">runtime.ts:597</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -241,7 +241,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L631">runtime.ts:631</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L631">runtime.ts:631</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -279,7 +279,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L644">runtime.ts:644</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L644">runtime.ts:644</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -310,7 +310,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L621">runtime.ts:621</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L621">runtime.ts:621</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -332,7 +332,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L609">runtime.ts:609</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L609">runtime.ts:609</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/instance.html b/docs/reference/api/typedoc/classes/instance.html
index 482fe68243..e742d570a9 100644
--- a/docs/reference/api/typedoc/classes/instance.html
+++ b/docs/reference/api/typedoc/classes/instance.html
@@ -139,7 +139,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L692">runtime.ts:692</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L692">runtime.ts:692</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -202,7 +202,7 @@
<div class="tsd-signature tsd-kind-icon">exports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol"><</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">></span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L684">runtime.ts:684</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L684">runtime.ts:684</a></li>
</ul>
</aside>
</section>
@@ -212,7 +212,7 @@
<div class="tsd-signature tsd-kind-icon">memory<span class="tsd-signature-symbol">:</span> <a href="memory.html" class="tsd-signature-type">Memory</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L683">runtime.ts:683</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L683">runtime.ts:683</a></li>
</ul>
</aside>
</section>
@@ -229,7 +229,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L932">runtime.ts:932</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L932">runtime.ts:932</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -260,7 +260,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L994">runtime.ts:994</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L994">runtime.ts:994</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -303,7 +303,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L924">runtime.ts:924</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L924">runtime.ts:924</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -341,7 +341,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L732">runtime.ts:732</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L732">runtime.ts:732</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -358,7 +358,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L952">runtime.ts:952</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L952">runtime.ts:952</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -402,7 +402,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L816">runtime.ts:816</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L816">runtime.ts:816</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -434,7 +434,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L1033">runtime.ts:1033</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L1033">runtime.ts:1033</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -465,7 +465,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L846">runtime.ts:846</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L846">runtime.ts:846</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -497,7 +497,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L750">runtime.ts:750</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L750">runtime.ts:750</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -520,7 +520,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L1013">runtime.ts:1013</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L1013">runtime.ts:1013</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -568,7 +568,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L789">runtime.ts:789</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L789">runtime.ts:789</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -608,7 +608,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L914">runtime.ts:914</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L914">runtime.ts:914</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -646,7 +646,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L1145">runtime.ts:1145</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L1145">runtime.ts:1145</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -698,7 +698,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L740">runtime.ts:740</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L740">runtime.ts:740</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -722,7 +722,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L868">runtime.ts:868</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L868">runtime.ts:868</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -754,7 +754,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L857">runtime.ts:857</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L857">runtime.ts:857</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -786,7 +786,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L940">runtime.ts:940</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L940">runtime.ts:940</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/memory.html b/docs/reference/api/typedoc/classes/memory.html
index 725378d9c7..4ef1276739 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/c8973a550/web/src/memory.ts#L40">memory.ts:40</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L32">memory.ts:32</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L33">memory.ts:33</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L154">memory.ts:154</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L90">memory.ts:90</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L97">memory.ts:97</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L74">memory.ts:74</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L81">memory.ts:81</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L104">memory.ts:104</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L132">memory.ts:132</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L145">memory.ts:145</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L60">memory.ts:60</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L67">memory.ts:67</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L53">memory.ts:53</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L114">memory.ts:114</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L124">memory.ts:124</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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/c8973a550/web/src/memory.ts#L175">memory.ts:175</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/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 3b55b44bed..d9f5e97ba0 100644
--- a/docs/reference/api/typedoc/classes/module.html
+++ b/docs/reference/api/typedoc/classes/module.html
@@ -124,7 +124,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L504">runtime.ts:504</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L504">runtime.ts:504</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -170,7 +170,7 @@
<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L502">runtime.ts:502</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L502">runtime.ts:502</a></li>
</ul>
</aside>
</section>
@@ -187,7 +187,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L516">runtime.ts:516</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L516">runtime.ts:516</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -204,7 +204,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L530">runtime.ts:530</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L530">runtime.ts:530</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -236,7 +236,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L561">runtime.ts:561</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L561">runtime.ts:561</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/ndarray.html b/docs/reference/api/typedoc/classes/ndarray.html
index 5e993ebefd..e2e02a6c1c 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/c8973a550/web/src/runtime.ts#L304">runtime.ts:304</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L304">runtime.ts:304</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -158,7 +158,7 @@
<div class="tsd-signature tsd-kind-icon">device<span class="tsd-signature-symbol">:</span> <a href="dldevice.html" class="tsd-signature-type">DLDevice</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L297">runtime.ts:297</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L297">runtime.ts:297</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -173,7 +173,7 @@
<div class="tsd-signature tsd-kind-icon">dtype<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L293">runtime.ts:293</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L293">runtime.ts:293</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -188,7 +188,7 @@
<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L289">runtime.ts:289</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L289">runtime.ts:289</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -203,7 +203,7 @@
<div class="tsd-signature tsd-kind-icon">ndim<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L291">runtime.ts:291</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L291">runtime.ts:291</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -218,7 +218,7 @@
<div class="tsd-signature tsd-kind-icon">shape<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Array</span><span class="tsd-signature-symbol"><</span><span class="tsd-signature-type">number</span><span class="tsd-signature-symbol">></span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L295">runtime.ts:295</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L295">runtime.ts:295</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -240,7 +240,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L370">runtime.ts:370</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L370">runtime.ts:370</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -273,7 +273,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L414">runtime.ts:414</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L414">runtime.ts:414</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -305,7 +305,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L355">runtime.ts:355</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L355">runtime.ts:355</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
@@ -322,7 +322,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L474">runtime.ts:474</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L474">runtime.ts:474</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -346,7 +346,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L443">runtime.ts:443</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L443">runtime.ts:443</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
diff --git a/docs/reference/api/typedoc/classes/packedfunccell.html b/docs/reference/api/typedoc/classes/packedfunccell.html
index 165a9dbd6e..aca9b1a162 100644
--- a/docs/reference/api/typedoc/classes/packedfunccell.html
+++ b/docs/reference/api/typedoc/classes/packedfunccell.html
@@ -122,7 +122,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L158">runtime.ts:158</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L158">runtime.ts:158</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -147,7 +147,7 @@
<div class="tsd-signature tsd-kind-icon">handle<span class="tsd-signature-symbol">:</span> <a href="../index.html#pointer" class="tsd-signature-type">Pointer</a></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L157">runtime.ts:157</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L157">runtime.ts:157</a></li>
</ul>
</aside>
</section>
@@ -164,7 +164,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/runtime.ts#L165">runtime.ts:165</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/runtime.ts#L165">runtime.ts:165</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">void</span></h4>
diff --git a/docs/reference/api/typedoc/classes/rpcserver.html b/docs/reference/api/typedoc/classes/rpcserver.html
index 097543f60d..a169b5601c 100644
--- a/docs/reference/api/typedoc/classes/rpcserver.html
+++ b/docs/reference/api/typedoc/classes/rpcserver.html
@@ -115,7 +115,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L92">rpc_server.ts:92</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L92">rpc_server.ts:92</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -176,7 +176,7 @@
<div class="tsd-signature tsd-kind-icon">get<wbr>Imports<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> => </span><span class="tsd-signature-type">Record</span><span class="tsd-signature-symbol"><</span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">, </span><span class="tsd-signature-type">unknown</span><span class="tsd-signat [...]
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L82">rpc_server.ts:82</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L82">rpc_server.ts:82</a></li>
</ul>
</aside>
<div class="tsd-type-declaration">
@@ -201,7 +201,7 @@
<div class="tsd-signature tsd-kind-icon">key<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">string</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/c8973a550/web/src/rpc_server.ts#L78">rpc_server.ts:78</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L78">rpc_server.ts:78</a></li>
</ul>
</aside>
</section>
@@ -211,7 +211,7 @@
<div class="tsd-signature tsd-kind-icon">logger<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-symbol">(</span>msg<span class="tsd-signature-symbol">: </span><span class="tsd-signature-type">string</span><span class="tsd-signature-symbol">)</span><span class="tsd-signature-symbol"> => </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/c8973a550/web/src/rpc_server.ts#L81">rpc_server.ts:81</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/fe372a730/web/src/rpc_server.ts#L81">rpc_server.ts:81</a></li>
</ul>
</aside>
... 1815 lines suppressed ...