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 2023/04/04 05:05:31 UTC
[tvm-site] branch asf-site updated: deploying docs (apache/tvm@99a5734a9ec25d5fc263fd1b61fc2687c83a2f92)
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 8cfddcd89b deploying docs (apache/tvm@99a5734a9ec25d5fc263fd1b61fc2687c83a2f92)
8cfddcd89b is described below
commit 8cfddcd89b1b359664470544c0f0285341bb355a
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Tue Apr 4 05:05:24 2023 +0000
deploying docs (apache/tvm@99a5734a9ec25d5fc263fd1b61fc2687c83a2f92)
---
docs/_images/sphx_glr_micro_train_001.png | Bin 343660 -> 329926 bytes
docs/_images/sphx_glr_micro_train_thumb.png | Bin 23847 -> 23812 bytes
.../how_to/compile_models/from_darknet.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 | 2 +-
.../deploy_model_on_adreno_tvmc.rst.txt | 2 +-
.../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 | 48 +-
.../extend_tvm/bring_your_own_datatypes.rst.txt | 2 +-
.../how_to/extend_tvm/sg_execution_times.rst.txt | 10 +-
.../how_to/extend_tvm/use_pass_instrument.rst.txt | 16 +-
.../optimize_operators/opt_conv_cuda.rst.txt | 2 +-
.../optimize_operators/opt_conv_tensorcore.rst.txt | 2 +-
.../how_to/optimize_operators/opt_gemm.rst.txt | 16 +-
.../optimize_operators/sg_execution_times.rst.txt | 8 +-
.../sg_execution_times.rst.txt | 14 +-
.../tune_conv2d_layer_cuda.rst.txt | 1877 +++++++++-----------
.../tune_network_cuda.rst.txt | 4 +-
.../tune_network_x86.rst.txt | 4 +-
.../tune_sparse_x86.rst.txt | 103 +-
.../tune_with_autotvm/sg_execution_times.rst.txt | 6 +-
.../tune_with_autotvm/tune_conv2d_cuda.rst.txt | 811 +--------
.../work_with_microtvm/micro_autotune.rst.txt | 18 +-
.../work_with_microtvm/micro_pytorch.rst.txt | 4 +-
.../how_to/work_with_microtvm/micro_train.rst.txt | 18 +-
.../work_with_microtvm/sg_execution_times.rst.txt | 14 +-
.../work_with_relay/sg_execution_times.rst.txt | 8 +-
.../how_to/work_with_schedules/intrin_math.rst.txt | 2 +-
.../work_with_schedules/sg_execution_times.rst.txt | 12 +-
.../tutorials/autotvm/sg_execution_times.rst.txt | 4 +-
.../frontend/deploy_classification.rst.txt | 2 +-
.../tutorials/frontend/deploy_detection.rst.txt | 2 +-
.../tutorials/frontend/sg_execution_times.rst.txt | 6 +-
.../tutorials/optimize/sg_execution_times.rst.txt | 6 +-
.../topic/vta/tutorials/sg_execution_times.rst.txt | 6 +-
.../tutorial/auto_scheduler_matmul_x86.rst.txt | 13 +-
docs/_sources/tutorial/autotvm_matmul_x86.rst.txt | 20 +-
docs/_sources/tutorial/autotvm_relay_x86.rst.txt | 56 +-
.../tutorial/cross_compilation_and_rpc.rst.txt | 2 +-
docs/_sources/tutorial/intro_topi.rst.txt | 2 +-
docs/_sources/tutorial/sg_execution_times.rst.txt | 16 +-
.../tutorial/tensor_expr_get_started.rst.txt | 51 +-
docs/commit_hash | 2 +-
docs/genindex.html | 2 +
docs/how_to/compile_models/from_darknet.html | 2 +-
docs/how_to/compile_models/from_mxnet.html | 2 +-
docs/how_to/compile_models/from_oneflow.html | 14 +-
docs/how_to/compile_models/from_pytorch.html | 10 +-
docs/how_to/compile_models/from_tensorflow.html | 2 +-
docs/how_to/compile_models/sg_execution_times.html | 22 +-
.../deploy_models/deploy_model_on_adreno.html | 2 +-
.../deploy_models/deploy_model_on_adreno_tvmc.html | 43 +-
.../deploy_models/deploy_model_on_android.html | 2 +-
.../deploy_object_detection_pytorch.html | 62 +-
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 | 42 +-
docs/how_to/deploy_models/sg_execution_times.html | 26 +-
.../extend_tvm/bring_your_own_datatypes.html | 2 +-
docs/how_to/extend_tvm/sg_execution_times.html | 10 +-
docs/how_to/extend_tvm/use_pass_instrument.html | 16 +-
docs/how_to/optimize_operators/opt_conv_cuda.html | 2 +-
.../optimize_operators/opt_conv_tensorcore.html | 2 +-
docs/how_to/optimize_operators/opt_gemm.html | 16 +-
.../optimize_operators/sg_execution_times.html | 8 +-
.../sg_execution_times.html | 14 +-
.../tune_conv2d_layer_cuda.html | 1877 +++++++++-----------
.../tune_with_autoscheduler/tune_network_cuda.html | 4 +-
.../tune_with_autoscheduler/tune_network_x86.html | 4 +-
.../tune_with_autoscheduler/tune_sparse_x86.html | 103 +-
.../tune_with_autotvm/sg_execution_times.html | 6 +-
.../how_to/tune_with_autotvm/tune_conv2d_cuda.html | 811 +--------
docs/how_to/work_with_microtvm/micro_autotune.html | 18 +-
docs/how_to/work_with_microtvm/micro_pytorch.html | 6 +-
docs/how_to/work_with_microtvm/micro_train.html | 16 +-
.../work_with_microtvm/sg_execution_times.html | 14 +-
.../how_to/work_with_relay/sg_execution_times.html | 8 +-
docs/how_to/work_with_schedules/intrin_math.html | 2 +-
.../work_with_schedules/sg_execution_times.html | 12 +-
docs/objects.inv | Bin 24757 -> 24763 bytes
.../classtvm_1_1tir_1_1ScheduleNode-members.html | 109 +-
.../doxygen/classtvm_1_1tir_1_1ScheduleNode.html | 38 +
...lasstvm_1_1tir_1_1ScheduleNode__coll__graph.svg | 2 +-
...stvm_1_1tir_1_1ScheduleNode__inherit__graph.svg | 2 +-
docs/reference/api/doxygen/database_8h_source.html | 2 +-
docs/reference/api/doxygen/functions_func_m.html | 3 +-
docs/reference/api/doxygen/functions_m.html | 1 +
.../api/doxygen/measure__candidate_8h_source.html | 2 +-
docs/reference/api/doxygen/postproc_8h_source.html | 2 +-
.../api/doxygen/schedule__rule_8h_source.html | 2 +-
docs/reference/api/doxygen/search/all_e.js | 2 +-
docs/reference/api/doxygen/search/functions_d.js | 2 +-
.../doxygen/tir_2schedule_2schedule_8h_source.html | 5 +-
docs/reference/api/doxygen/trace_8h_source.html | 2 +-
docs/reference/api/python/auto_scheduler.html | 4 +-
docs/reference/api/python/tir.html | 145 +-
.../api/typedoc/classes/bytestreamreader.html | 12 +-
.../api/typedoc/classes/cachedcallstack.html | 34 +-
docs/reference/api/typedoc/classes/dldatatype.html | 12 +-
docs/reference/api/typedoc/classes/dldevice.html | 10 +-
.../reference/api/typedoc/classes/environment.html | 12 +-
docs/reference/api/typedoc/classes/ffilibrary.html | 20 +-
docs/reference/api/typedoc/classes/instance.html | 58 +-
docs/reference/api/typedoc/classes/memory.html | 34 +-
docs/reference/api/typedoc/classes/module.html | 10 +-
docs/reference/api/typedoc/classes/ndarray.html | 22 +-
.../api/typedoc/classes/packedfunccell.html | 6 +-
docs/reference/api/typedoc/classes/rpcserver.html | 14 +-
.../api/typedoc/classes/runtimecontext.html | 22 +-
docs/reference/api/typedoc/classes/scalar.html | 6 +-
docs/reference/api/typedoc/classes/tvmarray.html | 16 +-
docs/reference/api/typedoc/classes/tvmobject.html | 12 +-
.../api/typedoc/classes/webgpucontext.html | 12 +-
docs/reference/api/typedoc/enums/argtypecode.html | 30 +-
.../api/typedoc/enums/aynccallbackcode.html | 4 +-
.../api/typedoc/enums/dldatatypecode.html | 8 +-
.../api/typedoc/enums/rpcserverstate.html | 12 +-
docs/reference/api/typedoc/enums/sizeof.html | 18 +-
docs/reference/api/typedoc/index.html | 124 +-
.../api/typedoc/interfaces/disposable.html | 2 +-
.../api/typedoc/interfaces/functioninfo.html | 6 +-
.../api/typedoc/interfaces/libraryprovider.html | 4 +-
docs/searchindex.js | 2 +-
.../vta/tutorials/autotvm/sg_execution_times.html | 4 +-
.../tutorials/frontend/deploy_classification.html | 2 +-
.../vta/tutorials/frontend/deploy_detection.html | 2 +-
.../vta/tutorials/frontend/sg_execution_times.html | 10 +-
.../vta/tutorials/optimize/sg_execution_times.html | 6 +-
docs/topic/vta/tutorials/sg_execution_times.html | 6 +-
docs/tutorial/auto_scheduler_matmul_x86.html | 8 +-
docs/tutorial/autotvm_matmul_x86.html | 20 +-
docs/tutorial/autotvm_relay_x86.html | 264 +--
docs/tutorial/cross_compilation_and_rpc.html | 2 +-
docs/tutorial/intro_topi.html | 2 +-
docs/tutorial/sg_execution_times.html | 16 +-
docs/tutorial/tensor_expr_get_started.html | 47 +-
146 files changed, 3108 insertions(+), 4567 deletions(-)
diff --git a/docs/_images/sphx_glr_micro_train_001.png b/docs/_images/sphx_glr_micro_train_001.png
index 4f3246f529..cf6a64d19f 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 3ef5dcacc2..5378603442 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 d2ebcbb61b..17c4f13601 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 23.229 seconds)
+ **Total running time of the script:** ( 1 minutes 20.630 seconds)
.. _sphx_glr_download_how_to_compile_models_from_darknet.py:
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 ea8fd2201b..082244be3a 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.zip14e69f2c-37c2-46c8-9f51-31c44101d8e9 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+ Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zipfae97a33-2c77-494d-bc7a-8f359693ab8c 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 2a2a825a55..7bb14b6c60 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, 55.6MB/s]
35%|###4 | 14.3M/41.5M [00:00<00:00, 46.3MB/s]
51%|##### | 21.0M/41.5M [00:00<00:00, 54.5MB/s]
64%|######3 | 26.5M/41.5M [00:00<00:00, 41.3MB/s]
77%|#######7 | 32.0M/41.5M [00:00<00:00, 44.6MB/s]
92%|#########2| 38.3M/41.5M [00:00<00:00, 42.2MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 45.1MB/s]
+
0%| | 0.00/41.5M [00:00<?, ?B/s]
19%|#9 | 7.99M/41.5M [00:00<00:00, 51.3MB/s]
35%|###4 | 14.3M/41.5M [00:00<00:00, 58.3MB/s]
48%|####8 | 20.1M/41.5M [00:00<00:00, 49.8MB/s]
60%|###### | 25.0M/41.5M [00:00<00:00, 43.8MB/s]
77%|#######7 | 32.0M/41.5M [00:00<00:00, 44.9MB/s]
92%|#########2| 38.3M/41.5M [00:00<00:00, 50.1MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 48.0MB/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 a1ff6c6dc2..60ee96cf31 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 | 7.99M/44.7M [00:00<00:00, 65.8MB/s]
46%|####5 | 20.4M/44.7M [00:00<00:00, 99.7MB/s]
68%|######7 | 30.3M/44.7M [00:00<00:00, 57.6MB/s]
83%|########3 | 37.3M/44.7M [00:00<00:00, 50.6MB/s]
96%|#########6| 43.0M/44.7M [00:00<00:00, 48.6MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 49.0MB/s]
+
0%| | 0.00/44.7M [00:00<?, ?B/s]
5%|4 | 2.05M/44.7M [00:00<00:02, 18.5MB/s]
54%|#####3 | 24.0M/44.7M [00:00<00:00, 133MB/s]
83%|########3 | 37.2M/44.7M [00:00<00:00, 118MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 108MB/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 25858ea25c..36fe014c48 100644
--- a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
@@ -430,7 +430,7 @@ Run the corresponding model on tensorflow
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 31.647 seconds)
+ **Total running time of the script:** ( 1 minutes 31.229 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 13f3fe37b6..50fa304d1e 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
=================
-**06:56.667** total execution time for **how_to_compile_models** files:
+**06:49.351** total execution time for **how_to_compile_models** files:
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:31.647 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:31.229 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:23.229 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:20.630 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:57.563 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:56.683 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:38.247 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:38.067 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:32.859 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:32.357 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:30.582 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:30.190 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:28.678 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:28.589 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:26.444 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:26.151 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:24.681 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:22.759 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.738 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.696 | 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 375e6680b5..0fe56cb19c 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
@@ -637,7 +637,7 @@ 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)
- 2753.6753 2754.0597 2756.0955 2750.8269 1.7248
+ 2539.4006 2538.4581 2544.2771 2536.9132 2.1375
diff --git a/docs/_sources/how_to/deploy_models/deploy_model_on_adreno_tvmc.rst.txt b/docs/_sources/how_to/deploy_models/deploy_model_on_adreno_tvmc.rst.txt
index 7eae44e6b8..26d3d5fbc4 100644
--- a/docs/_sources/how_to/deploy_models/deploy_model_on_adreno_tvmc.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_model_on_adreno_tvmc.rst.txt
@@ -127,7 +127,7 @@ Make a Keras Resnet50 Model
.. code-block:: none
Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels.h5
-
8192/102967424 [..............................] - ETA: 0s
2187264/102967424 [..............................] - ETA: 3s
8380416/102967424 [=>............................] - ETA: 1s
10575872/102967424 [==>...........................] - ETA: 2s
16769024/102967424 [===>..........................] - ETA: 2s
25157632/102967424 [======>.......................] - ETA: 1s
32481280/102967424 [========>.....................] - ETA: 1s
33546240/102967424 [========>.....................] - ETA: 1s
40189952/102967424 [==========>...................] - ETA: 1s
41934848/102967424 [===========>..................] - ETA: 1s
48578560/102967424 [=============>................] - ETA: 1s
50323456/102967424 [=============>................] - ETA: 1s
56967168/102967424 [===============>..............] - ETA: 0s
58712064/102967424 [================>.............] - ETA: 0s
65355776/102967424 [==================>...........] - ETA: 0s
67100672/102967424 [==================>...........] -
ETA: 0s
69296128/102967424 [===================>..........] - ETA: 0s
75489280/102967424 [====================>.........] - ETA: 0s
82124800/102967424 [======================>.......] - ETA: 0s
83877888/102967424 [=======================>......] - ETA: 0s
90521600/102967424 [=========================>....] - ETA: 0s
92266496/102967424 [=========================>....] - ETA: 0s
95059968/102967424 [==========================>...] - ETA: 0s
98910208/102967424
[===========================>..] - ETA: 0s
100646912/102967424 [============================>.] - ETA: 0s
102850560/102967424 [============================>.] - ETA: 0s
102967424/102967424 [==============================] - 2s 0us/step
+
8192/102967424 [..............................] - ETA: 0s
11804672/102967424 [==>...........................] - ETA: 0s
16769024/102967424 [===>..........................] - ETA: 0s
23011328/102967424 [=====>........................] - ETA: 0s
28106752/102967424 [=======>......................] - ETA: 0s
33120256/102967424 [========>.....................] - ETA: 0s
38223872/102967424 [==========>...................] - ETA: 0s
43196416/102967424 [===========>..................] - ETA: 0s
48300032/102967424 [=============>................] - ETA: 0s
50987008/102967424 [=============>................] - ETA: 0s
58220544/102967424 [===============>..............] - ETA: 0s
63324160/102967424 [=================>............] - ETA: 0s
68296704/102967424 [==================>...........] - ETA: 0s
73334784/102967424 [====================>.........] - ETA: 0s
78381056/102967424 [=====================>........] - ETA: 0s
83329024/102967424 [=======================>......] -
ETA: 0s
88416256/102967424 [========================>.....] - ETA: 0s
92266496/102967424 [=========================>....] - ETA: 0s
99287040/102967424 [===========================>..] - ETA: 0s
102967424/102967424 [==============================] - 1s 0us/step
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 093df6e7be..9e2ab2acf5 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)
- 15.9861 15.8256 17.0500 15.6477 0.4107
+ 15.9646 15.7778 16.5501 15.6210 0.3490
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 2df96c0c66..88c014a118 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<00:04, 41.0MB/s]
8%|8 | 14.3M/170M [00:00<00:04, 38.0MB/s]
11%|# | 17.9M/170M [00:00<00:04, 36.5MB/s]
13%|#3 | 22.3M/170M [00:00<00:04, 35.6MB/s]
15%|#5 | 25.7M/170M [00:00<00:04, 35.1MB/s]
19%|#8 | 32.0M/170M [00:00<00:03, 36.7MB/s]
24%|##3 | 40.0M/170M [00:01<00:03, 44.9MB/s]
27%|##7 | 46.3M/170M [00:01<00:02, 45.5MB/s]
30%|##9 | 50.6M/170M [00:01<00:03, 35.4MB/s]
33%|###2 | 56.0M/170M [00:01<00:03, 37.6MB/s]
38%|###7 | 64.0M/170M [00:01<00:02, 41.2MB/s]
42%|####2 | 72.0M/170M [00:01<00:02, 46.3MB/s]
46%|####6 | 78.3M/170M [00:01<00:01, 49.5MB/s]
49%|####8 | 83.2M/170M [00:02<00:01, 46.2MB/s]
52%|#####1 | 87.7M/170M [00:02<00:01, 44.8MB/s]
54%|#####4 | 92.0M/170M [00:02<00:02, 39.0MB/s]
57%|#####6 | 96.0M/170M [00:02<00:02, 33.6MB/
s]
60%|###### | 102M/170M [00:02<00:01, 36.3MB/s]
62%|######2 | 106M/170M [00:02<00:01, 34.5MB/s]
66%|######5 | 112M/170M [00:02<00:01, 36.4MB/s]
71%|####### | 120M/170M [00:03<00:01, 43.8MB/s]
73%|#######3 | 124M/170M [00:03<00:01, 38.1MB/s]
75%|#######5 | 128M/170M [00:03<00:01, 35.7MB/s]
80%|######## | 136M/170M [00:03<00:00, 35.9MB/s]
85%|########4 | 144M/170M [00:03<00:00, 42.9MB/s]
89%|########8 | 151M/170M [00:03<00:00, 49.6MB/s]
92%|#########1| 156M/170M [00:04<00:00, 46.5MB/s]
95%|#########4| 161M/170M [00:04<00:00, 42.0MB/s]
98%|#########7| 166M/170M [00:04<00:00, 42.8MB/s]
100%|##########| 170M/170M [00:04<00:00, 41.1MB/s]
+
0%| | 0.00/170M [00:00<?, ?B/s]
5%|4 | 7.99M/170M [00:00<00:03, 53.3MB/s]
9%|9 | 16.0M/170M [00:00<00:03, 50.9MB/s]
14%|#4 | 24.0M/170M [00:00<00:03, 49.8MB/s]
18%|#7 | 30.3M/170M [00:00<00:02, 50.5MB/s]
21%|## | 35.1M/170M [00:00<00:02, 48.6MB/s]
24%|##3 | 40.0M/170M [00:00<00:02, 46.6MB/s]
26%|##6 | 44.4M/170M [00:00<00:03, 42.5MB/s]
29%|##8 | 48.4M/170M [00:01<00:03, 38.9MB/s]
33%|###2 | 56.0M/170M [00:01<00:02, 41.5MB/s]
37%|###6 | 62.3M/170M [00:01<00:02, 44.8MB/s]
41%|#### | 69.4M/170M [00:01<00:02, 52.0MB/s]
44%|####3 | 74.6M/170M [00:01<00:02, 47.0MB/s]
47%|####6 | 79.2M/170M [00:01<00:02, 44.8MB/s]
49%|####9 | 83.6M/170M [00:01<00:02, 37.2MB/s]
52%|#####1 | 88.1M/170M [00:02<00:02, 39.4MB/s]
57%|#####6 | 96.0M/170M [00:02<00:01, 48.0MB/s]
61%|######1 | 104M/170M [00:02<00:01, 44.2MB/s
]
65%|######5 | 111M/170M [00:02<00:01, 50.0MB/s]
68%|######8 | 116M/170M [00:02<00:01, 49.5MB/s]
72%|#######1 | 122M/170M [00:02<00:01, 49.5MB/s]
75%|#######4 | 127M/170M [00:02<00:00, 45.3MB/s]
77%|#######7 | 131M/170M [00:03<00:00, 42.0MB/s]
80%|######## | 136M/170M [00:03<00:00, 42.3MB/s]
82%|########2 | 140M/170M [00:03<00:00, 40.8MB/s]
85%|########4 | 144M/170M [00:03<00:00, 32.4MB/s]
88%|########8 | 150M/170M [00:03<00:00, 38.4MB/s]
91%|######### | 154M/170M [00:03<00:00, 37.2MB/s]
94%|#########4| 160M/170M [00:03<00:00, 39.9MB/s]
99%|#########8| 167M/170M [00:03<00:00, 46.9MB/s]
100%|##########| 170M/170M [00:04<00:00, 44.4MB/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 36.406 seconds)
+ **Total running time of the script:** ( 3 minutes 35.059 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 6aae7f5a5a..a820341e96 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, 51.6MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 71.3MB/s]
+
0%| | 0.00/13.6M [00:00<?, ?B/s]
59%|#####8 | 7.99M/13.6M [00:00<00:00, 74.9MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 103MB/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)
- 90.0879 90.0527 91.3993 89.8019 0.2330
+ 90.5942 90.5546 96.1535 89.9876 0.6161
@@ -458,7 +458,7 @@ TODO
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 16.787 seconds)
+ **Total running time of the script:** ( 1 minutes 16.326 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 0ecadca8f7..fbdd4a7cd2 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)
- 119.4720 119.4017 121.8871 117.6485 0.8324
+ 119.6478 119.6423 121.0938 118.5766 0.3817
@@ -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 30.077 seconds)
+ **Total running time of the script:** ( 2 minutes 35.885 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 7aff703040..be0b85bc3f 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 46.660 seconds)
+ **Total running time of the script:** ( 1 minutes 49.286 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 634fd0c26d..2fe4d1bb83 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]
1%| | 1234/132723 [00:00<00:10, 12287.69KB/s]
3%|3 | 3994/132723 [00:00<00:06, 21278.05KB/s]
8%|7 | 9986/132723 [00:00<00:03, 38901.60KB/s]
14%|#3 | 18556/132723 [00:00<00:01, 57363.52KB/s]
21%|## | 27223/132723 [00:00<00:01, 67924.57KB/s]
26%|##5 | 34018/132723 [00:00<00:01, 56581.03KB/s]
32%|###2 | 42764/132723 [00:00<00:01, 65470.69KB/s]
39%|###8 | 51506/132723 [00:00<00:01, 71866.40KB/s]
45%|####5 | 60239/132723 [00:00<00:00, 76408.92KB/s]
51%|#####1 | 68252/132723 [00:01<00:00, 77507.40KB/s]
58%|#####8 | 77015/132723 [00:01<00:00, 80511.69KB/s]
64%|######4 | 85182/132723 [00:01<00:00, 75087.72KB/s]
71%|####### | 93981/132723 [00:01<00:00, 78735.34KB/s]
77%|#######6 | 101984/132723 [00:01<00:00, 61049.47KB/s]
83%|########3 | 110765/132723 [00:01<00:00, 67496.49KB/s]
89%|########9 |
118147/132723 [00:01<00:00, 68088.06KB/s]
96%|#########5| 126956/132723 [00:01<00:00, 73382.43KB/s]
100%|##########| 132723/132723 [00:01<00:00, 67668.32KB/s]
+
0%| | 0/132723 [00:00<?, ?KB/s]
1%|1 | 1364/132723 [00:00<00:09, 13592.63KB/s]
3%|3 | 4324/132723 [00:00<00:05, 22944.10KB/s]
8%|8 | 10833/132723 [00:00<00:02, 42148.08KB/s]
12%|#2 | 16244/132723 [00:00<00:02, 46022.01KB/s]
16%|#5 | 20839/132723 [00:00<00:02, 43286.87KB/s]
20%|#9 | 26540/132723 [00:00<00:02, 47728.24KB/s]
25%|##4 | 32711/132723 [00:00<00:01, 52148.10KB/s]
30%|### | 40468/132723 [00:00<00:01, 60057.65KB/s]
35%|###5 | 46518/132723 [00:00<00:01, 55275.22KB/s]
39%|###9 | 52144/132723 [00:01<00:01, 50337.60KB/s]
43%|####3 | 57418/132723 [00:01<00:01, 50987.76KB/s]
49%|####8 | 64707/132723 [00:01<00:01, 57104.03KB/s]
55%|#####4 | 72608/132723 [00:01<00:00, 63350.24KB/s]
60%|#####9 | 79069/132723 [00:01<00:00, 62766.84KB/s]
64%|######4 | 85433/132723 [00:01<00:00, 60001.84KB/s]
69%|######9 |
91860/132723 [00:01<00:00, 61200.41KB/s]
76%|#######5 | 100413/132723 [00:01<00:00, 68192.86KB/s]
82%|########2 | 109048/132723 [00:01<00:00, 73476.59KB/s]
88%|########7 | 116711/132723 [00:02<00:00, 74400.89KB/s]
94%|#########3| 124203/132723 [00:02<00:00, 68711.59KB/s]
99%|#########8| 131193/132723 [00:02<00:00, 66494.46KB/s]
100%|##########| 132723/132723 [00:02<00:00, 58396.11KB/s]
@@ -246,7 +246,7 @@ Display result
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 3 minutes 53.000 seconds)
+ **Total running time of the script:** ( 3 minutes 50.099 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 fd7f3a8452..2a4c29d531 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
@@ -5,28 +5,28 @@
Computation times
=================
-**16:32.793** total execution time for **how_to_deploy_models** files:
+**16:32.551** total execution time for **how_to_deploy_models** files:
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:52.1000 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:36.406 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:30.077 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:46.660 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:16.787 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 00:57.626 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno_tvmc.py` (``deploy_model_on_adreno_tvmc.py``) | 00:52.671 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:43.241 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:28.298 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:28.021 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``) | 00:00.006 | 0.0 MB |
-+------------------------------------------------------------------------------------------------------------------+------------+--------+
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:50.099 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:35.059 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:35.885 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:49.286 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:16.326 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 00:54.784 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno_tvmc.py` (``deploy_model_on_adreno_tvmc.py``) | 00:51.258 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:42.669 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:28.845 | 0.0 MB |
++------------------------------------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:28.335 | 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 e8bd80773f..f4eb10a646 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.zip3f4906e1-27a1-487c-bed5-2ab2f9a53345 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+ Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip9c87e86e-7778-429c-ab44-ef7922d9f5b9 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 fd1246c907..d1e8123bd9 100644
--- a/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/extend_tvm/sg_execution_times.rst.txt
@@ -5,14 +5,14 @@
Computation times
=================
-**00:55.065** total execution time for **how_to_extend_tvm** files:
+**00:54.287** total execution time for **how_to_extend_tvm** files:
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:51.171 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:50.477 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.796 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.726 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.089 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.077 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``) | 00:00.008 | 0.0 MB |
+| :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 2af8ce9184..d864d72560 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: 22227us [22227us] (48.46%; 48.46%)
- FoldScaleAxis: 23637us [6us] (51.54%; 51.54%)
- FoldConstant: 23631us [1706us] (51.52%; 99.97%)
- InferType: 21925us [21925us] (47.80%; 92.78%)
+ InferType: 22449us [22449us] (48.08%; 48.08%)
+ FoldScaleAxis: 24238us [8us] (51.92%; 51.92%)
+ FoldConstant: 24230us [1734us] (51.90%; 99.97%)
+ InferType: 22496us [22496us] (48.18%; 92.84%)
@@ -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: 21698us [21698us] (47.75%; 47.75%)
- FoldScaleAxis: 23747us [5us] (52.25%; 52.25%)
- FoldConstant: 23741us [1707us] (52.24%; 99.98%)
- InferType: 22034us [22034us] (48.49%; 92.81%)
+ InferType: 21872us [21872us] (48.21%; 48.21%)
+ FoldScaleAxis: 23500us [6us] (51.79%; 51.79%)
+ FoldConstant: 23494us [1731us] (51.78%; 99.97%)
+ InferType: 21762us [21762us] (47.96%; 92.63%)
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 e8e1bc1109..1356733cc4 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: 53.507839 ms
+ Convolution: 36.583423 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 041e6dc5c3..ffa257b098 100644
--- a/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_conv_tensorcore.rst.txt
@@ -598,7 +598,7 @@ be able to run on our build server
.. code-block:: none
- conv2d with tensor core: 12.246058 ms
+ conv2d with tensor core: 11.332147 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 4131f568e8..bd2c63acdd 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: 0.017962
- Baseline: 3.255473
+ Numpy running time: 0.018174
+ Baseline: 3.442605
@@ -227,7 +227,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
.. code-block:: none
- Opt1: 0.301741
+ Opt1: 0.296080
@@ -318,7 +318,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
.. code-block:: none
- Opt2: 0.341142
+ Opt2: 0.328482
@@ -406,7 +406,7 @@ the access pattern for A matrix is more cache friendly.
.. code-block:: none
- Opt3: 0.120074
+ Opt3: 0.116717
@@ -523,7 +523,7 @@ flattening.
.. code-block:: none
- Opt4: 0.109988
+ Opt4: 0.109713
@@ -635,7 +635,7 @@ write to C when all the block results are ready.
.. code-block:: none
- Opt5: 0.112287
+ Opt5: 0.112274
@@ -748,7 +748,7 @@ Furthermore, we can also utilize multi-core processors to do the thread-level pa
.. code-block:: none
- Opt6: 0.146727
+ Opt6: 0.147084
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 8430f382ea..75a3ac2f00 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
=================
-**00:34.963** total execution time for **how_to_optimize_operators** files:
+**00:35.296** total execution time for **how_to_optimize_operators** files:
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:31.996 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:32.341 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.859 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.891 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.108 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.063 | 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 eb2794f6ac..8d4ce685e4 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
=================
-**10:28.765** total execution time for **how_to_tune_with_autoscheduler** files:
+**09:59.269** total execution time for **how_to_tune_with_autoscheduler** files:
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 06:21.426 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 06:05.984 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:43.370 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:42.224 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:11.382 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:10.847 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:44.654 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:32.532 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:14.253 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:14.090 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:13.680 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:13.591 | 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 29b04616e3..e753dd4234 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
@@ -243,11 +243,11 @@ cooperative fetching, unrolling and operator fusion.
@T.prim_func
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": T.bool(True), "global_symbol": "main", "tir.noalias": T.bool(True)})
- blockIdx_x = T.launch_thread("blockIdx.x", 16)
+ blockIdx_x = T.launch_thread("blockIdx.x", 28)
conv2d_nchw = T.allocate([14], "float32", "local")
- pad_temp_shared = T.allocate([1568], "float32", "shared")
- kernel_shared = T.allocate([1024], "float32", "shared")
- threadIdx_x = T.launch_thread("threadIdx.x", 112)
+ pad_temp_shared = T.allocate([72], "float32", "shared")
+ kernel_shared = T.allocate([3072], "float32", "shared")
+ threadIdx_x = T.launch_thread("threadIdx.x", 64)
conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
@@ -263,516 +263,459 @@ cooperative fetching, unrolling and operator fusion.
conv2d_nchw_1[11] = T.float32(0)
conv2d_nchw_1[12] = T.float32(0)
conv2d_nchw_1[13] = T.float32(0)
- for rc_outer_outer, ry_outer_outer, rx_outer_outer in T.grid(16, 3, 3):
- cse_var_2: T.int32 = rc_outer_outer * 288
+ for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
+ cse_var_2: T.int32 = rc_outer_outer * 72
cse_var_1: T.int32 = ry_outer_outer * 3
+ pad_temp_shared_1 = T.Buffer((72,), data=pad_temp_shared, scope="shared")
+ with T.launch_thread("threadIdx.x", 64) as threadIdx_x_1:
+ data_1 = T.Buffer((25088,), data=data.data)
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= threadIdx_x_1 * 4 % 9 and threadIdx_x_1 * 4 % 9 < 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 * 4 // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + threadIdx_x_1 * 4 % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 1] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 1) % 9 and (threadIdx_x_1 * 4 + 1) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 1) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 1) % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 2] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 2) % 9 and (threadIdx_x_1 * 4 + 2) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 2) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 2) % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 3] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 3) % 9 and (threadIdx_x_1 * 4 + 3) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 3) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 3) % 9 - 8], T.float32(0))
threadIdx_x_1 = T.env_thread("threadIdx.x")
- pad_temp_shared_1 = T.Buffer((1568,), data=pad_temp_shared, scope="shared")
- data_1 = T.Buffer((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 % 49 // 7 + ry_outer_outer and threadIdx_x_1 % 49 // 7 + ry_outer_outer < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer - 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 <= ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 104], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 216], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 328], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 440], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 560] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 552], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 672] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 664], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 784] = T.if_then_else(1 <= threadIdx_x_1 % 49 // 7 + ry_outer_outer and threadIdx_x_1 % 49 // 7 + ry_outer_outer < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 776], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 896] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 888], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1008] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1000], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1120] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1112], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1232] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1224], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1344] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1336], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1456] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1448], T.float32(0))
- threadIdx_x_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.Buffer((1024,), data=kernel_shared, scope="shared")
+ kernel_shared_1 = T.Buffer((3072,), data=kernel_shared, scope="shared")
kernel_1 = T.Buffer((2359296,), data=kernel.data)
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 112] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 112) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 224] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 32256]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 336) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 560) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 672] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 96768]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 784) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 16):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 147456 + cse_var_2 + threadIdx_x_2 * 9 + cse_var_1 + rx_outer_outer + 142992]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 2] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 3] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 4] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 5] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 6] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 2] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 3] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 4] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 5] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 6] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 50] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 51] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 52] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 53] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 54] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 55] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 50] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 51] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 52] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 53] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 54] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 55] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 98] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 99] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 100] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 101] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 102] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 103] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 104] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 98] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 99] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 100] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 101] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 102] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 103] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 104] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 147] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 148] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 149] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 150] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 151] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 152] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 153] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 147] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 148] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 149] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 150] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 151] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 152] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 153] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 196] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 197] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 198] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 199] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 200] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 201] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 202] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 196] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 197] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 198] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 199] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 200] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 201] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 202] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 245] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 246] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 247] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 248] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 249] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 250] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 251] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 245] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 246] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 247] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 248] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 249] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 250] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 251] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 294] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 295] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 296] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 297] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 298] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 299] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 300] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 294] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 295] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 296] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 297] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 298] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 299] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 300] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 343] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 344] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 345] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 346] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 347] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 348] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 349] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 343] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 344] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 345] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 346] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 347] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 348] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 349] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 392] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 393] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 394] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 395] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 396] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 397] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 398] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 392] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 393] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 394] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 395] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 396] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 397] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 398] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 441] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 442] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 443] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 444] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 445] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 446] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 447] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 441] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 442] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 443] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 444] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 445] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 446] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 447] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 490] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 491] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 492] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 493] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 494] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 495] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 496] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 490] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 491] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 492] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 493] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 494] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 495] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 496] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 539] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 540] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 541] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 542] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 543] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 544] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 545] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 539] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 540] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 541] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 542] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 543] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 544] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 545] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 588] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 589] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 590] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 591] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 592] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 593] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 594] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 588] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 589] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 590] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 591] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 592] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 593] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 594] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 637] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 638] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 639] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 640] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 641] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 642] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 643] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 637] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 638] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 639] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 640] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 641] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 642] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 643] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 686] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 687] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 688] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 689] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 690] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 691] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 692] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 686] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 687] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 688] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 689] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 690] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 691] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 692] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 735] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 736] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 737] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 738] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 739] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 740] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 741] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 735] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 736] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 737] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 738] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 739] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 740] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 741] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 784] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 785] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 786] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 787] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 788] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 789] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 790] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 784] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 785] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 786] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 787] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 788] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 789] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 790] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 833] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 834] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 835] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 836] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 837] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 838] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 839] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 833] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 834] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 835] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 836] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 837] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 838] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 839] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 882] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 883] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 884] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 885] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 886] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 887] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 888] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 882] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 883] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 884] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 885] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 886] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 887] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 888] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 931] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 932] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 933] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 934] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 935] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 936] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 937] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 931] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 932] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 933] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 934] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 935] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 936] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 937] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 980] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 981] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 982] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 983] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 984] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 985] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 986] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 980] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 981] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 982] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 983] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 984] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 985] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 986] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1029] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1030] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1031] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1032] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1033] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1034] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1035] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1029] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1030] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1031] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1032] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1033] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1034] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1035] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1078] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1079] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1080] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1081] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1082] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1083] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1084] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1078] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1079] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1080] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1081] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1082] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1083] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1084] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1127] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1128] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1129] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1130] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1131] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1132] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1133] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1127] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1128] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1129] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1130] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1131] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1132] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1133] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1176] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1177] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1178] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1179] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1180] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1181] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1182] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1176] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1177] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1178] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1179] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1180] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1181] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1182] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1225] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1226] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1227] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1228] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1229] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1230] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1231] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1225] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1226] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1227] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1228] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1229] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1230] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1231] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1274] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1275] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1276] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1277] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1278] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1279] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1280] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1274] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1275] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1276] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1277] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1278] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1279] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1280] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1323] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1324] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1325] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1326] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1327] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1328] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1329] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1323] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1324] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1325] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1326] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1327] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1328] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1329] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1372] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1373] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1374] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1375] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1376] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1377] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1378] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1372] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1373] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1374] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1375] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1376] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1377] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1378] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1421] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1422] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1423] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1424] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1425] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1426] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1427] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1421] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1422] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1423] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1424] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1425] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1426] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1427] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1470] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1471] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1472] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1473] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1474] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1475] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1476] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1470] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1471] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1472] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1473] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1474] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1475] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1476] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1519] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1520] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1521] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1522] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1523] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1524] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1525] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1519] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1520] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1521] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1522] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1523] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1524] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1525] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
for i1_inner, i3_inner in T.grid(2, 7):
compute_1 = T.Buffer((25088,), data=compute.data)
bias_1 = T.Buffer((512,), data=bias.data)
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 98 + i1_inner * 49 + threadIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 * 2 + 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))
@@ -822,7 +765,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 0.315 ms
+ Execution time of this operator: 0.350 ms
@@ -870,33 +813,33 @@ 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=2)
- conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
- conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=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=7)
- conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
+ conv2d_nchw_xx_o_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=4)
- 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)
- conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+ conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2 [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
- compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
+ 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)
@@ -919,14 +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:
@@ -944,10 +887,10 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
#define int64_t long long
#define uint64_t unsigned long long
#endif
- extern "C" __global__ void __launch_bounds__(112) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+ 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[1568];
- __shared__ float kernel_shared[1024];
+ __shared__ float pad_temp_shared[72];
+ __shared__ float kernel_shared[3072];
conv2d_nchw[0] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
@@ -962,491 +905,411 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
conv2d_nchw[11] = 0.000000e+00f;
conv2d_nchw[12] = 0.000000e+00f;
conv2d_nchw[13] = 0.000000e+00f;
- for (int rc_outer_outer = 0; rc_outer_outer < 16; ++rc_outer_outer) {
+ for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
- for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
- __syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 49) / 7) + ry_outer_outer)) && ((((((int)threadIdx.x) % 49) / 7) + ry_outer_outer) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 104)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 216)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 328)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 440)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 560)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 552)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 672)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 664)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 784)] = (((((1 <= (((((int)threadIdx.x) % 49) / 7) + ry_outer_outer)) && ((((((int)threadIdx.x) % 49) / 7) + ry_outer_outer) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 776)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 896)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 888)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1008)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1000)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1120)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1112)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1232)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1224)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1344)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1336)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1456)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1448)] : 0.000000e+00f);
- kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 112)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 112) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 32256)];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 336) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 560)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 560) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 672)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 96768)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 784) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 129024)];
- if (((int)threadIdx.x) < 16) {
- kernel_shared[(((int)threadIdx.x) + 1008)] = kernel[((((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 288)) + (((int)threadIdx.x) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 142992)];
- }
- __syncthreads();
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 7)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 2)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 3)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 4)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 5)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 6)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 49)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 50)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 51)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 52)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 53)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 55)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 49)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 50)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 51)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 52)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 53)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 55)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 98)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 99)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 100)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 101)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 102)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 103)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 104)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 98)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 99)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 100)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 101)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 102)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 103)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 104)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 147)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 148)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 149)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 150)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 151)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 152)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 153)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 147)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 148)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 149)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 150)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 151)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 152)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 153)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 198)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 199)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 200)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 201)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 202)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 198)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 199)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 200)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 201)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 202)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 245)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 246)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 247)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 248)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 249)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 250)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 251)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 245)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 246)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 247)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 248)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 249)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 250)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 251)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 294)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 295)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 296)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 297)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 298)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 299)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 300)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 294)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 295)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 296)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 297)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 298)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 299)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 300)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 343)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 344)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 345)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 346)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 347)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 348)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 349)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 343)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 344)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 345)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 346)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 347)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 348)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 349)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 392)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 393)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 394)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 395)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 396)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 397)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 398)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 392)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 393)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 394)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 395)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 396)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 397)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 398)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 490)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 491)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 492)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 493)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 494)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 495)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 496)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 490)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 491)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 492)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 493)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 494)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 495)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 496)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 539)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 540)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 541)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 542)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 543)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 544)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 545)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 539)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 540)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 541)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 542)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 543)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 544)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 545)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 588)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 589)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 590)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 591)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 592)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 593)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 594)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 588)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 589)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 590)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 591)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 592)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 593)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 594)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 637)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 638)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 639)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 640)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 641)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 642)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 643)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 637)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 638)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 639)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 640)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 641)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 642)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 643)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 686)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 687)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 688)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 689)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 690)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 691)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 692)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 686)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 687)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 688)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 689)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 690)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 691)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 692)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 735)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 736)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 737)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 738)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 739)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 740)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 741)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 735)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 736)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 737)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 738)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 739)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 740)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 741)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 784)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 785)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 786)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 787)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 788)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 789)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 790)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 784)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 785)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 786)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 787)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 788)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 789)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 790)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 833)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 834)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 835)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 836)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 837)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 838)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 839)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 833)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 834)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 835)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 836)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 837)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 838)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 839)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 882)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 883)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 884)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 885)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 886)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 887)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 888)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 882)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 883)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 884)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 885)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 886)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 887)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 888)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 931)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 932)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 933)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 934)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 935)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 936)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 937)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 931)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 932)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 933)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 934)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 935)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 936)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 937)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 980)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 981)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 982)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 983)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 984)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 985)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 986)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 980)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 981)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 982)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 983)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 984)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 985)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 986)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1029)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1030)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1031)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1032)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1033)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1034)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1035)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1029)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1030)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1031)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1032)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1033)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1034)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1035)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1078)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1079)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1080)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1081)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1082)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1083)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1084)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1078)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1079)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1080)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1081)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1082)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1083)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1084)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1176)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1177)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1178)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1179)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1180)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1181)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1182)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1176)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1177)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1178)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1179)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1180)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1181)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1182)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1225)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1226)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1227)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1228)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1229)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1230)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1231)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1225)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1226)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1227)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1228)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1229)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1230)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1231)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1274)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1275)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1276)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1277)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1278)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1279)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1280)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1274)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1275)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1276)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1277)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1278)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1279)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1280)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1324)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1325)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1326)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1327)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1328)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1329)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1324)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1325)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1326)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1327)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1328)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1329)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1372)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1373)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1374)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1375)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1376)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1377)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1372)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1373)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1374)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1375)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1376)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1377)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1421)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1422)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1423)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1424)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1425)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1426)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1427)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1421)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1422)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1423)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1424)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1425)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1426)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1427)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1470)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1471)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1472)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1473)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1474)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1475)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1476)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1470)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1471)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1472)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1473)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1474)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1475)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1476)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1519)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1520)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1521)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1522)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1523)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1524)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1525)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1519)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1520)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1521)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1522)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1523)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1524)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1525)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
+ __syncthreads();
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) * 4) % 9))) && (((((int)threadIdx.x) * 4) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
}
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 1) % 9))) && ((((((int)threadIdx.x) * 4) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
+ }
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 2) % 9))) && ((((((int)threadIdx.x) * 4) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
+ }
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 3) % 9))) && ((((((int)threadIdx.x) * 4) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
+ }
+ kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 64) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 128) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
+ kernel_shared[(((((((int)threadIdx.x) + 256) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 320) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
+ kernel_shared[(((((((int)threadIdx.x) + 448) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 512) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
+ kernel_shared[(((((((int)threadIdx.x) + 640) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 704) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
+ kernel_shared[(((((((int)threadIdx.x) + 832) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 896) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
+ kernel_shared[(((((((int)threadIdx.x) + 1024) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1088) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
+ kernel_shared[(((((((int)threadIdx.x) + 1216) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1280) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
+ kernel_shared[(((((((int)threadIdx.x) + 1408) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1472) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
+ kernel_shared[(((((((int)threadIdx.x) + 1600) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1664) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
+ kernel_shared[(((((((int)threadIdx.x) + 1792) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1856) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
+ kernel_shared[(((((((int)threadIdx.x) + 1984) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2048) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
+ kernel_shared[(((((((int)threadIdx.x) + 2176) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2240) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
+ kernel_shared[(((((((int)threadIdx.x) + 2368) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2432) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
+ kernel_shared[(((((((int)threadIdx.x) + 2560) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2624) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
+ kernel_shared[(((((((int)threadIdx.x) + 2752) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2816) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
+ kernel_shared[(((((((int)threadIdx.x) + 2944) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 3008) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ __syncthreads();
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[1] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[2] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[4] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[5] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[0] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
}
}
for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[(((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 98)) + (i1_inner * 49)) + ((((int)threadIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[(((((int)blockIdx.x) * 32) + ((((int)threadIdx.x) / 7) * 2)) + 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);
}
}
}
@@ -1507,7 +1370,7 @@ In the example below we resume the status and do more 5 trials.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 6 minutes 21.426 seconds)
+ **Total running time of the script:** ( 6 minutes 5.984 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 12f5cc2294..cedc82e826 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)
- 8.1186 8.1181 8.1212 8.1165 0.0019
+ 8.1404 8.1394 8.1507 8.1311 0.0080
@@ -675,7 +675,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 11.382 seconds)
+ **Total running time of the script:** ( 1 minutes 10.847 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 16419ea551..0a921911e4 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)
- 759.7677 760.1449 763.1201 756.0381 2.9035
+ 754.6195 754.4586 756.4399 752.9600 1.4252
@@ -694,7 +694,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 43.370 seconds)
+ **Total running time of the script:** ( 1 minutes 42.224 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 eccaffc22b..2b1d63c117 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
@@ -389,27 +389,86 @@ 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": T.bool(True), "global_symbol": "main", "tir.noalias": T.bool(True)})
- for i0_outer in T.parallel(64):
- compute_1 = T.allocate([64], "float32", "global")
- for i1_outer in range(16):
- compute_2 = T.Buffer((64,), data=compute_1)
- for nb_j_inner in range(2):
- for i_inner_init, j_init in T.grid(2, 16):
- compute_2[i_inner_init * 32 + nb_j_inner * 16 + j_init] = T.float32(0)
- for elem_idx, i_inner, j in T.grid(T.Let(placeholder_5[cse_var_1 + 1] - placeholder_5[cse_var_1], where={cse_var_1: i1_outer * 2 + nb_j_inner}), 2, 16):
- cse_var_1 = T.int32()
- placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
- cse_var_3: T.int32 = i1_outer * 2 + nb_j_inner
- cse_var_2: T.int32 = i_inner * 32 + nb_j_inner * 16 + j
- placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
- placeholder_7 = T.Buffer((32768,), data=placeholder.data)
- placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_2] = compute_2[cse_var_2] + placeholder_6[placeholder_5[cse_var_3] * 16 + elem_idx * 16 + j] * T.max(placeholder_7[i0_outer * 512 + i_inner * 256 + placeholder_8[placeholder_5[cse_var_3] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(2, 32):
- cse_var_4: T.int32 = i0_outer * 1024 + i0_inner * 512 + i1_outer * 32 + i1_inner
- compute_3 = T.Buffer((65536,), data=compute.data)
- placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
- compute_3[cse_var_4] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_4], T.float32(0))
+ for i0_outer_i1_outer_fused in T.parallel(32):
+ compute_1 = T.allocate([2048], "float32", "global")
+ compute_2 = T.Buffer((2048,), data=compute_1)
+ for i_outer_inner in range(2):
+ for i_inner_init in range(64):
+ cse_var_1: T.int32 = i_outer_inner * 1024 + i_inner_init * 16
+ compute_2[cse_var_1] = T.float32(0)
+ compute_2[cse_var_1 + 1] = T.float32(0)
+ compute_2[cse_var_1 + 2] = T.float32(0)
+ compute_2[cse_var_1 + 3] = T.float32(0)
+ compute_2[cse_var_1 + 4] = T.float32(0)
+ compute_2[cse_var_1 + 5] = T.float32(0)
+ compute_2[cse_var_1 + 6] = T.float32(0)
+ compute_2[cse_var_1 + 7] = T.float32(0)
+ compute_2[cse_var_1 + 8] = T.float32(0)
+ compute_2[cse_var_1 + 9] = T.float32(0)
+ compute_2[cse_var_1 + 10] = T.float32(0)
+ compute_2[cse_var_1 + 11] = T.float32(0)
+ compute_2[cse_var_1 + 12] = T.float32(0)
+ compute_2[cse_var_1 + 13] = T.float32(0)
+ compute_2[cse_var_1 + 14] = T.float32(0)
+ compute_2[cse_var_1 + 15] = T.float32(0)
+ for elem_idx, i_inner in T.grid(placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused], 64):
+ placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
+ placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
+ placeholder_7 = T.Buffer((32768,), data=placeholder.data)
+ placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_2: T.int32 = i_outer_inner * 1024 + i_inner * 16
+ compute_2[cse_var_2] = compute_2[cse_var_2] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_3: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 1
+ compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 1] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_4: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 2
+ compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 2] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_5: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 3
+ compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 3] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_6: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 4
+ compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 4] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_7: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 5
+ compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 5] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_8: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 6
+ compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 6] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_9: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 7
+ compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 7] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_10: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 8
+ compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 8] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_11: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 9
+ compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 9] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_12: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 10
+ compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 10] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_13: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 11
+ compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 11] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_14: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 12
+ compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 12] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_15: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 13
+ compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 13] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_16: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 14
+ compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 14] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_17: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 15
+ compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 15] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ for i0_inner in range(128):
+ cse_var_18: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 16
+ compute_3 = T.Buffer((65536,), data=compute.data)
+ placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
+ compute_3[cse_var_18:cse_var_18 + 16] = T.max(compute_2[i0_inner * 16:i0_inner * 16 + 16] + placeholder_5[cse_var_18:cse_var_18 + 16], T.Broadcast(T.float32(0), 16))
@@ -459,7 +518,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 1.873 ms
+ Execution time of this operator: 1.824 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 d19ca5c2e0..e5e7fa3731 100644
--- a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
@@ -5,16 +5,16 @@
Computation times
=================
-**00:36.350** total execution time for **how_to_tune_with_autotvm** files:
+**00:57.262** 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``) | 00:36.315 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``) | 00:57.227 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``) | 00:00.021 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``) | 00:00.005 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``) | 00:00.005 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``) | 00:00.004 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``) | 00:00.004 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
index 0dec39b2b8..369bf2871d 100644
--- a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
@@ -268,8 +268,25 @@ for this template
waiting for device...
device available
Get devices for measurement successfully!
- No: 1 GFLOPS: 35.00/35.00 result: MeasureResult(costs=(0.0066147660625,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2832298278808594, timestamp=1680521223.4808674) [('tile_f', [-1, 1, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5281179
- No: 2 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ No: 1 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 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, 2, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7711234
+ 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)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -391,8 +408,13 @@ 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, 4, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9415737
- No: 3 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 1, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3956030
+ No: 3 GFLOPS: 183.79/183.79 result: MeasureResult(costs=(0.0012595632702702701,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.519602537155151, timestamp=1680581738.1102219) [('tile_f', [-1, 1, 8, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4502421
+ No: 4 GFLOPS: 154.18/183.79 result: MeasureResult(costs=(0.001501489656716418,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.0189778804779053, timestamp=1680581738.9676352) [('tile_f', [-1, 2, 2, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3488649
+ No: 5 GFLOPS: 110.76/183.79 result: MeasureResult(costs=(0.00209018620754717,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.892058372497559, timestamp=1680581748.0066292) [('tile_f', [-1, 1, 1, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3786116
+ No: 6 GFLOPS: 31.66/183.79 result: MeasureResult(costs=(0.007312564285714286,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5477538108825684, timestamp=1680581748.8912892) [('tile_f', [-1, 2, 1, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3116181
+ No: 7 GFLOPS: 429.48/429.48 result: MeasureResult(costs=(0.0005390288449612404,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.55623459815979, timestamp=1680581751.377373) [('tile_f', [-1, 1, 64, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,455885
+ No: 8 GFLOPS: 0.00/429.48 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
@@ -514,8 +536,8 @@ 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, 32, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,279053
- No: 4 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7140687
+ No: 9 GFLOPS: 0.00/429.48 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
@@ -637,10 +659,8 @@ 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, 8, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6627624
- No: 5 GFLOPS: 30.79/35.00 result: MeasureResult(costs=(0.007519179352941176,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.6613967418670654, timestamp=1680521231.781294) [('tile_f', [-1, 64, 2, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3527056
- No: 6 GFLOPS: 4.08/35.00 result: MeasureResult(costs=(0.05678187175,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.161860227584839, timestamp=1680521233.0591633) [('tile_f', [-1, 8, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3680007
- No: 7 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6841672
+ No: 10 GFLOPS: 0.00/429.48 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
@@ -762,8 +782,11 @@ 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, 1, 4, 64]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5646067
- No: 8 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9655792
+ No: 11 GFLOPS: 655.02/655.02 result: MeasureResult(costs=(0.00035342363841807915,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7047991752624512, timestamp=1680581753.9538736) [('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3391971
+ No: 12 GFLOPS: 11.59/655.02 result: MeasureResult(costs=(0.0199772745,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.3188257217407227, timestamp=1680581754.9549854) [('tile_f', [-1, 4, 2, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7409666
+ No: 13 GFLOPS: 169.40/655.02 result: MeasureResult(costs=(0.001366572396226415,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.081042766571045, timestamp=1680581760.2097487) [('tile_f', [-1, 8, 4, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9891442
+ No: 14 GFLOPS: 0.00/655.02 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 +908,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 32, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8272821
- No: 9 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7674285
+ No: 15 GFLOPS: 0.00/655.02 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 +1031,8 @@ 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, 64, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9457914
- No: 10 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6699522
+ No: 16 GFLOPS: 0.00/655.02 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 +1154,9 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 16, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5563887
- No: 11 GFLOPS: 0.00/35.00 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, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1484399
+ No: 17 GFLOPS: 385.93/655.02 result: MeasureResult(costs=(0.0005998543142857143,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5332989692687988, timestamp=1680581763.394771) [('tile_f', [-1, 1, 32, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3719630
+ No: 18 GFLOPS: 0.00/655.02 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 +1278,9 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 2, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8052191
- No: 12 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 2, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3752654
+ No: 19 GFLOPS: 128.18/655.02 result: MeasureResult(costs=(0.0018060951754385963,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.7704761028289795, timestamp=1680581764.264103) [('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,8978924
+ No: 20 GFLOPS: 0.00/655.02 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
@@ -1377,747 +1402,7 @@ 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, 1, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10319545
- No: 13 GFLOPS: 0.00/35.00 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
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2039327
- No: 14 GFLOPS: 139.11/139.11 result: MeasureResult(costs=(0.0016641625238095238,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.1421754360198975, timestamp=1680521236.668181) [('tile_f', [-1, 1, 16, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3416905
- No: 15 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 16, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4868100
- No: 16 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,5169747
- No: 17 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3206886
- No: 18 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8220205
- No: 19 GFLOPS: 17.50/139.11 result: MeasureResult(costs=(0.013231103875,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2924680709838867, timestamp=1680521238.2001758) [('tile_f', [-1, 16, 4, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6001719
- No: 20 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
- tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
- Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,538236
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4815751
@@ -2172,9 +1457,9 @@ and measure running time.
Finish loading 20 records
Best config:
- [('tile_f', [-1, 1, 16, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3416905
+ [('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3391971
Finish loading 20 records
- Time cost of this operator: 0.001949
+ Time cost of this operator: 0.000776
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 2472fb7867..e1c6168cce 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_autotune.rst.txt
@@ -360,10 +360,10 @@ Timing the untuned program
########## Build without Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
- tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 316.0 98.745 (1, 2, 10, 10, 3) 2 1 [316.0]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.061 0.957 (1, 6, 10, 10) 1 1 [3.061]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.954 0.298 (1, 1, 10, 10, 3) 1 1 [0.954]
- Total_time - 320.015 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 314.6 98.734 (1, 2, 10, 10, 3) 2 1 [314.6]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.082 0.967 (1, 6, 10, 10) 1 1 [3.082]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.952 0.299 (1, 1, 10, 10, 3) 1 1 [0.952]
+ Total_time - 318.634 - - - - -
@@ -428,10 +428,10 @@ Timing the tuned program
########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
- tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 103.0 97.503 (1, 6, 10, 10, 1) 2 1 [103.0]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.781 1.686 (1, 6, 10, 10) 1 1 [1.781]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.857 0.811 (1, 3, 10, 10, 1) 1 1 [0.857]
- Total_time - 105.638 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 100.9 97.327 (1, 6, 10, 10, 1) 2 1 [100.9]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.8 1.736 (1, 6, 10, 10) 1 1 [1.8]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.972 0.937 (1, 1, 10, 10, 3) 1 1 [0.972]
+ Total_time - 103.671 - - - - -
@@ -439,7 +439,7 @@ Timing the tuned program
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 22.617 seconds)
+ **Total running time of the script:** ( 1 minutes 21.699 seconds)
.. _sphx_glr_download_how_to_work_with_microtvm_micro_autotune.py:
diff --git a/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt b/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
index 73ea5ac384..b44cd7712d 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
@@ -118,7 +118,7 @@ download a cat image and preprocess it to use as the model input.
/venv/apache-tvm-py3.7/lib/python3.7/site-packages/torch/ao/quantization/utils.py:281: UserWarning: must run observer before calling calculate_qparams. Returning default values.
"must run observer before calling calculate_qparams. " +
Downloading: "https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2_qnnpack_37f702c5.pth
-
0%| | 0.00/3.42M [00:00<?, ?B/s]
90%|######### | 3.09M/3.42M [00:00<00:00, 32.3MB/s]
100%|##########| 3.42M/3.42M [00:00<00:00, 33.1MB/s]
+
0%| | 0.00/3.42M [00:00<?, ?B/s]
61%|###### | 2.09M/3.42M [00:00<00:00, 11.8MB/s]
100%|##########| 3.42M/3.42M [00:00<00:00, 18.8MB/s]
/workspace/python/tvm/relay/frontend/pytorch_utils.py:47: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
return LooseVersion(torch_ver) > ver
/venv/apache-tvm-py3.7/lib/python3.7/site-packages/setuptools/_distutils/version.py:346: DeprecationWarning: distutils Version classes are deprecated. Use packaging.version instead.
@@ -324,7 +324,7 @@ Look up prediction top 1 index in 1000 class synset.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 18.650 seconds)
+ **Total running time of the script:** ( 1 minutes 17.974 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 0b19fb93dd..5099291685 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
@@ -217,7 +217,7 @@ take about **2 minutes** to download the Stanford Cars, while COCO 2017 validati
.. code-block:: none
- '/tmp/tmpf6o0yyh0/images/random'
+ '/tmp/tmpaxabpir0/images/random'
@@ -308,7 +308,7 @@ 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: [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0]
+ :alt: [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]
:srcset: /how_to/work_with_microtvm/images/sphx_glr_micro_train_001.png
:class: sphx-glr-single-img
@@ -317,8 +317,8 @@ objects to other stuff? We can display some examples from our datasets using ``m
.. code-block:: none
- /tmp/tmpf6o0yyh0/images/target contains 8144 images
- /tmp/tmpf6o0yyh0/images/random contains 5000 images
+ /tmp/tmpaxabpir0/images/target contains 8144 images
+ /tmp/tmpaxabpir0/images/random contains 5000 images
@@ -493,13 +493,13 @@ the time on our validation set).
.. code-block:: none
Epoch 1/3
- 328/328 - 41s - loss: 0.2421 - accuracy: 0.9175 - val_loss: 0.1331 - val_accuracy: 0.9535 - 41s/epoch - 124ms/step
+ 328/328 - 41s - loss: 0.2252 - accuracy: 0.9244 - val_loss: 0.1283 - val_accuracy: 0.9562 - 41s/epoch - 125ms/step
Epoch 2/3
- 328/328 - 35s - loss: 0.1024 - accuracy: 0.9624 - val_loss: 0.1931 - val_accuracy: 0.9358 - 35s/epoch - 105ms/step
+ 328/328 - 35s - loss: 0.1040 - accuracy: 0.9614 - val_loss: 0.0961 - val_accuracy: 0.9675 - 35s/epoch - 105ms/step
Epoch 3/3
- 328/328 - 34s - loss: 0.0714 - accuracy: 0.9728 - val_loss: 0.3400 - val_accuracy: 0.8716 - 34s/epoch - 105ms/step
+ 328/328 - 34s - loss: 0.0595 - accuracy: 0.9767 - val_loss: 0.1097 - val_accuracy: 0.9645 - 34s/epoch - 105ms/step
- <keras.callbacks.History object at 0x7f896e28c6d0>
+ <keras.callbacks.History object at 0x7fcdcd711790>
@@ -860,7 +860,7 @@ 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 26.612 seconds)
+ **Total running time of the script:** ( 4 minutes 30.838 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 840cbb51b8..4e9154b9f4 100644
--- a/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/sg_execution_times.rst.txt
@@ -5,20 +5,20 @@
Computation times
=================
-**07:33.972** total execution time for **how_to_work_with_microtvm** files:
+**07:36.295** total execution time for **how_to_work_with_microtvm** files:
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:26.612 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:30.838 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 01:22.617 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 01:21.699 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:18.650 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:17.974 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:10.409 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:10.265 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_custom_ide.py` (``micro_custom_ide.py``) | 00:08.267 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_custom_ide.py` (``micro_custom_ide.py``) | 00:08.078 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:07.416 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:07.442 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``) | 00:00.000 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt b/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
index bb47ae54af..9ded999c8e 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
=================
-**00:36.850** total execution time for **how_to_work_with_relay** files:
+**00:36.825** total execution time for **how_to_work_with_relay** files:
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:32.167 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:32.208 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:02.888 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:02.876 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:01.789 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:01.735 | 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 12f0fd747d..12ccd16f55 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 0x7f85b9eb20e0>
+ <function my_cuda_math_rule at 0x7fca2c73c290>
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 838fa03f7d..a1499970d3 100644
--- a/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/sg_execution_times.rst.txt
@@ -5,22 +5,22 @@
Computation times
=================
-**00:09.169** total execution time for **how_to_work_with_schedules** files:
+**00:08.131** total execution time for **how_to_work_with_schedules** files:
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:06.384 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:05.312 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.280 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.304 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.617 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.623 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``) | 00:00.610 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``) | 00:00.130 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``) | 00:00.132 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``) | 00:00.064 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.056 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.055 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``) | 00:00.030 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
index 88389e1453..9796374774 100644
--- a/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
Computation times
=================
-**00:31.529** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:31.059** total execution time for **topic_vta_tutorials_autotvm** files:
+---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.522 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.052 | 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 303350f403..2aeb0aa23d 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
@@ -293,7 +293,7 @@ The compilation steps are:
DeprecationWarning,
/workspace/vta/tutorials/frontend/deploy_classification.py:213: DeprecationWarning: legacy graph executor behavior of producing json / lib / params will be removed in the next release. Please see documents of tvm.contrib.graph_executor.GraphModule for the new recommended usage.
relay_prog, target=tvm.target.Target(target, host=env.target_host), params=params
- resnet18_v1 inference graph built in 33.66s!
+ resnet18_v1 inference graph built in 33.19s!
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 e5e5a7fd54..23554a1d2d 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
@@ -337,7 +337,7 @@ The compilation steps are:
/workspace/python/tvm/relay/build_module.py:348: DeprecationWarning: Please use input parameter mod (tvm.IRModule) instead of deprecated parameter mod (tvm.relay.function.Function)
DeprecationWarning,
- yolov3-tiny inference graph built in 22.99s!
+ yolov3-tiny inference graph built in 22.58s!
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 c635920c06..311265ca1e 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
Computation times
=================
-**01:40.849** total execution time for **topic_vta_tutorials_frontend** files:
+**01:40.122** total execution time for **topic_vta_tutorials_frontend** files:
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.626 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:50.064 | 0.0 MB |
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:50.223 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.058 | 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 2ddde8b4ad..4fb506d5f5 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
=================
-**00:03.209** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.171** total execution time for **topic_vta_tutorials_optimize** files:
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.735 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.678 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.474 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.493 | 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 2f3cca5bba..1245c26df7 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
=================
-**00:00.830** total execution time for **topic_vta_tutorials** files:
+**00:00.827** total execution time for **topic_vta_tutorials** files:
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.422 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.427 | 0.0 MB |
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.408 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.401 | 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 db4ca737ec..4b578540a0 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -207,6 +207,13 @@ trials, we can load the best schedule from the log file and apply it.
+.. rst-class:: sphx-glr-script-out
+
+ .. code-block:: none
+
+
+ *E
+
@@ -318,7 +325,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 94.953 ms
+ Execution time of this operator: 94.545 ms
@@ -416,7 +423,7 @@ resume the status and do more 5 trials.
.. code-block:: none
Resume search:
-
+ .T
@@ -434,7 +441,7 @@ operations.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 44.840 seconds)
+ **Total running time of the script:** ( 1 minutes 56.214 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 6504a8be27..855779c6d7 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: 3.93/3.93 result: MeasureResult(costs=(0.06828601479999999,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.3776822090148926, timestamp=1680519541.316674) [('tile_y', [-1, 32]), ('tile_x', [-1, 16])],None,45
- No: 2 GFLOPS: 11.71/11.71 result: MeasureResult(costs=(0.0229281526,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.625922441482544, timestamp=1680519541.9470005) [('tile_y', [-1, 8]), ('tile_x', [-1, 512])],None,93
- No: 3 GFLOPS: 11.09/11.71 result: MeasureResult(costs=(0.0242145548,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6321713924407959, timestamp=1680519543.8310676) [('tile_y', [-1, 2]), ('tile_x', [-1, 512])],None,91
- No: 4 GFLOPS: 6.42/11.71 result: MeasureResult(costs=(0.041814656,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.9472103118896484, timestamp=1680519546.0236704) [('tile_y', [-1, 512]), ('tile_x', [-1, 64])],None,69
- No: 5 GFLOPS: 4.38/11.71 result: MeasureResult(costs=(0.061313176999999996,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2453370094299316, timestamp=1680519547.3828332) [('tile_y', [-1, 8]), ('tile_x', [-1, 16])],None,43
- No: 6 GFLOPS: 9.69/11.71 result: MeasureResult(costs=(0.0277092016,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.689699649810791, timestamp=1680519549.3533072) [('tile_y', [-1, 2]), ('tile_x', [-1, 32])],None,51
- No: 7 GFLOPS: 0.96/11.71 result: MeasureResult(costs=(0.2808641504,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.748321056365967, timestamp=1680519554.1088438) [('tile_y', [-1, 32]), ('tile_x', [-1, 2])],None,15
- No: 8 GFLOPS: 2.07/11.71 result: MeasureResult(costs=(0.12944264479999998,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.3285298347473145, timestamp=1680519556.4387348) [('tile_y', [-1, 128]), ('tile_x', [-1, 4])],None,27
- No: 9 GFLOPS: 0.51/11.71 result: MeasureResult(costs=(0.5296113604,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.675562381744385, timestamp=1680519565.2277825) [('tile_y', [-1, 256]), ('tile_x', [-1, 1])],None,8
- No: 10 GFLOPS: 11.38/11.71 result: MeasureResult(costs=(0.0235853958,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5853269100189209, timestamp=1680519565.8662276) [('tile_y', [-1, 32]), ('tile_x', [-1, 512])],None,95
+ No: 1 GFLOPS: 9.78/9.78 result: MeasureResult(costs=(0.027459741200000005,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.686920166015625, timestamp=1680580045.3394732) [('tile_y', [-1, 1]), ('tile_x', [-1, 256])],None,80
+ No: 2 GFLOPS: 10.11/10.11 result: MeasureResult(costs=(0.026553794799999995,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.667900800704956, timestamp=1680580046.0218923) [('tile_y', [-1, 4]), ('tile_x', [-1, 32])],None,52
+ No: 3 GFLOPS: 0.90/10.11 result: MeasureResult(costs=(0.2988416214,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.032846450805664, timestamp=1680580052.3265057) [('tile_y', [-1, 64]), ('tile_x', [-1, 2])],None,16
+ No: 4 GFLOPS: 14.72/14.72 result: MeasureResult(costs=(0.0182314066,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.8457832336425781, timestamp=1680580054.1388175) [('tile_y', [-1, 64]), ('tile_x', [-1, 64])],None,66
+ No: 5 GFLOPS: 2.68/14.72 result: MeasureResult(costs=(0.1000616446,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.8402447700500488, timestamp=1680580057.367008) [('tile_y', [-1, 4]), ('tile_x', [-1, 2])],None,12
+ No: 6 GFLOPS: 10.93/14.72 result: MeasureResult(costs=(0.0245507748,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6259403228759766, timestamp=1680580058.0188527) [('tile_y', [-1, 2]), ('tile_x', [-1, 512])],None,91
+ No: 7 GFLOPS: 0.96/14.72 result: MeasureResult(costs=(0.28075304759999997,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.727412700653076, timestamp=1680580062.765185) [('tile_y', [-1, 32]), ('tile_x', [-1, 2])],None,15
+ No: 8 GFLOPS: 3.07/14.72 result: MeasureResult(costs=(0.0873125868,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.6450128555297852, timestamp=1680580064.4225287) [('tile_y', [-1, 256]), ('tile_x', [-1, 8])],None,38
+ No: 9 GFLOPS: 13.78/14.72 result: MeasureResult(costs=(0.019481429,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.5480635166168213, timestamp=1680580065.1220503) [('tile_y', [-1, 128]), ('tile_x', [-1, 64])],None,67
+ No: 10 GFLOPS: 12.94/14.72 result: MeasureResult(costs=(0.020752020200000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.607900857925415, timestamp=1680580065.7162917) [('tile_y', [-1, 128]), ('tile_x', [-1, 128])],None,77
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index fde62c9dc1..01d1868ed3 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': 512.4494685900003, 'median': 512.0274692999942, 'std': 2.756563460026631}
+ {'mean': 513.2705139899997, 'median': 512.6211620999982, 'std': 3.249682685153772}
@@ -545,30 +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: 11.59/ 23.89 GFLOPS | Progress: (4/20) | 11.54 s
[Task 1/25] Current/Best: 3.17/ 23.89 GFLOPS | Progress: (8/20) | 16.19 s
[Task 1/25] Current/Best: 9.27/ 23.89 GFLOPS | Progress: (12/20) | 18.67 s
[Task 1/25] Current/Best: 7.27/ 23.89 GFLOPS | Progress: (16/20) | 21.83 s
[Task 1/25] Current/Best: 7.15/ 23.89 GFLOPS | Progress: (20/20) | 24.04 s Done.
-
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 20.45/ 20.45 GFLOPS | Progress: (4/20) | 4.83 s
[Task 2/25] Current/Best: 13.93/ 20.45 GFLOPS | Progress: (8/20) | 6.30 s
[Task 2/25] Current/Best: 3.45/ 20.45 GFLOPS | Progress: (12/20) | 7.68 s
[Task 2/25] Current/Best: 7.79/ 20.45 GFLOPS | Progress: (16/20) | 9.31 s
[Task 2/25] Current/Best: 14.50/ 20.45 GFLOPS | Progress: (20/20) | 11.30 s Done.
-
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 1.61/ 12.36 GFLOPS | Progress: (4/20) | 6.75 s
[Task 3/25] Current/Best: 14.00/ 17.23 GFLOPS | Progress: (8/20) | 9.20 s
[Task 3/25] Current/Best: 19.92/ 19.92 GFLOPS | Progress: (12/20) | 11.40 s
[Task 3/25] Current/Best: 5.78/ 19.92 GFLOPS | Progress: (16/20) | 15.33 s
[Task 3/25] Current/Best: 11.76/ 19.92 GFLOPS | Progress: (20/20) | 17.97 s Done.
-
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 11.70/ 13.41 GFLOPS | Progress: (4/20) | 5.67 s
[Task 4/25] Current/Best: 7.76/ 17.38 GFLOPS | Progress: (8/20) | 8.10 s
[Task 4/25] Current/Best: 8.08/ 17.38 GFLOPS | Progress: (12/20) | 11.27 s
[Task 4/25] Current/Best: 13.98/ 21.57 GFLOPS | Progress: (16/20) | 13.73 s
[Task 4/25] Current/Best: 10.02/ 21.57 GFLOPS | Progress: (20/20) | 16.49 s Done.
-
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 13.04/ 16.29 GFLOPS | Progress: (4/20) | 5.47 s
[Task 5/25] Current/Best: 17.75/ 19.46 GFLOPS | Progress: (8/20) | 7.19 s
[Task 5/25] Current/Best: 2.68/ 19.46 GFLOPS | Progress: (12/20) | 9.47 s
[Task 5/25] Current/Best: 10.03/ 19.46 GFLOPS | Progress: (16/20) | 11.43 s
[Task 5/25] Current/Best: 5.24/ 19.46 GFLOPS | Progress: (20/20) | 14.41 s Done.
-
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 11.91/ 13.22 GFLOPS | Progress: (4/20) | 6.30 s
[Task 6/25] Current/Best: 13.02/ 16.42 GFLOPS | Progress: (8/20) | 9.70 s
[Task 6/25] Current/Best: 6.36/ 16.42 GFLOPS | Progress: (12/20) | 14.60 s
[Task 6/25] Current/Best: 3.75/ 18.31 GFLOPS | Progress: (16/20) | 17.75 s
[Task 6/25] Current/Best: 2.74/ 18.31 GFLOPS | Progress: (20/20) | 21.88 s Done.
-
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 11.98/ 18.14 GFLOPS | Progress: (4/20) | 5.63 s
[Task 7/25] Current/Best: 17.31/ 18.79 GFLOPS | Progress: (8/20) | 7.84 s
[Task 7/25] Current/Best: 10.19/ 19.13 GFLOPS | Progress: (12/20) | 10.15 s
[Task 7/25] Current/Best: 15.69/ 19.13 GFLOPS | Progress: (16/20) | 12.83 s
[Task 7/25] Current/Best: 8.63/ 19.13 GFLOPS | Progress: (20/20) | 15.06 s Done.
-
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 3.53/ 10.14 GFLOPS | Progress: (4/20) | 8.87 s
[Task 8/25] Current/Best: 11.67/ 13.19 GFLOPS | Progress: (8/20) | 13.14 s
[Task 8/25] Current/Best: 15.15/ 20.83 GFLOPS | Progress: (12/20) | 20.78 s
[Task 8/25] Current/Best: 6.31/ 20.83 GFLOPS | Progress: (16/20) | 23.32 s
[Task 8/25] Current/Best: 10.32/ 20.83 GFLOPS | Progress: (20/20) | 30.79 s Done.
-
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 10.03/ 14.54 GFLOPS | Progress: (4/20) | 6.63 s
[Task 9/25] Current/Best: 16.27/ 16.27 GFLOPS | Progress: (8/20) | 10.94 s
[Task 9/25] Current/Best: 9.84/ 19.74 GFLOPS | Progress: (12/20) | 12.69 s
[Task 9/25] Current/Best: 6.86/ 19.74 GFLOPS | Progress: (16/20) | 16.62 s
[Task 9/25] Current/Best: 13.92/ 22.22 GFLOPS | Progress: (20/20) | 18.51 s Done.
-
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 3.20/ 14.79 GFLOPS | Progress: (4/20) | 7.03 s
[Task 10/25] Current/Best: 3.94/ 15.20 GFLOPS | Progress: (8/20) | 9.97 s
[Task 10/25] Current/Best: 5.54/ 15.20 GFLOPS | Progress: (12/20) | 12.31 s
[Task 10/25] Current/Best: 15.27/ 17.75 GFLOPS | Progress: (16/20) | 13.97 s
[Task 10/25] Current/Best: 14.87/ 17.75 GFLOPS | Progress: (20/20) | 16.38 s Done.
-
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 15.00/ 23.62 GFLOPS | Progress: (4/20) | 4.98 s
[Task 11/25] Current/Best: 13.50/ 23.62 GFLOPS | Progress: (8/20) | 7.54 s
[Task 11/25] Current/Best: 19.49/ 23.62 GFLOPS | Progress: (12/20) | 10.93 s
[Task 11/25] Current/Best: 8.96/ 23.62 GFLOPS | Progress: (16/20) | 14.08 s
[Task 11/25] Current/Best: 13.27/ 23.62 GFLOPS | Progress: (20/20) | 16.81 s Done.
-
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 13.95/ 13.95 GFLOPS | Progress: (4/20) | 5.46 s
[Task 12/25] Current/Best: 14.74/ 14.74 GFLOPS | Progress: (8/20) | 8.75 s
[Task 12/25] Current/Best: 11.01/ 14.74 GFLOPS | Progress: (12/20) | 12.17 s
[Task 12/25] Current/Best: 3.33/ 18.40 GFLOPS | Progress: (16/20) | 14.60 s
[Task 12/25] Current/Best: 14.64/ 18.40 GFLOPS | Progress: (20/20) | 18.10 s Done.
-
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 20.12/ 20.12 GFLOPS | Progress: (4/20) | 7.99 s
[Task 13/25] Current/Best: 17.60/ 20.12 GFLOPS | Progress: (8/20) | 11.24 s
[Task 13/25] Current/Best: 12.40/ 20.12 GFLOPS | Progress: (12/20) | 13.63 s
[Task 13/25] Current/Best: 6.07/ 20.12 GFLOPS | Progress: (16/20) | 17.58 s
[Task 13/25] Current/Best: 3.09/ 20.12 GFLOPS | Progress: (20/20) | 21.86 s Done.
-
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 15.21/ 17.41 GFLOPS | Progress: (4/20) | 5.53 s
[Task 14/25] Current/Best: 8.09/ 22.82 GFLOPS | Progress: (8/20) | 11.68 s
[Task 14/25] Current/Best: 14.51/ 22.82 GFLOPS | Progress: (12/20) | 13.86 s
[Task 14/25] Current/Best: 9.01/ 22.82 GFLOPS | Progress: (16/20) | 19.95 s
[Task 14/25] Current/Best: 9.45/ 22.82 GFLOPS | Progress: (20/20) | 22.57 s Done.
-
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 12.06/ 19.07 GFLOPS | Progress: (4/20) | 4.49 s
[Task 15/25] Current/Best: 13.38/ 19.07 GFLOPS | Progress: (8/20) | 7.40 s
[Task 15/25] Current/Best: 14.98/ 19.07 GFLOPS | Progress: (12/20) | 12.19 s
[Task 15/25] Current/Best: 10.08/ 19.07 GFLOPS | Progress: (16/20) | 17.83 s
[Task 15/25] Current/Best: 17.11/ 19.79 GFLOPS | Progress: (20/20) | 24.20 s
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 16.08/ 18.04 GFLOPS | Progress: (4/20) | 4.63 s
[Task 16/25] Current/Best: 3.18/ 18.04 GFLOPS | Progress: (8/20) | 7.71 s
[Task 16/25] Current/Best: 11.80/ 19.94 GFLOPS | Progress: (12/20) | 10.03 s
[Task 16/25] Current/Best: 15.88/ 19.94 GFLOPS | Progress: (16/20) | 13.09 s
[Task 16/25] Current/Best: 16.05/ 19.94 GFLOPS | Progress: (20/20
) | 15.46 s Done.
-
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 18.66/ 18.66 GFLOPS | Progress: (4/20) | 5.28 s
[Task 17/25] Current/Best: 4.58/ 19.12 GFLOPS | Progress: (8/20) | 8.18 s
[Task 17/25] Current/Best: 10.31/ 19.12 GFLOPS | Progress: (12/20) | 10.74 s
[Task 17/25] Current/Best: 11.34/ 19.12 GFLOPS | Progress: (16/20) | 13.91 s
[Task 17/25] Current/Best: 6.22/ 19.12 GFLOPS | Progress: (20/20) | 16.64 s Done.
-
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 4.99/ 21.47 GFLOPS | Progress: (4/20) | 8.51 s
[Task 18/25] Current/Best: 15.35/ 21.47 GFLOPS | Progress: (8/20) | 10.37 s
[Task 18/25] Current/Best: 4.89/ 21.47 GFLOPS | Progress: (12/20) | 13.16 s
[Task 18/25] Current/Best: 16.11/ 21.47 GFLOPS | Progress: (16/20) | 15.75 s
[Task 18/25] Current/Best: 17.30/ 21.47 GFLOPS | Progress: (20/20) | 18.56 s Done.
-
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 10.07/ 16.50 GFLOPS | Progress: (4/20) | 5.53 s
[Task 19/25] Current/Best: 22.54/ 22.54 GFLOPS | Progress: (8/20) | 9.60 s
[Task 19/25] Current/Best: 9.96/ 22.54 GFLOPS | Progress: (12/20) | 12.68 s
[Task 19/25] Current/Best: 12.10/ 22.54 GFLOPS | Progress: (16/20) | 15.07 s
[Task 19/25] Current/Best: 11.08/ 22.54 GFLOPS | Progress: (20/20) | 18.77 s Done.
-
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 18.47/ 18.47 GFLOPS | Progress: (4/20) | 4.89 s
[Task 20/25] Current/Best: 4.95/ 18.47 GFLOPS | Progress: (8/20) | 8.74 s
[Task 20/25] Current/Best: 9.34/ 19.46 GFLOPS | Progress: (12/20) | 13.44 s
[Task 20/25] Current/Best: 13.04/ 19.46 GFLOPS | Progress: (16/20) | 16.85 s
[Task 20/25] Current/Best: 16.08/ 19.46 GFLOPS | Progress: (20/20) | 19.63 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 20.83/ 20.83 GFLOPS | Progress: (4/20) | 4.55 s
[Task 21/25] Current/Best: 3.15/ 20.83 GFLOPS | Progress: (8/20) | 6.41 s
[Task 21/25] Current/Best: 2.17/ 20.83 GFLOPS | Progress: (12/20) | 12.97 s Done.
-
[Task 21/25] Current/Best: 22.20/ 22.20 GFLOPS | Progress: (16/20) | 16.03 s
[Task 21/25] Current/Best: 17.79/ 22.20 GFLOPS | Progress: (20/20) | 17.67 s Done.
-
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 6.61/ 10.68 GFLOPS | Progress: (4/20) | 6.32 s
[Task 22/25] Current/Best: 16.11/ 18.05 GFLOPS | Progress: (8/20) | 7.97 s
[Task 22/25] Current/Best: 5.24/ 18.05 GFLOPS | Progress: (12/20) | 12.08 s
[Task 22/25] Current/Best: 5.37/ 18.05 GFLOPS | Progress: (16/20) | 14.74 s
[Task 22/25] Current/Best: 10.45/ 19.75 GFLOPS | Progress: (20/20) | 17.11 s Done.
-
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 19.72/ 21.89 GFLOPS | Progress: (4/20) | 7.95 s
[Task 23/25] Current/Best: 8.91/ 21.89 GFLOPS | Progress: (8/20) | 13.38 s
[Task 23/25] Current/Best: 20.52/ 21.89 GFLOPS | Progress: (12/20) | 17.27 s
[Task 23/25] Current/Best: 10.64/ 21.89 GFLOPS | Progress: (16/20) | 19.89 s
[Task 23/25] Current/Best: 9.66/ 21.89 GFLOPS | Progress: (20/20) | 23.03 s Done.
-
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 9.76/ 9.76 GFLOPS | Progress: (4/20) | 11.96 s
[Task 24/25] Current/Best: 8.64/ 9.76 GFLOPS | Progress: (8/20) | 18.26 s
[Task 24/25] Current/Best: 2.32/ 9.76 GFLOPS | Progress: (12/20) | 28.93 s
[Task 24/25] Current/Best: 2.85/ 9.76 GFLOPS | Progress: (16/20) | 34.65 s
[Task 24/25] Current/Best: 4.91/ 10.23 GFLOPS | Progress: (20/20) | 45.31 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 25/25] Current/Best: 7.55/ 7.55 GFLOPS | Progress: (4/20) | 15.43 s
[Task 25/25] Current/Best: 9.37/ 9.37 GFLOPS | Progress: (8/20) | 18.22 s Done.
-
[Task 25/25] Current/Best: 5.64/ 9.37 GFLOPS | Progress: (12/20) | 20.93 s
[Task 25/25] Current/Best: 9.61/ 9.61 GFLOPS | Progress: (16/20) | 31.91 s
[Task 25/25] Current/Best: 8.32/ 9.61 GFLOPS | Progress: (20/20) | 34.52 s
+
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 1/25] Current/Best: 12.37/ 18.37 GFLOPS | Progress: (4/20) | 9.65 s
[Task 1/25] Current/Best: 19.20/ 19.20 GFLOPS | Progress: (8/20) | 14.55 s
[Task 1/25] Current/Best: 9.95/ 19.20 GFLOPS | Progress: (12/20) | 17.61 s
[Task 1/25] Current/Best: 17.60/ 19.20 GFLOPS | Progress: (16/20) | 20.33 s
[Task 1/25] Current/Best: 8.31/ 19.22 GFLOPS | Progress: (20/20) | 22.51 s Done.
+
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 8.72/ 15.43 GFLOPS | Progress: (4/20) | 4.42 s
[Task 2/25] Current/Best: 18.55/ 18.55 GFLOPS | Progress: (8/20) | 5.86 s
[Task 2/25] Current/Best: 19.20/ 19.20 GFLOPS | Progress: (12/20) | 7.36 s
[Task 2/25] Current/Best: 18.17/ 20.41 GFLOPS | Progress: (16/20) | 8.73 s
[Task 2/25] Current/Best: 12.96/ 20.41 GFLOPS | Progress: (20/20) | 10.71 s Done.
+
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 7.35/ 18.22 GFLOPS | Progress: (4/20) | 5.29 s
[Task 3/25] Current/Best: 9.52/ 21.98 GFLOPS | Progress: (8/20) | 7.83 s
[Task 3/25] Current/Best: 6.36/ 21.98 GFLOPS | Progress: (12/20) | 10.25 s
[Task 3/25] Current/Best: 5.59/ 21.98 GFLOPS | Progress: (16/20) | 12.77 s
[Task 3/25] Current/Best: 12.64/ 21.98 GFLOPS | Progress: (20/20) | 14.92 s Done.
+
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 17.65/ 17.65 GFLOPS | Progress: (4/20) | 7.14 s
[Task 4/25] Current/Best: 18.36/ 18.36 GFLOPS | Progress: (8/20) | 9.04 s
[Task 4/25] Current/Best: 13.66/ 18.36 GFLOPS | Progress: (12/20) | 11.96 s
[Task 4/25] Current/Best: 3.80/ 18.36 GFLOPS | Progress: (16/20) | 14.40 s
[Task 4/25] Current/Best: 16.66/ 18.36 GFLOPS | Progress: (20/20) | 16.54 s Done.
+
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 9.68/ 20.73 GFLOPS | Progress: (4/20) | 5.15 s
[Task 5/25] Current/Best: 15.77/ 20.73 GFLOPS | Progress: (8/20) | 7.26 s
[Task 5/25] Current/Best: 9.41/ 20.73 GFLOPS | Progress: (12/20) | 9.46 s
[Task 5/25] Current/Best: 14.66/ 20.73 GFLOPS | Progress: (16/20) | 11.68 s
[Task 5/25] Current/Best: 2.90/ 20.73 GFLOPS | Progress: (20/20) | 14.18 s Done.
+
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 17.51/ 17.51 GFLOPS | Progress: (4/20) | 4.96 s
[Task 6/25] Current/Best: 5.60/ 17.51 GFLOPS | Progress: (8/20) | 7.97 s
[Task 6/25] Current/Best: 18.08/ 18.08 GFLOPS | Progress: (12/20) | 12.51 s
[Task 6/25] Current/Best: 15.05/ 18.08 GFLOPS | Progress: (16/20) | 15.01 s
[Task 6/25] Current/Best: 11.73/ 18.08 GFLOPS | Progress: (20/20) | 18.85 s Done.
+
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 15.43/ 17.85 GFLOPS | Progress: (4/20) | 5.01 s
[Task 7/25] Current/Best: 14.06/ 17.85 GFLOPS | Progress: (8/20) | 7.17 s
[Task 7/25] Current/Best: 8.76/ 20.36 GFLOPS | Progress: (12/20) | 9.58 s
[Task 7/25] Current/Best: 3.08/ 20.36 GFLOPS | Progress: (16/20) | 13.26 s
[Task 7/25] Current/Best: 6.35/ 20.36 GFLOPS | Progress: (20/20) | 15.94 s Done.
+
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 12.88/ 15.46 GFLOPS | Progress: (4/20) | 5.51 s
[Task 8/25] Current/Best: 13.98/ 15.46 GFLOPS | Progress: (8/20) | 9.54 s
[Task 8/25] Current/Best: 10.74/ 18.33 GFLOPS | Progress: (12/20) | 12.52 s
[Task 8/25] Current/Best: 4.39/ 18.33 GFLOPS | Progress: (16/20) | 16.91 s
[Task 8/25] Current/Best: 6.81/ 18.33 GFLOPS | Progress: (20/20) | 19.56 s Done.
+
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 11.31/ 11.31 GFLOPS | Progress: (4/20) | 4.98 s
[Task 9/25] Current/Best: 13.86/ 19.60 GFLOPS | Progress: (8/20) | 8.00 s
[Task 9/25] Current/Best: 8.31/ 20.20 GFLOPS | Progress: (12/20) | 10.96 s
[Task 9/25] Current/Best: 12.57/ 20.20 GFLOPS | Progress: (16/20) | 12.89 s
[Task 9/25] Current/Best: 14.38/ 20.20 GFLOPS | Progress: (20/20) | 14.74 s Done.
+
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 5.80/ 19.39 GFLOPS | Progress: (4/20) | 4.95 s
[Task 10/25] Current/Best: 5.29/ 19.39 GFLOPS | Progress: (8/20) | 7.01 s
[Task 10/25] Current/Best: 5.50/ 19.39 GFLOPS | Progress: (12/20) | 8.83 s
[Task 10/25] Current/Best: 16.32/ 19.39 GFLOPS | Progress: (16/20) | 10.82 s
[Task 10/25] Current/Best: 5.32/ 19.39 GFLOPS | Progress: (20/20) | 12.77 s Done.
+
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 20.74/ 20.74 GFLOPS | Progress: (4/20) | 5.56 s
[Task 11/25] Current/Best: 19.40/ 20.74 GFLOPS | Progress: (8/20) | 7.67 s
[Task 11/25] Current/Best: 18.97/ 21.04 GFLOPS | Progress: (12/20) | 9.84 s
[Task 11/25] Current/Best: 17.99/ 21.04 GFLOPS | Progress: (16/20) | 13.36 s
[Task 11/25] Current/Best: 18.39/ 21.04 GFLOPS | Progress: (20/20) | 15.47 s Done.
+
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 14.13/ 14.13 GFLOPS | Progress: (4/20) | 9.50 s
[Task 12/25] Current/Best: 17.06/ 19.57 GFLOPS | Progress: (8/20) | 11.78 s
[Task 12/25] Current/Best: 15.07/ 19.57 GFLOPS | Progress: (12/20) | 18.27 s
[Task 12/25] Current/Best: 21.23/ 21.23 GFLOPS | Progress: (16/20) | 20.33 s
[Task 12/25] Current/Best: 15.55/ 21.23 GFLOPS | Progress: (20/20) | 22.75 s Done.
+
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 18.57/ 18.57 GFLOPS | Progress: (4/20) | 6.52 s
[Task 13/25] Current/Best: 20.75/ 20.75 GFLOPS | Progress: (8/20) | 10.26 s
[Task 13/25] Current/Best: 3.10/ 20.75 GFLOPS | Progress: (12/20) | 13.61 s
[Task 13/25] Current/Best: 12.45/ 20.75 GFLOPS | Progress: (16/20) | 17.58 s
[Task 13/25] Current/Best: 20.26/ 20.75 GFLOPS | Progress: (20/20) | 20.97 s Done.
+
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 5.13/ 13.22 GFLOPS | Progress: (4/20) | 6.79 s
[Task 14/25] Current/Best: 11.66/ 20.51 GFLOPS | Progress: (8/20) | 10.45 s
[Task 14/25] Current/Best: 15.90/ 20.51 GFLOPS | Progress: (12/20) | 14.25 s
[Task 14/25] Current/Best: 9.85/ 20.51 GFLOPS | Progress: (16/20) | 16.36 s
[Task 14/25] Current/Best: 3.11/ 20.51 GFLOPS | Progress: (20/20) | 20.40 s Done.
+
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 15.92/ 15.92 GFLOPS | Progress: (4/20) | 7.46 s
[Task 15/25] Current/Best: 14.46/ 15.92 GFLOPS | Progress: (8/20) | 8.78 s
[Task 15/25] Current/Best: 11.18/ 23.06 GFLOPS | Progress: (12/20) | 13.62 s
[Task 15/25] Current/Best: 15.96/ 23.06 GFLOPS | Progress: (16/20) | 16.56 s
[Task 15/25] Current/Best: 5.46/ 23.06 GFLOPS | Progress: (20/20) | 21.15 s Done.
+
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 11.74/ 17.55 GFLOPS | Progress: (4/20) | 4.35 s
[Task 16/25] Current/Best: 17.64/ 17.64 GFLOPS | Progress: (8/20) | 6.25 s
[Task 16/25] Current/Best: 14.73/ 17.64 GFLOPS | Progress: (12/20) | 8.18 s
[Task 16/25] Current/Best: 11.17/ 19.34 GFLOPS | Progress: (16/20) | 10.63 s
[Task 16/25] Current/Best: 17.56/ 19.34 GFLOPS | Progress: (20/20) | 12.23 s Done.
+
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 11.25/ 19.98 GFLOPS | Progress: (4/20) | 5.78 s
[Task 17/25] Current/Best: 21.97/ 21.97 GFLOPS | Progress: (8/20) | 9.67 s
[Task 17/25] Current/Best: 4.48/ 21.97 GFLOPS | Progress: (12/20) | 12.30 s
[Task 17/25] Current/Best: 10.27/ 21.97 GFLOPS | Progress: (16/20) | 15.36 s
[Task 17/25] Current/Best: 20.26/ 21.97 GFLOPS | Progress: (20/20) | 19.17 s Done.
+
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 10.38/ 20.80 GFLOPS | Progress: (4/20) | 5.10 s
[Task 18/25] Current/Best: 6.31/ 20.80 GFLOPS | Progress: (8/20) | 7.19 s
[Task 18/25] Current/Best: 4.46/ 20.80 GFLOPS | Progress: (12/20) | 9.66 s
[Task 18/25] Current/Best: 10.77/ 20.80 GFLOPS | Progress: (16/20) | 13.16 s
[Task 18/25] Current/Best: 9.11/ 20.80 GFLOPS | Progress: (20/20) | 19.73 s Done.
+
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 21.90/ 21.90 GFLOPS | Progress: (4/20) | 6.37 s
[Task 19/25] Current/Best: 6.16/ 21.90 GFLOPS | Progress: (8/20) | 11.65 s
[Task 19/25] Current/Best: 8.82/ 21.90 GFLOPS | Progress: (12/20) | 15.37 s
[Task 19/25] Current/Best: 1.55/ 21.90 GFLOPS | Progress: (16/20) | 20.16 s
[Task 19/25] Current/Best: 10.22/ 21.90 GFLOPS | Progress: (20/20) | 23.64 s Done.
+
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 14.76/ 18.09 GFLOPS | Progress: (4/20) | 6.22 s
[Task 20/25] Current/Best: 5.00/ 18.09 GFLOPS | Progress: (8/20) | 8.81 s
[Task 20/25] Current/Best: 10.45/ 19.97 GFLOPS | Progress: (12/20) | 13.11 s
[Task 20/25] Current/Best: 5.11/ 20.67 GFLOPS | Progress: (16/20) | 19.06 s
[Task 20/25] Current/Best: 11.74/ 20.67 GFLOPS | Progress: (20/20) | 20.41 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 10.57/ 16.95 GFLOPS | Progress: (4/20) | 6.07 s
[Task 21/25] Current/Best: 5.28/ 19.11 GFLOPS | Progress: (8/20) | 10.72 s
[Task 21/25] Current/Best: 7.49/ 19.11 GFLOPS | Progress: (12/20) | 12.28 s Done.
+
[Task 21/25] Current/Best: 13.98/ 19.11 GFLOPS | Progress: (16/20) | 15.15 s
[Task 21/25] Current/Best: 10.02/ 19.11 GFLOPS | Progress: (20/20) | 19.13 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 9.72/ 17.81 GFLOPS | Progress: (4/20) | 4.96 s
[Task 22/25] Current/Best: 17.71/ 17.81 GFLOPS | Progress: (8/20) | 7.23 s
[Task 22/25] Current/Best: 14.61/ 17.81 GFLOPS | Progress: (12/20) | 8.93 s
[Task 22/25] Current/Best: 5.13/ 19.73 GFLOPS | Progress: (16/20) | 11.58 s
[Task 22/25] Current/Best: 18.76/ 19.73 GFLOPS | Progress: (20/20) | 13.82 s Done.
+
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 19.49/ 22.22 GFLOPS | Progress: (4/20) | 5.33 s
[Task 23/25] Current/Best: 22.33/ 22.33 GFLOPS | Progress: (8/20) | 8.31 s
[Task 23/25] Current/Best: 9.90/ 22.33 GFLOPS | Progress: (12/20) | 10.94 s
[Task 23/25] Current/Best: 23.48/ 23.48 GFLOPS | Progress: (16/20) | 13.95 s
[Task 23/25] Current/Best: 8.47/ 23.48 GFLOPS | Progress: (20/20) | 17.31 s Done.
+
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 2.68/ 8.55 GFLOPS | Progress: (4/20) | 13.75 s
[Task 24/25] Current/Best: 5.99/ 8.55 GFLOPS | Progress: (8/20) | 25.68 s
[Task 24/25] Current/Best: 2.66/ 8.55 GFLOPS | Progress: (12/20) | 38.21 s
[Task 24/25] Current/Best: 2.22/ 8.55 GFLOPS | Progress: (16/20) | 48.89 s
[Task 24/25] Current/Best: 1.73/ 8.55 GFLOPS | Progress: (20/20) | 56.30 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
+
[Task 25/25] Current/Best: 1.49/ 8.87 GFLOPS | Progress: (4/20) | 13.75 s
[Task 25/25] Current/Best: 5.96/ 8.87 GFLOPS | Progress: (8/20) | 24.72 s
[Task 25/25] Current/Best: 1.55/ 8.87 GFLOPS | Progress: (12/20) | 35.70 s
[Task 25/25] Current/Best: 7.55/ 8.87 GFLOPS | Progress: (16/20) | 42.00 s
[Task 25/25] Current/Best: 3.98/ 8.87 GFLOPS | Progress: (20/20) | 44.65 s
@@ -722,8 +722,8 @@ improvement in comparing the optimized model to the unoptimized model.
.. code-block:: none
- optimized: {'mean': 408.3219417699979, 'median': 407.67783649999956, 'std': 2.9376268913300176}
- unoptimized: {'mean': 512.4494685900003, 'median': 512.0274692999942, 'std': 2.756563460026631}
+ optimized: {'mean': 422.7289286899986, 'median': 420.4029523500026, 'std': 4.736483280623968}
+ unoptimized: {'mean': 513.2705139899997, 'median': 512.6211620999982, 'std': 3.249682685153772}
@@ -746,7 +746,7 @@ profiling/benchmarking.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 12 minutes 26.943 seconds)
+ **Total running time of the script:** ( 12 minutes 18.966 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 103cf8ccee..3ff31f3a9d 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
- 1.233e-07 secs/op
+ 1.25e-07 secs/op
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index 422437c10c..6e434e93d2 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -270,7 +270,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
.. code-block:: none
- [stage(a, placeholder(a, 0xf33f610)), stage(b, placeholder(b, 0x76a2880)), stage(T_add, compute(T_add, body=[a[ax0, ax1, ax2] + b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.Range(0, 10), "DataPar", ""), T.iter_var(ax2, T.Range(0, 10), "DataPar", "")], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[a[ax0, ax1, ax2] * b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.R [...]
+ [stage(a, placeholder(a, 0x167e2620)), stage(b, placeholder(b, 0x8064040)), 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 84e5877a61..cf2fc3a775 100644
--- a/docs/_sources/tutorial/sg_execution_times.rst.txt
+++ b/docs/_sources/tutorial/sg_execution_times.rst.txt
@@ -5,24 +5,24 @@
Computation times
=================
-**16:23.345** total execution time for **tutorial** files:
+**16:23.166** total execution time for **tutorial** files:
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 12:26.943 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 12:18.966 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:44.840 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:56.214 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 00:59.345 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 01:00.650 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:36.990 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:36.718 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:32.420 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:28.181 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:01.779 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:01.407 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``) | 00:00.858 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.170 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.172 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_tutorial_uma.py` (``uma.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 27ee738256..725d972d7f 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -285,8 +285,8 @@ helper function to run a profile of the TVM generated code.
.. code-block:: none
- Numpy running time: 0.000008
- naive: 0.000007
+ Numpy running time: 0.000007
+ naive: 0.000008
@@ -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: 0.000026
+ vector: 0.000025
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -498,10 +498,10 @@ We can now compare the different schedules
.. code-block:: none
Operator Timing Performance
- numpy 8.111570000437496e-06 1.0
- naive 6.6413e-06 0.8187440901874486
- parallel 6.885799999999999e-06 0.848886220500916
- vector 2.6319599999999998e-05 3.2446986216700906
+ numpy 7.279910000761447e-06 1.0
+ naive 8.2921e-06 1.1390388066792971
+ parallel 6.2142000000000006e-06 0.853609453873746
+ vector 2.4608300000000003e-05 3.3803027781148502
@@ -922,7 +922,7 @@ matrix multiplication.
.. code-block:: none
- Numpy running time: 0.018664
+ Numpy running time: 0.018311
@@ -980,7 +980,7 @@ optimizations.
.. code-block:: none
- none: 3.276021
+ none: 3.387673
@@ -1080,7 +1080,7 @@ schedule.
.. code-block:: none
- blocking: 0.306694
+ blocking: 0.301096
@@ -1164,7 +1164,7 @@ already cache friendly from our previous optimizations.
.. code-block:: none
- vectorization: 0.336994
+ vectorization: 0.336335
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -1230,7 +1230,7 @@ more cache friendly.
.. code-block:: none
- loop permutation: 0.118392
+ loop permutation: 0.116298
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -1321,7 +1321,7 @@ optimized schedule.
.. code-block:: none
- array packing: 0.109105
+ array packing: 0.108435
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -1404,7 +1404,7 @@ to `C` when all the block results are ready.
.. code-block:: none
- block caching: 0.110114
+ block caching: 0.110452
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -1478,7 +1478,7 @@ of thread-level parallelization.
.. code-block:: none
- parallelization: 0.145534
+ parallelization: 0.146169
# from tvm.script import ir as I
# from tvm.script import tir as T
@@ -1548,13 +1548,13 @@ working, we can compare the results.
.. code-block:: none
Operator Timing Performance
- none 3.2760207547 1.0
- blocking 0.30669357340000003 0.09361771379500473
- vectorization 0.3369942954 0.10286695983733474
- loop permutation 0.118392466 0.03613910743091178
- array packing 0.1091052999 0.03330421510409242
- block caching 0.11011446630000002 0.03361226150415023
- parallelization 0.1455338389 0.04442396730582593
+ none 3.3876733566000006 1.0
+ blocking 0.3010964253 0.08888000512605265
+ vectorization 0.33633455749999996 0.09928187345593387
+ loop permutation 0.1162978298 0.03432970583584274
+ array packing 0.1084351938 0.03200875125364204
+ block caching 0.1104521714 0.03260413852616949
+ parallelization 0.1461686477 0.04314720822042314
@@ -1594,6 +1594,11 @@ operations with tunable parameters that allows you to automatically optimize
the computation for specific platforms.
+.. rst-class:: sphx-glr-timing
+
+ **Total running time of the script:** ( 1 minutes 0.650 seconds)
+
+
.. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
.. only:: html
diff --git a/docs/commit_hash b/docs/commit_hash
index 91247786a2..868eb3e419 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-2c052b20670087133cb598913692e026e94718ff
+99a5734a9ec25d5fc263fd1b61fc2687c83a2f92
diff --git a/docs/genindex.html b/docs/genindex.html
index 21e6fa6593..503269e86e 100644
--- a/docs/genindex.html
+++ b/docs/genindex.html
@@ -2723,6 +2723,8 @@
</li>
</ul></li>
<li><a href="reference/api/python/contrib.html#tvm.contrib.pickle_memoize.memoize">memoize() (in module tvm.contrib.pickle_memoize)</a>
+</li>
+ <li><a href="reference/api/python/tir.html#tvm.tir.Schedule.merge">merge() (tvm.tir.Schedule method)</a>
</li>
<li><a href="reference/api/python/relay/transform.html#tvm.relay.transform.MergeCompilerRegions">MergeCompilerRegions() (in module tvm.relay.transform)</a>
</li>
diff --git a/docs/how_to/compile_models/from_darknet.html b/docs/how_to/compile_models/from_darknet.html
index 54a8f97c06..e0ce46ddf1 100644
--- a/docs/how_to/compile_models/from_darknet.html
+++ b/docs/how_to/compile_models/from_darknet.html
@@ -590,7 +590,7 @@ 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 23.229 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 20.630 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_mxnet.html b/docs/how_to/compile_models/from_mxnet.html
index 4058ce2d67..352f1fc7de 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -444,7 +444,7 @@
<span class="nb">print</span><span class="p">(</span><span class="s2">"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.zip14e69f2c-37c2-46c8-9f51-31c44101d8e9 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.zipfae97a33-2c77-494d-bc7a-8f359693ab8c 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 a12555fa95..b4b76d42ca 100644
--- a/docs/how_to/compile_models/from_oneflow.html
+++ b/docs/how_to/compile_models/from_oneflow.html
@@ -454,13 +454,13 @@ 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, 55.6MB/s]
- 35%|###4 | 14.3M/41.5M [00:00<00:00, 46.3MB/s]
- 51%|##### | 21.0M/41.5M [00:00<00:00, 54.5MB/s]
- 64%|######3 | 26.5M/41.5M [00:00<00:00, 41.3MB/s]
- 77%|#######7 | 32.0M/41.5M [00:00<00:00, 44.6MB/s]
- 92%|#########2| 38.3M/41.5M [00:00<00:00, 42.2MB/s]
-100%|##########| 41.5M/41.5M [00:00<00:00, 45.1MB/s]
+ 19%|#9 | 7.99M/41.5M [00:00<00:00, 51.3MB/s]
+ 35%|###4 | 14.3M/41.5M [00:00<00:00, 58.3MB/s]
+ 48%|####8 | 20.1M/41.5M [00:00<00:00, 49.8MB/s]
+ 60%|###### | 25.0M/41.5M [00:00<00:00, 43.8MB/s]
+ 77%|#######7 | 32.0M/41.5M [00:00<00:00, 44.9MB/s]
+ 92%|#########2| 38.3M/41.5M [00:00<00:00, 50.1MB/s]
+100%|##########| 41.5M/41.5M [00:00<00:00, 48.0MB/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 6ea00c32c1..e92c88599a 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -437,12 +437,10 @@ 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 | 7.99M/44.7M [00:00<00:00, 65.8MB/s]
- 46%|####5 | 20.4M/44.7M [00:00<00:00, 99.7MB/s]
- 68%|######7 | 30.3M/44.7M [00:00<00:00, 57.6MB/s]
- 83%|########3 | 37.3M/44.7M [00:00<00:00, 50.6MB/s]
- 96%|#########6| 43.0M/44.7M [00:00<00:00, 48.6MB/s]
-100%|##########| 44.7M/44.7M [00:00<00:00, 49.0MB/s]
+ 5%|4 | 2.05M/44.7M [00:00<00:02, 18.5MB/s]
+ 54%|#####3 | 24.0M/44.7M [00:00<00:00, 133MB/s]
+ 83%|########3 | 37.2M/44.7M [00:00<00:00, 118MB/s]
+100%|##########| 44.7M/44.7M [00:00<00:00, 108MB/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 2e7010433e..219fe23e9d 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -657,7 +657,7 @@ 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 31.647 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 31.229 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 bf1bba9114..9622ccd08f 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-compile-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>06:56.667</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>06:49.351</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 81%" />
@@ -354,43 +354,43 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="from_tensorflow.html#sphx-glr-how-to-compile-models-from-tensorflow-py"><span class="std std-ref">Compile Tensorflow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tensorflow.py</span></code>)</p></td>
-<td><p>01:31.647</p></td>
+<td><p>01:31.229</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>01:23.229</p></td>
+<td><p>01:20.630</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>00:57.563</p></td>
+<td><p>00:56.683</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_oneflow.html#sphx-glr-how-to-compile-models-from-oneflow-py"><span class="std std-ref">Compile OneFlow Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_oneflow.py</span></code>)</p></td>
-<td><p>00:38.247</p></td>
+<td><p>00:38.067</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="from_coreml.html#sphx-glr-how-to-compile-models-from-coreml-py"><span class="std std-ref">Compile CoreML Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_coreml.py</span></code>)</p></td>
-<td><p>00:32.859</p></td>
+<td><p>00:32.357</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_mxnet.html#sphx-glr-how-to-compile-models-from-mxnet-py"><span class="std std-ref">Compile MXNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_mxnet.py</span></code>)</p></td>
-<td><p>00:30.582</p></td>
+<td><p>00:30.190</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>00:28.678</p></td>
+<td><p>00:28.589</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>00:26.444</p></td>
+<td><p>00:26.151</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>00:24.681</p></td>
+<td><p>00:22.759</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>00:02.738</p></td>
+<td><p>00:02.696</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 9ac610598b..e3c72defb8 100644
--- a/docs/how_to/deploy_models/deploy_model_on_adreno.html
+++ b/docs/how_to/deploy_models/deploy_model_on_adreno.html
@@ -799,7 +799,7 @@ Top5 predictions:
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 2753.6753 2754.0597 2756.0955 2750.8269 1.7248
+ 2539.4006 2538.4581 2544.2771 2536.9132 2.1375
</pre></div>
</div>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-model-on-adreno-py">
diff --git a/docs/how_to/deploy_models/deploy_model_on_adreno_tvmc.html b/docs/how_to/deploy_models/deploy_model_on_adreno_tvmc.html
index 384b68a2a5..31bd13709c 100644
--- a/docs/how_to/deploy_models/deploy_model_on_adreno_tvmc.html
+++ b/docs/how_to/deploy_models/deploy_model_on_adreno_tvmc.html
@@ -443,32 +443,25 @@ to run this tutorial with a real device over rpc.</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading data from https://storage.googleapis.com/tensorflow/keras-applications/resnet/resnet50_weights_tf_dim_ordering_tf_kernels.h5
8192/102967424 [..............................] - ETA: 0s
- 2187264/102967424 [..............................] - ETA: 3s
- 8380416/102967424 [=>............................] - ETA: 1s
- 10575872/102967424 [==>...........................] - ETA: 2s
- 16769024/102967424 [===>..........................] - ETA: 2s
- 25157632/102967424 [======>.......................] - ETA: 1s
- 32481280/102967424 [========>.....................] - ETA: 1s
- 33546240/102967424 [========>.....................] - ETA: 1s
- 40189952/102967424 [==========>...................] - ETA: 1s
- 41934848/102967424 [===========>..................] - ETA: 1s
- 48578560/102967424 [=============>................] - ETA: 1s
- 50323456/102967424 [=============>................] - ETA: 1s
- 56967168/102967424 [===============>..............] - ETA: 0s
- 58712064/102967424 [================>.............] - ETA: 0s
- 65355776/102967424 [==================>...........] - ETA: 0s
- 67100672/102967424 [==================>...........] - ETA: 0s
- 69296128/102967424 [===================>..........] - ETA: 0s
- 75489280/102967424 [====================>.........] - ETA: 0s
- 82124800/102967424 [======================>.......] - ETA: 0s
- 83877888/102967424 [=======================>......] - ETA: 0s
- 90521600/102967424 [=========================>....] - ETA: 0s
+ 11804672/102967424 [==>...........................] - ETA: 0s
+ 16769024/102967424 [===>..........................] - ETA: 0s
+ 23011328/102967424 [=====>........................] - ETA: 0s
+ 28106752/102967424 [=======>......................] - ETA: 0s
+ 33120256/102967424 [========>.....................] - ETA: 0s
+ 38223872/102967424 [==========>...................] - ETA: 0s
+ 43196416/102967424 [===========>..................] - ETA: 0s
+ 48300032/102967424 [=============>................] - ETA: 0s
+ 50987008/102967424 [=============>................] - ETA: 0s
+ 58220544/102967424 [===============>..............] - ETA: 0s
+ 63324160/102967424 [=================>............] - ETA: 0s
+ 68296704/102967424 [==================>...........] - ETA: 0s
+ 73334784/102967424 [====================>.........] - ETA: 0s
+ 78381056/102967424 [=====================>........] - ETA: 0s
+ 83329024/102967424 [=======================>......] - ETA: 0s
+ 88416256/102967424 [========================>.....] - ETA: 0s
92266496/102967424 [=========================>....] - ETA: 0s
- 95059968/102967424 [==========================>...] - ETA: 0s
- 98910208/102967424 [===========================>..] - ETA: 0s
-100646912/102967424 [============================>.] - ETA: 0s
-102850560/102967424 [============================>.] - ETA: 0s
-102967424/102967424 [==============================] - 2s 0us/step
+ 99287040/102967424 [===========================>..] - ETA: 0s
+102967424/102967424 [==============================] - 1s 0us/step
</pre></div>
</div>
</div>
diff --git a/docs/how_to/deploy_models/deploy_model_on_android.html b/docs/how_to/deploy_models/deploy_model_on_android.html
index ecb70a3d3c..fe24c9b082 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -667,7 +667,7 @@ to the remote android device.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 15.9861 15.8256 17.0500 15.6477 0.4107
+ 15.9646 15.7778 16.5501 15.6210 0.3490
</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 fcb33e40d7..a38b899306 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -459,36 +459,36 @@ 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<00:04, 41.0MB/s]
- 8%|8 | 14.3M/170M [00:00<00:04, 38.0MB/s]
- 11%|# | 17.9M/170M [00:00<00:04, 36.5MB/s]
- 13%|#3 | 22.3M/170M [00:00<00:04, 35.6MB/s]
- 15%|#5 | 25.7M/170M [00:00<00:04, 35.1MB/s]
- 19%|#8 | 32.0M/170M [00:00<00:03, 36.7MB/s]
- 24%|##3 | 40.0M/170M [00:01<00:03, 44.9MB/s]
- 27%|##7 | 46.3M/170M [00:01<00:02, 45.5MB/s]
- 30%|##9 | 50.6M/170M [00:01<00:03, 35.4MB/s]
- 33%|###2 | 56.0M/170M [00:01<00:03, 37.6MB/s]
- 38%|###7 | 64.0M/170M [00:01<00:02, 41.2MB/s]
- 42%|####2 | 72.0M/170M [00:01<00:02, 46.3MB/s]
- 46%|####6 | 78.3M/170M [00:01<00:01, 49.5MB/s]
- 49%|####8 | 83.2M/170M [00:02<00:01, 46.2MB/s]
- 52%|#####1 | 87.7M/170M [00:02<00:01, 44.8MB/s]
- 54%|#####4 | 92.0M/170M [00:02<00:02, 39.0MB/s]
- 57%|#####6 | 96.0M/170M [00:02<00:02, 33.6MB/s]
- 60%|###### | 102M/170M [00:02<00:01, 36.3MB/s]
- 62%|######2 | 106M/170M [00:02<00:01, 34.5MB/s]
- 66%|######5 | 112M/170M [00:02<00:01, 36.4MB/s]
- 71%|####### | 120M/170M [00:03<00:01, 43.8MB/s]
- 73%|#######3 | 124M/170M [00:03<00:01, 38.1MB/s]
- 75%|#######5 | 128M/170M [00:03<00:01, 35.7MB/s]
- 80%|######## | 136M/170M [00:03<00:00, 35.9MB/s]
- 85%|########4 | 144M/170M [00:03<00:00, 42.9MB/s]
- 89%|########8 | 151M/170M [00:03<00:00, 49.6MB/s]
- 92%|#########1| 156M/170M [00:04<00:00, 46.5MB/s]
- 95%|#########4| 161M/170M [00:04<00:00, 42.0MB/s]
- 98%|#########7| 166M/170M [00:04<00:00, 42.8MB/s]
-100%|##########| 170M/170M [00:04<00:00, 41.1MB/s]
+ 5%|4 | 7.99M/170M [00:00<00:03, 53.3MB/s]
+ 9%|9 | 16.0M/170M [00:00<00:03, 50.9MB/s]
+ 14%|#4 | 24.0M/170M [00:00<00:03, 49.8MB/s]
+ 18%|#7 | 30.3M/170M [00:00<00:02, 50.5MB/s]
+ 21%|## | 35.1M/170M [00:00<00:02, 48.6MB/s]
+ 24%|##3 | 40.0M/170M [00:00<00:02, 46.6MB/s]
+ 26%|##6 | 44.4M/170M [00:00<00:03, 42.5MB/s]
+ 29%|##8 | 48.4M/170M [00:01<00:03, 38.9MB/s]
+ 33%|###2 | 56.0M/170M [00:01<00:02, 41.5MB/s]
+ 37%|###6 | 62.3M/170M [00:01<00:02, 44.8MB/s]
+ 41%|#### | 69.4M/170M [00:01<00:02, 52.0MB/s]
+ 44%|####3 | 74.6M/170M [00:01<00:02, 47.0MB/s]
+ 47%|####6 | 79.2M/170M [00:01<00:02, 44.8MB/s]
+ 49%|####9 | 83.6M/170M [00:01<00:02, 37.2MB/s]
+ 52%|#####1 | 88.1M/170M [00:02<00:02, 39.4MB/s]
+ 57%|#####6 | 96.0M/170M [00:02<00:01, 48.0MB/s]
+ 61%|######1 | 104M/170M [00:02<00:01, 44.2MB/s]
+ 65%|######5 | 111M/170M [00:02<00:01, 50.0MB/s]
+ 68%|######8 | 116M/170M [00:02<00:01, 49.5MB/s]
+ 72%|#######1 | 122M/170M [00:02<00:01, 49.5MB/s]
+ 75%|#######4 | 127M/170M [00:02<00:00, 45.3MB/s]
+ 77%|#######7 | 131M/170M [00:03<00:00, 42.0MB/s]
+ 80%|######## | 136M/170M [00:03<00:00, 42.3MB/s]
+ 82%|########2 | 140M/170M [00:03<00:00, 40.8MB/s]
+ 85%|########4 | 144M/170M [00:03<00:00, 32.4MB/s]
+ 88%|########8 | 150M/170M [00:03<00:00, 38.4MB/s]
+ 91%|######### | 154M/170M [00:03<00:00, 37.2MB/s]
+ 94%|#########4| 160M/170M [00:03<00:00, 39.9MB/s]
+ 99%|#########8| 167M/170M [00:03<00:00, 46.9MB/s]
+100%|##########| 170M/170M [00:04<00:00, 44.4MB/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=& [...]
@@ -586,7 +586,7 @@ 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 36.406 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes 35.059 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 32b44d6cf3..7b06db4b62 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -500,8 +500,8 @@ training. Other models require a full post training calibration.</p>
Downloading: "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, 51.6MB/s]
-100%|##########| 13.6M/13.6M [00:00<00:00, 71.3MB/s]
+ 59%|#####8 | 7.99M/13.6M [00:00<00:00, 74.9MB/s]
+100%|##########| 13.6M/13.6M [00:00<00:00, 103MB/s]
</pre></div>
</div>
</div>
@@ -592,7 +592,7 @@ output values are identical out of 1000 outputs from mobilenet v2.</p>
</div>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 90.0879 90.0527 91.3993 89.8019 0.2330
+ 90.5942 90.5546 96.1535 89.9876 0.6161
</pre></div>
</div>
<div class="admonition note">
@@ -631,7 +631,7 @@ This includes support for the VNNI 8 bit dot product instruction (CascadeLake or
<div class="section" id="deploy-a-quantized-tflite-model">
<h2>Deploy a quantized TFLite Model<a class="headerlink" href="#deploy-a-quantized-tflite-model" title="Permalink to this headline">¶</a></h2>
<p>TODO</p>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 16.787 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 16.326 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 b8a4a3667f..6702f363c3 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -585,7 +585,7 @@ TFLite Top-5 labels: [387 102 386 341 349]
</div>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 119.4720 119.4017 121.8871 117.6485 0.8324
+ 119.6478 119.6423 121.0938 118.5766 0.3817
</pre></div>
</div>
<div class="admonition note">
@@ -613,7 +613,7 @@ network for ARM CPU</span></a>.</p></li>
</ul>
</div></blockquote>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 30.077 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 35.885 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 bd6c2b4bca..9eb481caa8 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -526,7 +526,7 @@ for calibration. But the accuracy might be impacted.</p>
DeprecationWarning,
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 46.660 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 49.286 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 338a124c56..4db78aa214 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -468,24 +468,28 @@ 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]
- 1%| | 1234/132723 [00:00<00:10, 12287.69KB/s]
- 3%|3 | 3994/132723 [00:00<00:06, 21278.05KB/s]
- 8%|7 | 9986/132723 [00:00<00:03, 38901.60KB/s]
- 14%|#3 | 18556/132723 [00:00<00:01, 57363.52KB/s]
- 21%|## | 27223/132723 [00:00<00:01, 67924.57KB/s]
- 26%|##5 | 34018/132723 [00:00<00:01, 56581.03KB/s]
- 32%|###2 | 42764/132723 [00:00<00:01, 65470.69KB/s]
- 39%|###8 | 51506/132723 [00:00<00:01, 71866.40KB/s]
- 45%|####5 | 60239/132723 [00:00<00:00, 76408.92KB/s]
- 51%|#####1 | 68252/132723 [00:01<00:00, 77507.40KB/s]
- 58%|#####8 | 77015/132723 [00:01<00:00, 80511.69KB/s]
- 64%|######4 | 85182/132723 [00:01<00:00, 75087.72KB/s]
- 71%|####### | 93981/132723 [00:01<00:00, 78735.34KB/s]
- 77%|#######6 | 101984/132723 [00:01<00:00, 61049.47KB/s]
- 83%|########3 | 110765/132723 [00:01<00:00, 67496.49KB/s]
- 89%|########9 | 118147/132723 [00:01<00:00, 68088.06KB/s]
- 96%|#########5| 126956/132723 [00:01<00:00, 73382.43KB/s]
-100%|##########| 132723/132723 [00:01<00:00, 67668.32KB/s]
+ 1%|1 | 1364/132723 [00:00<00:09, 13592.63KB/s]
+ 3%|3 | 4324/132723 [00:00<00:05, 22944.10KB/s]
+ 8%|8 | 10833/132723 [00:00<00:02, 42148.08KB/s]
+ 12%|#2 | 16244/132723 [00:00<00:02, 46022.01KB/s]
+ 16%|#5 | 20839/132723 [00:00<00:02, 43286.87KB/s]
+ 20%|#9 | 26540/132723 [00:00<00:02, 47728.24KB/s]
+ 25%|##4 | 32711/132723 [00:00<00:01, 52148.10KB/s]
+ 30%|### | 40468/132723 [00:00<00:01, 60057.65KB/s]
+ 35%|###5 | 46518/132723 [00:00<00:01, 55275.22KB/s]
+ 39%|###9 | 52144/132723 [00:01<00:01, 50337.60KB/s]
+ 43%|####3 | 57418/132723 [00:01<00:01, 50987.76KB/s]
+ 49%|####8 | 64707/132723 [00:01<00:01, 57104.03KB/s]
+ 55%|#####4 | 72608/132723 [00:01<00:00, 63350.24KB/s]
+ 60%|#####9 | 79069/132723 [00:01<00:00, 62766.84KB/s]
+ 64%|######4 | 85433/132723 [00:01<00:00, 60001.84KB/s]
+ 69%|######9 | 91860/132723 [00:01<00:00, 61200.41KB/s]
+ 76%|#######5 | 100413/132723 [00:01<00:00, 68192.86KB/s]
+ 82%|########2 | 109048/132723 [00:01<00:00, 73476.59KB/s]
+ 88%|########7 | 116711/132723 [00:02<00:00, 74400.89KB/s]
+ 94%|#########3| 124203/132723 [00:02<00:00, 68711.59KB/s]
+ 99%|#########8| 131193/132723 [00:02<00:00, 66494.46KB/s]
+100%|##########| 132723/132723 [00:02<00:00, 58396.11KB/s]
</pre></div>
</div>
<p>Create TVM runtime and do inference
@@ -524,7 +528,7 @@ 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 53.000 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 50.099 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 c5057653a7..f8e751eba0 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -345,52 +345,52 @@
<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>16:32.793</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>16:32.551</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
<table class="docutils align-default">
<colgroup>
-<col style="width: 85%" />
-<col style="width: 9%" />
+<col style="width: 86%" />
+<col style="width: 8%" />
<col style="width: 6%" />
</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>03:52.1000</p></td>
+<td><p>03:50.099</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>03:36.406</p></td>
+<td><p>03:35.059</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>02:30.077</p></td>
+<td><p>02:35.885</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>01:46.660</p></td>
+<td><p>01:49.286</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>01:16.787</p></td>
+<td><p>01:16.326</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="deploy_model_on_adreno.html#sphx-glr-how-to-deploy-models-deploy-model-on-adreno-py"><span class="std std-ref">Deploy the Pretrained Model on Adreno™</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_adreno.py</span></code>)</p></td>
-<td><p>00:57.626</p></td>
+<td><p>00:54.784</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_model_on_adreno_tvmc.html#sphx-glr-how-to-deploy-models-deploy-model-on-adreno-tvmc-py"><span class="std std-ref">Deploy the Pretrained Model on Adreno™ with tvmc Interface</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_adreno_tvmc.py</span></code>)</p></td>
-<td><p>00:52.671</p></td>
+<td><p>00:51.258</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="deploy_model_on_android.html#sphx-glr-how-to-deploy-models-deploy-model-on-android-py"><span class="std std-ref">Deploy the Pretrained Model on Android</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_android.py</span></code>)</p></td>
-<td><p>00:43.241</p></td>
+<td><p>00:42.669</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_model_on_nano.html#sphx-glr-how-to-deploy-models-deploy-model-on-nano-py"><span class="std std-ref">Deploy the Pretrained Model on Jetson Nano</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_nano.py</span></code>)</p></td>
-<td><p>00:28.298</p></td>
+<td><p>00:28.845</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="deploy_model_on_rasp.html#sphx-glr-how-to-deploy-models-deploy-model-on-rasp-py"><span class="std std-ref">Deploy the Pretrained Model on Raspberry Pi</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_rasp.py</span></code>)</p></td>
-<td><p>00:28.021</p></td>
+<td><p>00:28.335</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_sparse.html#sphx-glr-how-to-deploy-models-deploy-sparse-py"><span class="std std-ref">Deploy a Hugging Face Pruned Model on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_sparse.py</span></code>)</p></td>
diff --git a/docs/how_to/extend_tvm/bring_your_own_datatypes.html b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
index e162931fc1..96f0e8a32d 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -624,7 +624,7 @@ In this alpha state of the Bring Your Own Datatypes framework, we have not imple
<span class="n">module</span><span class="p">,</span> <a href="https://docs.python.org/3/library/stdtypes.html#dict" title="builtins.dict" class="sphx-glr-backref-module-builtins sphx-glr-backref-type-py-class sphx-glr-backref-instance"><span class="n">params</span></a> <span class="o">=</span> <span class="n">get_mobilenet</span><span class="p">()</span>
</pre></div>
</div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip3f4906e1-27a1-487c-bed5-2ab2f9a53345 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.zip9c87e86e-7778-429c-ab44-ef7922d9f5b9 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 563b468671..68746c7b83 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-extend-tvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:55.065</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:54.287</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -354,19 +354,19 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="bring_your_own_datatypes.html#sphx-glr-how-to-extend-tvm-bring-your-own-datatypes-py"><span class="std std-ref">Bring Your Own Datatypes to TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">bring_your_own_datatypes.py</span></code>)</p></td>
-<td><p>00:51.171</p></td>
+<td><p>00:50.477</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>00:02.796</p></td>
+<td><p>00:02.726</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>00:01.089</p></td>
+<td><p>00:01.077</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="low_level_custom_pass.html#sphx-glr-how-to-extend-tvm-low-level-custom-pass-py"><span class="std std-ref">Writing a Customized Pass</span></a> (<code class="docutils literal notranslate"><span class="pre">low_level_custom_pass.py</span></code>)</p></td>
-<td><p>00:00.008</p></td>
+<td><p>00:00.007</p></td>
<td><p>0.0 MB</p></td>
</tr>
</tbody>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index 9b035c61be..42908c1e0d 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -531,10 +531,10 @@ profile the execution time of each passes.</p>
</pre></div>
</div>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 22227us [22227us] (48.46%; 48.46%)
-FoldScaleAxis: 23637us [6us] (51.54%; 51.54%)
- FoldConstant: 23631us [1706us] (51.52%; 99.97%)
- InferType: 21925us [21925us] (47.80%; 92.78%)
+InferType: 22449us [22449us] (48.08%; 48.08%)
+FoldScaleAxis: 24238us [8us] (51.92%; 51.92%)
+ FoldConstant: 24230us [1734us] (51.90%; 99.97%)
+ InferType: 22496us [22496us] (48.18%; 92.84%)
</pre></div>
</div>
</div>
@@ -556,10 +556,10 @@ Refer to following sections and <a class="reference internal" href="../../refere
</pre></div>
</div>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 21698us [21698us] (47.75%; 47.75%)
-FoldScaleAxis: 23747us [5us] (52.25%; 52.25%)
- FoldConstant: 23741us [1707us] (52.24%; 99.98%)
- InferType: 22034us [22034us] (48.49%; 92.81%)
+InferType: 21872us [21872us] (48.21%; 48.21%)
+FoldScaleAxis: 23500us [6us] (51.79%; 51.79%)
+ FoldConstant: 23494us [1731us] (51.78%; 99.97%)
+ InferType: 21762us [21762us] (47.96%; 92.63%)
</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 57cda9e0e2..8390ef6db8 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -580,7 +580,7 @@ latency of convolution.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 53.507839 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 36.583423 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 aa751bd8ad..26fd31e89a 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -862,7 +862,7 @@ be able to run on our build server</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 12.246058 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 11.332147 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 4028bc5712..6593e04448 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -477,8 +477,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.017962
-Baseline: 3.255473
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.018174
+Baseline: 3.442605
</pre></div>
</div>
<p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -537,7 +537,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.301741
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.296080
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -594,7 +594,7 @@ vastly.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.341142
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.328482
</pre></div>
</div>
<p>Here is the generated IR after vectorization.</p>
@@ -649,7 +649,7 @@ the access pattern for A matrix is more cache friendly.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.120074
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.116717
</pre></div>
</div>
<p>Here is the generated IR after loop permutation.</p>
@@ -726,7 +726,7 @@ flattening.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.109988
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.109713
</pre></div>
</div>
<p>Here is the generated IR after array packing.</p>
@@ -804,7 +804,7 @@ write to C when all the block results are ready.</p>
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.112287
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.112274
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -884,7 +884,7 @@ class Module:
<span class="nb">print</span><span class="p">(</span><span class="s2">"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: 0.146727
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.147084
</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 2a6f8f3186..462f68ddd2 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-optimize-operators-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:34.963</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:35.296</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -354,15 +354,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="opt_gemm.html#sphx-glr-how-to-optimize-operators-opt-gemm-py"><span class="std std-ref">How to optimize GEMM on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_gemm.py</span></code>)</p></td>
-<td><p>00:31.996</p></td>
+<td><p>00:32.341</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>00:01.859</p></td>
+<td><p>00:01.891</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>00:01.108</p></td>
+<td><p>00:01.063</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 1c38406d48..c1c1ca4ce5 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-tune-with-autoscheduler-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>10:28.765</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>09:59.269</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 85%" />
@@ -354,27 +354,27 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_conv2d_layer_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-conv2d-layer-cuda-py"><span class="std std-ref">Auto-scheduling a Convolution Layer for GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_layer_cuda.py</span></code>)</p></td>
-<td><p>06:21.426</p></td>
+<td><p>06:05.984</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>01:43.370</p></td>
+<td><p>01:42.224</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>01:11.382</p></td>
+<td><p>01:10.847</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>00:44.654</p></td>
+<td><p>00:32.532</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_network_arm.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-arm-py"><span class="std std-ref">Auto-scheduling a Neural Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_arm.py</span></code>)</p></td>
-<td><p>00:14.253</p></td>
+<td><p>00:14.090</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="tune_network_mali.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-mali-py"><span class="std std-ref">Auto-scheduling a Neural Network for mali GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_mali.py</span></code>)</p></td>
-<td><p>00:13.680</p></td>
+<td><p>00:13.591</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 0aff78251b..ec25e5c826 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
@@ -510,11 +510,11 @@ class Module:
@T.prim_func
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": T.bool(True), "global_symbol": "main", "tir.noalias": T.bool(True)})
- blockIdx_x = T.launch_thread("blockIdx.x", 16)
+ blockIdx_x = T.launch_thread("blockIdx.x", 28)
conv2d_nchw = T.allocate([14], "float32", "local")
- pad_temp_shared = T.allocate([1568], "float32", "shared")
- kernel_shared = T.allocate([1024], "float32", "shared")
- threadIdx_x = T.launch_thread("threadIdx.x", 112)
+ pad_temp_shared = T.allocate([72], "float32", "shared")
+ kernel_shared = T.allocate([3072], "float32", "shared")
+ threadIdx_x = T.launch_thread("threadIdx.x", 64)
conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
@@ -530,516 +530,459 @@ class Module:
conv2d_nchw_1[11] = T.float32(0)
conv2d_nchw_1[12] = T.float32(0)
conv2d_nchw_1[13] = T.float32(0)
- for rc_outer_outer, ry_outer_outer, rx_outer_outer in T.grid(16, 3, 3):
- cse_var_2: T.int32 = rc_outer_outer * 288
+ for rc_outer_outer, ry_outer_outer in T.grid(64, 3):
+ cse_var_2: T.int32 = rc_outer_outer * 72
cse_var_1: T.int32 = ry_outer_outer * 3
+ pad_temp_shared_1 = T.Buffer((72,), data=pad_temp_shared, scope="shared")
+ with T.launch_thread("threadIdx.x", 64) as threadIdx_x_1:
+ data_1 = T.Buffer((25088,), data=data.data)
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= threadIdx_x_1 * 4 % 9 and threadIdx_x_1 * 4 % 9 < 8, data_1[rc_outer_outer * 392 + threadIdx_x_1 * 4 // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + threadIdx_x_1 * 4 % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 1] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 1) % 9 and (threadIdx_x_1 * 4 + 1) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 1) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 1) % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 2] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 2) % 9 and (threadIdx_x_1 * 4 + 2) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 2) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 2) % 9 - 8], T.float32(0))
+ if T.likely(threadIdx_x_1 < 18):
+ pad_temp_shared_1[threadIdx_x_1 * 4 + 3] = T.if_then_else(1 <= ry_outer_outer + blockIdx_x % 7 and ry_outer_outer + blockIdx_x % 7 < 8 and 1 <= (threadIdx_x_1 * 4 + 3) % 9 and (threadIdx_x_1 * 4 + 3) % 9 < 8, data_1[rc_outer_outer * 392 + (threadIdx_x_1 * 4 + 3) // 9 * 49 + ry_outer_outer * 7 + blockIdx_x % 7 * 7 + (threadIdx_x_1 * 4 + 3) % 9 - 8], T.float32(0))
threadIdx_x_1 = T.env_thread("threadIdx.x")
- pad_temp_shared_1 = T.Buffer((1568,), data=pad_temp_shared, scope="shared")
- data_1 = T.Buffer((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 % 49 // 7 + ry_outer_outer and threadIdx_x_1 % 49 // 7 + ry_outer_outer < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer - 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 <= ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 104], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 216], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 328], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 440], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 560] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 552], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 672] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 664], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 784] = T.if_then_else(1 <= threadIdx_x_1 % 49 // 7 + ry_outer_outer and threadIdx_x_1 % 49 // 7 + ry_outer_outer < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 776], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 896] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 2) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 888], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1008] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 4) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1000], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1120] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 6) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1112], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1232] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 1) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1224], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1344] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 3) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1336], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1456] = T.if_then_else(1 <= ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 and ry_outer_outer + (threadIdx_x_1 // 7 + 5) % 7 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[rc_outer_outer * 1568 + ry_outer_outer * 7 + threadIdx_x_1 + rx_outer_outer + 1448], T.float32(0))
- threadIdx_x_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.Buffer((1024,), data=kernel_shared, scope="shared")
+ kernel_shared_1 = T.Buffer((3072,), data=kernel_shared, scope="shared")
kernel_1 = T.Buffer((2359296,), data=kernel.data)
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 112] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 112) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 224] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 32256]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 336) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 448] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 560) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 672] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 96768]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 784) // 32 * 4608 + cse_var_2 + (threadIdx_x_2 + 16) % 32 * 9 + cse_var_1 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 147456 + threadIdx_x_2 // 32 * 4608 + cse_var_2 + threadIdx_x_2 % 32 * 9 + cse_var_1 + rx_outer_outer + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 16):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 147456 + cse_var_2 + threadIdx_x_2 * 9 + cse_var_1 + rx_outer_outer + 142992]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 2] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 3] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 4] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 5] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 6] * kernel_shared_1[threadIdx_x // 7 * 64]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 2] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 3] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 4] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 5] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 6] * kernel_shared_1[threadIdx_x // 7 * 64 + 32]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 50] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 51] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 52] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 53] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 54] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 55] * kernel_shared_1[threadIdx_x // 7 * 64 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 50] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 51] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 52] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 53] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 54] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 55] * kernel_shared_1[threadIdx_x // 7 * 64 + 33]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 98] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 99] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 100] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 101] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 102] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 103] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 104] * kernel_shared_1[threadIdx_x // 7 * 64 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 98] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 99] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 100] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 101] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 102] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 103] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 104] * kernel_shared_1[threadIdx_x // 7 * 64 + 34]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 147] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 148] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 149] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 150] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 151] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 152] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 153] * kernel_shared_1[threadIdx_x // 7 * 64 + 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 147] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 148] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 149] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 150] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 151] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 152] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 153] * kernel_shared_1[threadIdx_x // 7 * 64 + 35]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 196] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 197] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 198] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 199] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 200] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 201] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 202] * kernel_shared_1[threadIdx_x // 7 * 64 + 4]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 196] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 197] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 198] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 199] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 200] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 201] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 202] * kernel_shared_1[threadIdx_x // 7 * 64 + 36]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 245] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 246] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 247] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 248] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 249] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 250] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 251] * kernel_shared_1[threadIdx_x // 7 * 64 + 5]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 245] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 246] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 247] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 248] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 249] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 250] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 251] * kernel_shared_1[threadIdx_x // 7 * 64 + 37]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 294] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 295] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 296] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 297] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 298] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 299] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 300] * kernel_shared_1[threadIdx_x // 7 * 64 + 6]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 294] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 295] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 296] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 297] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 298] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 299] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 300] * kernel_shared_1[threadIdx_x // 7 * 64 + 38]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 343] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 344] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 345] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 346] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 347] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 348] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 349] * kernel_shared_1[threadIdx_x // 7 * 64 + 7]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 343] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 344] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 345] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 346] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 347] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 348] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 349] * kernel_shared_1[threadIdx_x // 7 * 64 + 39]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 392] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 393] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 394] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 395] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 396] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 397] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 398] * kernel_shared_1[threadIdx_x // 7 * 64 + 8]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 392] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 393] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 394] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 395] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 396] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 397] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 398] * kernel_shared_1[threadIdx_x // 7 * 64 + 40]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 441] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 442] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 443] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 444] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 445] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 446] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 447] * kernel_shared_1[threadIdx_x // 7 * 64 + 9]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 441] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 442] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 443] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 444] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 445] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 446] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 447] * kernel_shared_1[threadIdx_x // 7 * 64 + 41]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 490] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 491] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 492] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 493] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 494] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 495] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 496] * kernel_shared_1[threadIdx_x // 7 * 64 + 10]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 490] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 491] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 492] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 493] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 494] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 495] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 496] * kernel_shared_1[threadIdx_x // 7 * 64 + 42]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 539] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 540] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 541] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 542] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 543] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 544] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 545] * kernel_shared_1[threadIdx_x // 7 * 64 + 11]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 539] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 540] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 541] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 542] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 543] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 544] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 545] * kernel_shared_1[threadIdx_x // 7 * 64 + 43]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 588] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 589] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 590] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 591] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 592] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 593] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 594] * kernel_shared_1[threadIdx_x // 7 * 64 + 12]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 588] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 589] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 590] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 591] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 592] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 593] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 594] * kernel_shared_1[threadIdx_x // 7 * 64 + 44]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 637] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 638] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 639] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 640] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 641] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 642] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 643] * kernel_shared_1[threadIdx_x // 7 * 64 + 13]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 637] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 638] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 639] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 640] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 641] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 642] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 643] * kernel_shared_1[threadIdx_x // 7 * 64 + 45]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 686] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 687] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 688] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 689] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 690] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 691] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 692] * kernel_shared_1[threadIdx_x // 7 * 64 + 14]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 686] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 687] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 688] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 689] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 690] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 691] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 692] * kernel_shared_1[threadIdx_x // 7 * 64 + 46]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 735] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 736] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 737] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 738] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 739] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 740] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 741] * kernel_shared_1[threadIdx_x // 7 * 64 + 15]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 735] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 736] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 737] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 738] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 739] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 740] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 741] * kernel_shared_1[threadIdx_x // 7 * 64 + 47]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 784] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 785] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 786] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 787] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 788] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 789] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 790] * kernel_shared_1[threadIdx_x // 7 * 64 + 16]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 784] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 785] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 786] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 787] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 788] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 789] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 790] * kernel_shared_1[threadIdx_x // 7 * 64 + 48]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 833] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 834] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 835] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 836] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 837] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 838] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 839] * kernel_shared_1[threadIdx_x // 7 * 64 + 17]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 833] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 834] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 835] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 836] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 837] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 838] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 839] * kernel_shared_1[threadIdx_x // 7 * 64 + 49]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 882] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 883] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 884] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 885] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 886] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 887] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 888] * kernel_shared_1[threadIdx_x // 7 * 64 + 18]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 882] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 883] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 884] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 885] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 886] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 887] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 888] * kernel_shared_1[threadIdx_x // 7 * 64 + 50]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 931] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 932] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 933] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 934] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 935] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 936] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 937] * kernel_shared_1[threadIdx_x // 7 * 64 + 19]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 931] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 932] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 933] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 934] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 935] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 936] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 937] * kernel_shared_1[threadIdx_x // 7 * 64 + 51]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 980] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 981] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 982] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 983] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 984] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 985] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 986] * kernel_shared_1[threadIdx_x // 7 * 64 + 20]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 980] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 981] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 982] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 983] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 984] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 985] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 986] * kernel_shared_1[threadIdx_x // 7 * 64 + 52]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1029] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1030] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1031] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1032] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1033] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1034] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1035] * kernel_shared_1[threadIdx_x // 7 * 64 + 21]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1029] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1030] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1031] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1032] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1033] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1034] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1035] * kernel_shared_1[threadIdx_x // 7 * 64 + 53]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1078] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1079] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1080] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1081] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1082] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1083] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1084] * kernel_shared_1[threadIdx_x // 7 * 64 + 22]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1078] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1079] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1080] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1081] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1082] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1083] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1084] * kernel_shared_1[threadIdx_x // 7 * 64 + 54]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1127] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1128] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1129] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1130] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1131] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1132] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1133] * kernel_shared_1[threadIdx_x // 7 * 64 + 23]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1127] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1128] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1129] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1130] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1131] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1132] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1133] * kernel_shared_1[threadIdx_x // 7 * 64 + 55]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1176] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1177] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1178] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1179] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1180] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1181] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1182] * kernel_shared_1[threadIdx_x // 7 * 64 + 24]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1176] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1177] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1178] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1179] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1180] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1181] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1182] * kernel_shared_1[threadIdx_x // 7 * 64 + 56]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1225] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1226] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1227] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1228] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1229] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1230] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1231] * kernel_shared_1[threadIdx_x // 7 * 64 + 25]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1225] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1226] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1227] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1228] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1229] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1230] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1231] * kernel_shared_1[threadIdx_x // 7 * 64 + 57]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1274] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1275] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1276] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1277] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1278] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1279] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1280] * kernel_shared_1[threadIdx_x // 7 * 64 + 26]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1274] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1275] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1276] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1277] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1278] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1279] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1280] * kernel_shared_1[threadIdx_x // 7 * 64 + 58]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1323] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1324] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1325] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1326] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1327] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1328] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1329] * kernel_shared_1[threadIdx_x // 7 * 64 + 27]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1323] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1324] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1325] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1326] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1327] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1328] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1329] * kernel_shared_1[threadIdx_x // 7 * 64 + 59]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1372] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1373] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1374] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1375] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1376] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1377] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1378] * kernel_shared_1[threadIdx_x // 7 * 64 + 28]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1372] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1373] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1374] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1375] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1376] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1377] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1378] * kernel_shared_1[threadIdx_x // 7 * 64 + 60]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1421] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1422] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1423] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1424] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1425] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1426] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1427] * kernel_shared_1[threadIdx_x // 7 * 64 + 29]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1421] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1422] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1423] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1424] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1425] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1426] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1427] * kernel_shared_1[threadIdx_x // 7 * 64 + 61]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1470] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1471] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1472] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1473] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1474] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1475] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1476] * kernel_shared_1[threadIdx_x // 7 * 64 + 30]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1470] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1471] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1472] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1473] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1474] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1475] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1476] * kernel_shared_1[threadIdx_x // 7 * 64 + 62]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1519] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1520] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1521] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1522] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1523] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1524] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1525] * kernel_shared_1[threadIdx_x // 7 * 64 + 31]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1519] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1520] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1521] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1522] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1523] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1524] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[threadIdx_x % 7 * 7 + 1525] * kernel_shared_1[threadIdx_x // 7 * 64 + 63]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
for i1_inner, i3_inner in T.grid(2, 7):
compute_1 = T.Buffer((25088,), data=compute.data)
bias_1 = T.Buffer((512,), data=bias.data)
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 98 + i1_inner * 49 + threadIdx_x % 7 * 7 + i3_inner] = T.max(conv2d_nchw_1[i1_inner * 7 + i3_inner] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 * 2 + 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>
@@ -1073,7 +1016,7 @@ class Module:
<span class="p">)</span>
</pre></div>
</div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.315 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.350 ms
</pre></div>
</div>
</div>
@@ -1102,33 +1045,33 @@ 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=2)
-conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
-conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=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=7)
-conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
+conv2d_nchw_xx_o_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=4)
-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)
-conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2d_nc [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
+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)
@@ -1151,14 +1094,14 @@ 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:
@@ -1176,10 +1119,10 @@ 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) {
+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[1568];
- __shared__ float kernel_shared[1024];
+ __shared__ float pad_temp_shared[72];
+ __shared__ float kernel_shared[3072];
conv2d_nchw[0] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
@@ -1194,491 +1137,411 @@ extern "C" __global__ void __launch_bounds__(112) default_function_ker
conv2d_nchw[11] = 0.000000e+00f;
conv2d_nchw[12] = 0.000000e+00f;
conv2d_nchw[13] = 0.000000e+00f;
- for (int rc_outer_outer = 0; rc_outer_outer < 16; ++rc_outer_outer) {
+ for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
- for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
- __syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 49) / 7) + ry_outer_outer)) && ((((((int)threadIdx.x) % 49) / 7) + ry_outer_outer) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 104)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 216)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 328)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 440)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 560)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 552)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 672)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 664)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 784)] = (((((1 <= (((((int)threadIdx.x) % 49) / 7) + ry_outer_outer)) && ((((((int)threadIdx.x) % 49) / 7) + ry_outer_outer) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 776)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 896)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 2) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 888)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1008)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 4) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1000)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1120)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 6) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1112)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1232)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 1) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1224)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1344)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 3) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1336)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1456)] = (((((1 <= (ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7))) && ((ry_outer_outer + (((((int)threadIdx.x) / 7) + 5) % 7)) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 1568) + (ry_outer_outer * 7)) + ((int)threadIdx.x)) + rx_outer_outer) + 1448)] : 0.000000e+00f);
- kernel_shared[((int)threadIdx.x)] = kernel[((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 112)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 112) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 32256)];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 336) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 560)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 560) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 672)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 96768)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 784) >> 5) * 4608)) + (rc_outer_outer * 288)) + (((((int)threadIdx.x) + 16) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((((int)blockIdx.x) * 147456) + ((((int)threadIdx.x) >> 5) * 4608)) + (rc_outer_outer * 288)) + ((((int)threadIdx.x) & 31) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 129024)];
- if (((int)threadIdx.x) < 16) {
- kernel_shared[(((int)threadIdx.x) + 1008)] = kernel[((((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 288)) + (((int)threadIdx.x) * 9)) + (ry_outer_outer * 3)) + rx_outer_outer) + 142992)];
- }
- __syncthreads();
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 7)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 2)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 3)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 4)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 5)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 6)] * kernel_shared[((((int)threadIdx.x) / 7) * 64)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((((int)threadIdx.x) % 7) * 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 2)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 3)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 4)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 5)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 6)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 32)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 49)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 50)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 51)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 52)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 53)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 55)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 49)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 50)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 51)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 52)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 53)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 54)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 55)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 33)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 98)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 99)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 100)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 101)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 102)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 103)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 104)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 98)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 99)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 100)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 101)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 102)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 103)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 104)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 34)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 147)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 148)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 149)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 150)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 151)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 152)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 153)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 3)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 147)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 148)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 149)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 150)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 151)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 152)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 153)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 35)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 198)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 199)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 200)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 201)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 202)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 4)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 196)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 197)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 198)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 199)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 200)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 201)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 202)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 36)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 245)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 246)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 247)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 248)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 249)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 250)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 251)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 5)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 245)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 246)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 247)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 248)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 249)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 250)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 251)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 37)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 294)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 295)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 296)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 297)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 298)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 299)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 300)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 6)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 294)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 295)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 296)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 297)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 298)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 299)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 300)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 38)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 343)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 344)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 345)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 346)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 347)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 348)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 349)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 7)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 343)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 344)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 345)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 346)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 347)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 348)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 349)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 39)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 392)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 393)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 394)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 395)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 396)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 397)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 398)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 8)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 392)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 393)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 394)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 395)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 396)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 397)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 398)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 40)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 9)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 441)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 442)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 443)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 444)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 445)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 446)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 447)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 41)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 490)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 491)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 492)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 493)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 494)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 495)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 496)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 10)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 490)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 491)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 492)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 493)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 494)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 495)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 496)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 42)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 539)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 540)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 541)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 542)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 543)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 544)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 545)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 11)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 539)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 540)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 541)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 542)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 543)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 544)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 545)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 43)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 588)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 589)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 590)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 591)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 592)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 593)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 594)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 12)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 588)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 589)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 590)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 591)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 592)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 593)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 594)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 44)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 637)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 638)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 639)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 640)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 641)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 642)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 643)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 13)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 637)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 638)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 639)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 640)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 641)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 642)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 643)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 45)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 686)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 687)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 688)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 689)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 690)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 691)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 692)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 14)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 686)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 687)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 688)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 689)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 690)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 691)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 692)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 46)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 735)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 736)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 737)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 738)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 739)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 740)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 741)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 15)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 735)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 736)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 737)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 738)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 739)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 740)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 741)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 47)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 784)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 785)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 786)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 787)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 788)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 789)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 790)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 16)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 784)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 785)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 786)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 787)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 788)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 789)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 790)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 48)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 833)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 834)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 835)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 836)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 837)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 838)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 839)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 17)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 833)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 834)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 835)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 836)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 837)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 838)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 839)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 49)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 882)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 883)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 884)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 885)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 886)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 887)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 888)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 18)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 882)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 883)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 884)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 885)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 886)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 887)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 888)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 50)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 931)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 932)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 933)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 934)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 935)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 936)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 937)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 19)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 931)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 932)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 933)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 934)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 935)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 936)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 937)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 51)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 980)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 981)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 982)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 983)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 984)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 985)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 986)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 20)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 980)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 981)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 982)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 983)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 984)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 985)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 986)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 52)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1029)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1030)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1031)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1032)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1033)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1034)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1035)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 21)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1029)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1030)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1031)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1032)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1033)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1034)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1035)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 53)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1078)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1079)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1080)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1081)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1082)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1083)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1084)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 22)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1078)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1079)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1080)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1081)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1082)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1083)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1084)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 54)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 23)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1127)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1128)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1129)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1130)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1131)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1132)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1133)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 55)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1176)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1177)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1178)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1179)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1180)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1181)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1182)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 24)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1176)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1177)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1178)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1179)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1180)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1181)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1182)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 56)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1225)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1226)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1227)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1228)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1229)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1230)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1231)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 25)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1225)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1226)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1227)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1228)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1229)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1230)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1231)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 57)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1274)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1275)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1276)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1277)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1278)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1279)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1280)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 26)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1274)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1275)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1276)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1277)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1278)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1279)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1280)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 58)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1324)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1325)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1326)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1327)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1328)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1329)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 27)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1323)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1324)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1325)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1326)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1327)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1328)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1329)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 59)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1372)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1373)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1374)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1375)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1376)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1377)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 28)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1372)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1373)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1374)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1375)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1376)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1377)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1378)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 60)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1421)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1422)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1423)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1424)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1425)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1426)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1427)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 29)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1421)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1422)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1423)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1424)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1425)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1426)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1427)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 61)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1470)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1471)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1472)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1473)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1474)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1475)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1476)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 30)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1470)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1471)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1472)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1473)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1474)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1475)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1476)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 62)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1519)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1520)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1521)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1522)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1523)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1524)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1525)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 31)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1519)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1520)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1521)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1522)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1523)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1524)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((((int)threadIdx.x) % 7) * 7) + 1525)] * kernel_shared[(((((int)threadIdx.x) / 7) * 64) + 63)]));
+ __syncthreads();
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) * 4) % 9))) && (((((int)threadIdx.x) * 4) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
}
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 1) % 9))) && ((((((int)threadIdx.x) * 4) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
+ }
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 2) % 9))) && ((((((int)threadIdx.x) * 4) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
+ }
+ if (((int)threadIdx.x) < 18) {
+ pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 3) % 9))) && ((((((int)threadIdx.x) * 4) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
+ }
+ kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 64) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 128) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
+ kernel_shared[(((((((int)threadIdx.x) + 256) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 320) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
+ kernel_shared[(((((((int)threadIdx.x) + 448) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 512) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
+ kernel_shared[(((((((int)threadIdx.x) + 640) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 704) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
+ kernel_shared[(((((((int)threadIdx.x) + 832) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 896) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
+ kernel_shared[(((((((int)threadIdx.x) + 1024) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1088) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
+ kernel_shared[(((((((int)threadIdx.x) + 1216) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1280) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
+ kernel_shared[(((((((int)threadIdx.x) + 1408) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1472) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
+ kernel_shared[(((((((int)threadIdx.x) + 1600) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1664) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
+ kernel_shared[(((((((int)threadIdx.x) + 1792) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 1856) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
+ kernel_shared[(((((((int)threadIdx.x) + 1984) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2048) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
+ kernel_shared[(((((((int)threadIdx.x) + 2176) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2240) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
+ kernel_shared[(((((((int)threadIdx.x) + 2368) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2432) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
+ kernel_shared[(((((((int)threadIdx.x) + 2560) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2624) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
+ kernel_shared[(((((((int)threadIdx.x) + 2752) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 2816) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
+ kernel_shared[(((((((int)threadIdx.x) + 2944) / 24) * 24) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 1) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
+ kernel_shared[(((((((int)threadIdx.x) + 3008) / 24) * 24) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 3)) + ((((int)threadIdx.x) + 2) % 3))] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
+ __syncthreads();
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[0] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[1] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[2] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[3] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[4] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[5] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[6] * kernel_shared[(((int)threadIdx.x) * 48)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 3)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[0] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[9] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 24)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 27)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 1)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 4)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[1] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[10] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 25)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 28)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 2)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 5)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[2] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[11] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[3] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[12] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[4] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[13] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[5] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[14] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[6] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[15] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[7] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[16] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[8] * kernel_shared[((((int)threadIdx.x) * 48) + 26)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[17] * kernel_shared[((((int)threadIdx.x) * 48) + 29)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 6)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 9)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[18] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[27] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 30)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 33)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 7)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 10)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[19] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[28] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 31)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 34)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 8)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 11)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[20] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[29] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[21] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[30] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[22] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[31] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[23] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[32] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[24] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[33] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[25] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[34] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[26] * kernel_shared[((((int)threadIdx.x) * 48) + 32)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[35] * kernel_shared[((((int)threadIdx.x) * 48) + 35)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 12)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 15)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[36] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[45] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 36)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 39)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 13)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 16)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[37] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[46] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 37)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 40)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 14)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 17)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[38] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[47] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[39] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[48] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[40] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[49] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[41] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[50] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[42] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[51] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[43] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[52] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[44] * kernel_shared[((((int)threadIdx.x) * 48) + 38)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[53] * kernel_shared[((((int)threadIdx.x) * 48) + 41)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 18)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 21)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[54] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[63] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 42)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 45)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 19)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 22)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[55] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[64] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 43)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 46)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 20)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 23)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[56] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[65] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[57] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[66] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[58] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[67] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[59] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[68] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[60] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[69] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[61] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[70] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[62] * kernel_shared[((((int)threadIdx.x) * 48) + 44)]));
+ conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[71] * kernel_shared[((((int)threadIdx.x) * 48) + 47)]));
}
}
for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[(((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 98)) + (i1_inner * 49)) + ((((int)threadIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[(((((int)blockIdx.x) * 32) + ((((int)threadIdx.x) / 7) * 2)) + 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);
}
}
}
@@ -1714,7 +1577,7 @@ In the example below we resume the status and do more 5 trials.</p>
Get devices for measurement successfully!
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 6 minutes 21.426 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 6 minutes 5.984 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 36da06c6e4..a0ff90f817 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -921,7 +921,7 @@ so we can read the log file and load the best schedules.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 8.1186 8.1181 8.1212 8.1165 0.0019
+ 8.1404 8.1394 8.1507 8.1311 0.0080
</pre></div>
</div>
</div>
@@ -943,7 +943,7 @@ to learn how to use the RPC Tracker and RPC Server.
To use the RPC Tracker in auto-scheduler, replace the runner in <code class="code docutils literal notranslate"><span class="pre">TuningOptions</span></code>
with <a class="reference internal" href="../../reference/api/python/auto_scheduler.html#tvm.auto_scheduler.RPCRunner" title="tvm.auto_scheduler.RPCRunner"><code class="xref any py py-class docutils literal notranslate"><span class="pre">auto_scheduler.RPCRunner</span></code></a>.</p></li>
</ol>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 11.382 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 10.847 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 6ece35a060..2b6f9d1150 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -940,7 +940,7 @@ so we can read the log file and load the best schedules.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 759.7677 760.1449 763.1201 756.0381 2.9035
+ 754.6195 754.4586 756.4399 752.9600 1.4252
</pre></div>
</div>
</div>
@@ -962,7 +962,7 @@ to learn how to use the RPC Tracker and RPC Server.
To use the RPC Tracker in auto-scheduler, replace the runner in <code class="code docutils literal notranslate"><span class="pre">TuningOptions</span></code>
with <a class="reference internal" href="../../reference/api/python/auto_scheduler.html#tvm.auto_scheduler.RPCRunner" title="tvm.auto_scheduler.RPCRunner"><code class="xref any py py-class docutils literal notranslate"><span class="pre">auto_scheduler.RPCRunner</span></code></a>.</p></li>
</ol>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 43.370 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 42.224 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 05d7269b6a..4ea9a2beee 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -637,27 +637,86 @@ 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": T.bool(True), "global_symbol": "main", "tir.noalias": T.bool(True)})
- for i0_outer in T.parallel(64):
- compute_1 = T.allocate([64], "float32", "global")
- for i1_outer in range(16):
- compute_2 = T.Buffer((64,), data=compute_1)
- for nb_j_inner in range(2):
- for i_inner_init, j_init in T.grid(2, 16):
- compute_2[i_inner_init * 32 + nb_j_inner * 16 + j_init] = T.float32(0)
- for elem_idx, i_inner, j in T.grid(T.Let(placeholder_5[cse_var_1 + 1] - placeholder_5[cse_var_1], where={cse_var_1: i1_outer * 2 + nb_j_inner}), 2, 16):
- cse_var_1 = T.int32()
- placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
- cse_var_3: T.int32 = i1_outer * 2 + nb_j_inner
- cse_var_2: T.int32 = i_inner * 32 + nb_j_inner * 16 + j
- placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
- placeholder_7 = T.Buffer((32768,), data=placeholder.data)
- placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_2] = compute_2[cse_var_2] + placeholder_6[placeholder_5[cse_var_3] * 16 + elem_idx * 16 + j] * T.max(placeholder_7[i0_outer * 512 + i_inner * 256 + placeholder_8[placeholder_5[cse_var_3] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(2, 32):
- cse_var_4: T.int32 = i0_outer * 1024 + i0_inner * 512 + i1_outer * 32 + i1_inner
- compute_3 = T.Buffer((65536,), data=compute.data)
- placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
- compute_3[cse_var_4] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_4], T.float32(0))
+ for i0_outer_i1_outer_fused in T.parallel(32):
+ compute_1 = T.allocate([2048], "float32", "global")
+ compute_2 = T.Buffer((2048,), data=compute_1)
+ for i_outer_inner in range(2):
+ for i_inner_init in range(64):
+ cse_var_1: T.int32 = i_outer_inner * 1024 + i_inner_init * 16
+ compute_2[cse_var_1] = T.float32(0)
+ compute_2[cse_var_1 + 1] = T.float32(0)
+ compute_2[cse_var_1 + 2] = T.float32(0)
+ compute_2[cse_var_1 + 3] = T.float32(0)
+ compute_2[cse_var_1 + 4] = T.float32(0)
+ compute_2[cse_var_1 + 5] = T.float32(0)
+ compute_2[cse_var_1 + 6] = T.float32(0)
+ compute_2[cse_var_1 + 7] = T.float32(0)
+ compute_2[cse_var_1 + 8] = T.float32(0)
+ compute_2[cse_var_1 + 9] = T.float32(0)
+ compute_2[cse_var_1 + 10] = T.float32(0)
+ compute_2[cse_var_1 + 11] = T.float32(0)
+ compute_2[cse_var_1 + 12] = T.float32(0)
+ compute_2[cse_var_1 + 13] = T.float32(0)
+ compute_2[cse_var_1 + 14] = T.float32(0)
+ compute_2[cse_var_1 + 15] = T.float32(0)
+ for elem_idx, i_inner in T.grid(placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused], 64):
+ placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
+ placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
+ placeholder_7 = T.Buffer((32768,), data=placeholder.data)
+ placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_2: T.int32 = i_outer_inner * 1024 + i_inner * 16
+ compute_2[cse_var_2] = compute_2[cse_var_2] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_3: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 1
+ compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 1] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_4: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 2
+ compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 2] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_5: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 3
+ compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 3] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_6: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 4
+ compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 4] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_7: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 5
+ compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 5] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_8: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 6
+ compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 6] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_9: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 7
+ compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 7] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_10: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 8
+ compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 8] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_11: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 9
+ compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 9] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_12: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 10
+ compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 10] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_13: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 11
+ compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 11] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_14: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 12
+ compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 12] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_15: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 13
+ compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 13] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_16: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 14
+ compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 14] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ if T.likely(elem_idx < placeholder_5[i0_outer_i1_outer_fused + 1] - placeholder_5[i0_outer_i1_outer_fused]):
+ cse_var_17: T.int32 = i_outer_inner * 1024 + i_inner * 16 + 15
+ compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[i0_outer_i1_outer_fused] * 16 + elem_idx * 16 + 15] * T.max(placeholder_7[i_outer_inner * 16384 + i_inner * 256 + placeholder_8[placeholder_5[i0_outer_i1_outer_fused] + elem_idx]], T.float32(0))
+ for i0_inner in range(128):
+ cse_var_18: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 16
+ compute_3 = T.Buffer((65536,), data=compute.data)
+ placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
+ compute_3[cse_var_18:cse_var_18 + 16] = T.max(compute_2[i0_inner * 16:i0_inner * 16 + 16] + placeholder_5[cse_var_18:cse_var_18 + 16], T.Broadcast(T.float32(0), 16))
</pre></div>
</div>
</div>
@@ -691,7 +750,7 @@ class Module:
<span class="p">)</span>
</pre></div>
</div>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.873 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.824 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 abd9a5db97..49ce152df8 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-tune-with-autotvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:36.350</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:57.262</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -354,7 +354,7 @@
</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>00:36.315</p></td>
+<td><p>00:57.227</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>
@@ -366,7 +366,7 @@
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></td>
-<td><p>00:00.005</p></td>
+<td><p>00:00.004</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></td>
diff --git a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
index 6a03bffa64..7ac7f9c93e 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -573,8 +573,25 @@ for this template</p>
waiting for device...
device available
Get devices for measurement successfully!
-No: 1 GFLOPS: 35.00/35.00 result: MeasureResult(costs=(0.0066147660625,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.2832298278808594, timestamp=1680521223.4808674) [('tile_f', [-1, 1, 8, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5281179
-No: 2 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+No: 1 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 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, 2, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7711234
+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)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -696,8 +713,13 @@ 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, 4, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9415737
-No: 3 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 1, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3956030
+No: 3 GFLOPS: 183.79/183.79 result: MeasureResult(costs=(0.0012595632702702701,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.519602537155151, timestamp=1680581738.1102219) [('tile_f', [-1, 1, 8, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4502421
+No: 4 GFLOPS: 154.18/183.79 result: MeasureResult(costs=(0.001501489656716418,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.0189778804779053, timestamp=1680581738.9676352) [('tile_f', [-1, 2, 2, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3488649
+No: 5 GFLOPS: 110.76/183.79 result: MeasureResult(costs=(0.00209018620754717,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.892058372497559, timestamp=1680581748.0066292) [('tile_f', [-1, 1, 1, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3786116
+No: 6 GFLOPS: 31.66/183.79 result: MeasureResult(costs=(0.007312564285714286,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5477538108825684, timestamp=1680581748.8912892) [('tile_f', [-1, 2, 1, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3116181
+No: 7 GFLOPS: 429.48/429.48 result: MeasureResult(costs=(0.0005390288449612404,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.55623459815979, timestamp=1680581751.377373) [('tile_f', [-1, 1, 64, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,455885
+No: 8 GFLOPS: 0.00/429.48 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
@@ -819,8 +841,8 @@ 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, 32, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,279053
-No: 4 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7140687
+No: 9 GFLOPS: 0.00/429.48 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
@@ -942,10 +964,8 @@ 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, 8, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6627624
-No: 5 GFLOPS: 30.79/35.00 result: MeasureResult(costs=(0.007519179352941176,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.6613967418670654, timestamp=1680521231.781294) [('tile_f', [-1, 64, 2, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3527056
-No: 6 GFLOPS: 4.08/35.00 result: MeasureResult(costs=(0.05678187175,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.161860227584839, timestamp=1680521233.0591633) [('tile_f', [-1, 8, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3680007
-No: 7 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6841672
+No: 10 GFLOPS: 0.00/429.48 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
@@ -1067,8 +1087,11 @@ 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, 1, 4, 64]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5646067
-No: 8 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 128]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9655792
+No: 11 GFLOPS: 655.02/655.02 result: MeasureResult(costs=(0.00035342363841807915,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7047991752624512, timestamp=1680581753.9538736) [('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3391971
+No: 12 GFLOPS: 11.59/655.02 result: MeasureResult(costs=(0.0199772745,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.3188257217407227, timestamp=1680581754.9549854) [('tile_f', [-1, 4, 2, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7409666
+No: 13 GFLOPS: 169.40/655.02 result: MeasureResult(costs=(0.001366572396226415,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.081042766571045, timestamp=1680581760.2097487) [('tile_f', [-1, 8, 4, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9891442
+No: 14 GFLOPS: 0.00/655.02 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
@@ -1190,8 +1213,8 @@ 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, 32, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8272821
-No: 9 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7674285
+No: 15 GFLOPS: 0.00/655.02 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
@@ -1313,8 +1336,8 @@ 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, 64, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9457914
-No: 10 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6699522
+No: 16 GFLOPS: 0.00/655.02 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
@@ -1436,8 +1459,9 @@ 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, 16, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5563887
-No: 11 GFLOPS: 0.00/35.00 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, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1484399
+No: 17 GFLOPS: 385.93/655.02 result: MeasureResult(costs=(0.0005998543142857143,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5332989692687988, timestamp=1680581763.394771) [('tile_f', [-1, 1, 32, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3719630
+No: 18 GFLOPS: 0.00/655.02 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
@@ -1559,8 +1583,9 @@ 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, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8052191
-No: 12 GFLOPS: 0.00/35.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 2, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3752654
+No: 19 GFLOPS: 128.18/655.02 result: MeasureResult(costs=(0.0018060951754385963,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.7704761028289795, timestamp=1680581764.264103) [('tile_f', [-1, 1, 2, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,8978924
+No: 20 GFLOPS: 0.00/655.02 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
@@ -1682,747 +1707,7 @@ 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, 1, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10319545
-No: 13 GFLOPS: 0.00/35.00 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
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 8]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2039327
-No: 14 GFLOPS: 139.11/139.11 result: MeasureResult(costs=(0.0016641625238095238,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.1421754360198975, timestamp=1680521236.668181) [('tile_f', [-1, 1, 16, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3416905
-No: 15 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 16, 8]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4868100
-No: 16 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,5169747
-No: 17 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 1, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3206886
-No: 18 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 64, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8220205
-No: 19 GFLOPS: 17.50/139.11 result: MeasureResult(costs=(0.013231103875,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2924680709838867, timestamp=1680521238.2001758) [('tile_f', [-1, 16, 4, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6001719
-No: 20 GFLOPS: 0.00/139.11 result: Traceback (most recent call last):
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
- func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
- func = build(s, args, target=target, runtime=runtime)
- File "/workspace/python/tvm/driver/build_module.py", line 227, in build
- input_mod = lower(inputs, args, name=name, binds=binds)
- File "/workspace/python/tvm/driver/build_module.py", line 134, in lower
- return ffi.lower_schedule(inp, args, name, binds, simple_mode)
- File "tvm/_ffi/_cython/./packed_func.pxi", line 331, in tvm._ffi._cy3.core.PackedFuncBase.__call__
- File "tvm/_ffi/_cython/./packed_func.pxi", line 276, in tvm._ffi._cy3.core.FuncCall
- File "tvm/_ffi/_cython/./base.pxi", line 181, in tvm._ffi._cy3.core.CHECK_CALL
-tvm._ffi.base.TVMError: Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel
-
-Traceback (most recent call last):
- 24: TVMFuncCall
- at ../src/runtime/c_runtime_api.cc:477
- 23: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 22: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 21: operator()
- at ../include/tvm/runtime/packed_func.h:1734
- 20: unpack_call<tvm::IRModule, 5, tvm::<lambda(tvm::te::Schedule, const tvm::runtime::Array<tvm::runtime::ObjectRef>&, const tvm::runtime::String&, const tvm::runtime::Map<tvm::te::Tensor, tvm::tir::Buffer>&, bool)> >
- at ../include/tvm/runtime/packed_func.h:1674
- 19: run<>
- at ../include/tvm/runtime/packed_func.h:1634
- 18: run<tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 17: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 16: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 15: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1634
- 14: run<tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_, tvm::runtime::TVMMovableArgValueWithContext_>
- at ../include/tvm/runtime/packed_func.h:1649
- 13: operator()
- at ../src/driver/driver_api.cc:402
- 12: tvm::LowerSchedule(tvm::te::Schedule, tvm::runtime::Array<tvm::runtime::ObjectRef, void> const&, std::__cxx11::basic_string<char, std::char_traits<char>, std::allocator<char> > const&, std::unordered_map<tvm::te::Tensor, tvm::tir::Buffer, std::hash<tvm::te::Tensor>, std::equal_to<tvm::te::Tensor>, std::allocator<std::pair<tvm::te::Tensor const, tvm::tir::Buffer> > > const&, tvm::GlobalVarSupply, bool)
- at ../src/driver/driver_api.cc:388
- 11: tvm::LowerWithPassList(tvm::IRModule, tvm::runtime::Array<tvm::transform::Pass, void>)
- at ../src/driver/driver_api.cc:283
- 10: tvm::transform::Pass::operator()(tvm::IRModule) const
- at ../src/ir/transform.cc:258
- 9: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 8: tvm::transform::SequentialNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:451
- 7: tvm::transform::Pass::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/ir/transform.cc:274
- 6: tvm::tir::transform::PrimFuncPassNode::operator()(tvm::IRModule, tvm::transform::PassContext const&) const
- at ../src/tir/ir/transform.cc:100
- 5: tvm::runtime::TypedPackedFunc<tvm::tir::PrimFunc (tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext)>::operator()(tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext) const
- at ../include/tvm/runtime/packed_func.h:1753
- 4: tvm::tir::PrimFunc tvm::runtime::detail::typed_packed_call_dispatcher<tvm::tir::PrimFunc>::run<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::runtime::PackedFunc const&, tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&)
- at ../include/tvm/runtime/packed_func.h:1697
- 3: tvm::runtime::TVMRetValue tvm::runtime::PackedFunc::operator()<tvm::tir::PrimFunc, tvm::IRModule, tvm::transform::PassContext>(tvm::tir::PrimFunc&&, tvm::IRModule&&, tvm::transform::PassContext&&) const
- at ../include/tvm/runtime/packed_func.h:1621
- 2: tvm::runtime::PackedFuncObj::CallPacked(tvm::runtime::TVMArgs, tvm::runtime::TVMRetValue*) const
- at ../include/tvm/runtime/packed_func.h:1217
- 1: Call
- at ../include/tvm/runtime/packed_func.h:1213
- 0: operator()
- at ../src/runtime/c_runtime_api.cc:534
- File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
- File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
- raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,538236
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4815751
</pre></div>
</div>
<p>Finally we can inspect the best config from log file, check correctness,
@@ -2461,9 +1746,9 @@ 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, 16, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 16]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3416905
+[('tile_f', [-1, 2, 2, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3391971
Finish loading 20 records
-Time cost of this operator: 0.001949
+Time cost of this operator: 0.000776
</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 10f9333f0c..2749209203 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -649,10 +649,10 @@ 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 316.0 98.745 (1, 2, 10, 10, 3) 2 1 [316.0]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.061 0.957 (1, 6, 10, 10) 1 1 [3.061]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.954 0.298 (1, 1, 10, 10, 3) 1 1 [0.954]
-Total_time - 320.015 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 314.6 98.734 (1, 2, 10, 10, 3) 2 1 [314.6]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.082 0.967 (1, 6, 10, 10) 1 1 [3.082]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.952 0.299 (1, 1, 10, 10, 3) 1 1 [0.952]
+Total_time - 318.634 - - - - -
</pre></div>
</div>
</div>
@@ -704,13 +704,13 @@ Total_time -
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 103.0 97.503 (1, 6, 10, 10, 1) 2 1 [103.0]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.781 1.686 (1, 6, 10, 10) 1 1 [1.781]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.857 0.811 (1, 3, 10, 10, 1) 1 1 [0.857]
-Total_time - 105.638 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 100.9 97.327 (1, 6, 10, 10, 1) 2 1 [100.9]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.8 1.736 (1, 6, 10, 10) 1 1 [1.8]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.972 0.937 (1, 1, 10, 10, 3) 1 1 [0.972]
+Total_time - 103.671 - - - - -
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 22.617 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 21.699 seconds)</p>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-autotune-py">
<div class="sphx-glr-download sphx-glr-download-python docutils container">
<p><a class="reference download internal" download="" href="../../_downloads/9ccca8fd489a1486ac71b55a55c320c5/micro_autotune.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">micro_autotune.py</span></code></a></p>
diff --git a/docs/how_to/work_with_microtvm/micro_pytorch.html b/docs/how_to/work_with_microtvm/micro_pytorch.html
index 316d6c4afd..cec0d8dabb 100644
--- a/docs/how_to/work_with_microtvm/micro_pytorch.html
+++ b/docs/how_to/work_with_microtvm/micro_pytorch.html
@@ -460,8 +460,8 @@ 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]
- 90%|######### | 3.09M/3.42M [00:00<00:00, 32.3MB/s]
-100%|##########| 3.42M/3.42M [00:00<00:00, 33.1MB/s]
+ 61%|###### | 2.09M/3.42M [00:00<00:00, 11.8MB/s]
+100%|##########| 3.42M/3.42M [00:00<00:00, 18.8MB/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.
@@ -587,7 +587,7 @@ 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 18.650 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 17.974 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 83928df3d7..944248f5fa 100644
--- a/docs/how_to/work_with_microtvm/micro_train.html
+++ b/docs/how_to/work_with_microtvm/micro_train.html
@@ -528,7 +528,7 @@ take about <strong>2 minutes</strong> to download the Stanford Cars, while COCO
<a href="https://docs.python.org/3/library/shutil.html#shutil.move" title="shutil.move" class="sphx-glr-backref-module-shutil sphx-glr-backref-type-py-function"><span class="n">shutil</span><span class="o">.</span><span class="n">move</span></a><span class="p">(</span><span class="sa">f</span><span class="s2">"</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/tmpf6o0yyh0/images/random'
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>'/tmp/tmpaxabpir0/images/random'
</pre></div>
</div>
</div>
@@ -588,8 +588,8 @@ objects to other stuff? We can display some examples from our datasets using <co
<span class="n">plt</span><span class="o">.</span><span class="n">axis</span><span class="p">(</span><span class="s2">"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="[1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmpf6o0yyh0/images/target contains 8144 images
-/tmp/tmpf6o0yyh0/images/random contains 5000 images
+<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmpaxabpir0/images/target contains 8144 images
+/tmp/tmpaxabpir0/images/random contains 5000 images
</pre></div>
</div>
</div>
@@ -701,13 +701,13 @@ the time on our validation set).</p>
</pre></div>
</div>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Epoch 1/3
-328/328 - 41s - loss: 0.2421 - accuracy: 0.9175 - val_loss: 0.1331 - val_accuracy: 0.9535 - 41s/epoch - 124ms/step
+328/328 - 41s - loss: 0.2252 - accuracy: 0.9244 - val_loss: 0.1283 - val_accuracy: 0.9562 - 41s/epoch - 125ms/step
Epoch 2/3
-328/328 - 35s - loss: 0.1024 - accuracy: 0.9624 - val_loss: 0.1931 - val_accuracy: 0.9358 - 35s/epoch - 105ms/step
+328/328 - 35s - loss: 0.1040 - accuracy: 0.9614 - val_loss: 0.0961 - val_accuracy: 0.9675 - 35s/epoch - 105ms/step
Epoch 3/3
-328/328 - 34s - loss: 0.0714 - accuracy: 0.9728 - val_loss: 0.3400 - val_accuracy: 0.8716 - 34s/epoch - 105ms/step
+328/328 - 34s - loss: 0.0595 - accuracy: 0.9767 - val_loss: 0.1097 - val_accuracy: 0.9645 - 34s/epoch - 105ms/step
-<keras.callbacks.History object at 0x7f896e28c6d0>
+<keras.callbacks.History object at 0x7fcdcd711790>
</pre></div>
</div>
</div>
@@ -971,7 +971,7 @@ as intended.</p>
<p>From here, we could modify the model to read live images from the camera - we have another
Arduino tutorial for how to do that <a class="reference external" href="https://github.com/guberti/tvm-arduino-demos/tree/master/examples/person_detection">on GitHub</a>. Alternatively, we could also
<a class="reference external" href="https://tvm.apache.org/docs/how_to/work_with_microtvm/micro_autotune.html">use TVM’s autotuning capabilities</a> to dramatically improve the model’s performance.</p>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes 26.612 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes 30.838 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 a62d3d03ea..f7d341b9d0 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-microtvm-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>07:33.972</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>07:36.295</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 82%" />
@@ -354,27 +354,27 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_train.html#sphx-glr-how-to-work-with-microtvm-micro-train-py"><span class="std std-ref">5. Training Vision Models for microTVM on Arduino</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_train.py</span></code>)</p></td>
-<td><p>04:26.612</p></td>
+<td><p>04:30.838</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">6. Model Tuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></td>
-<td><p>01:22.617</p></td>
+<td><p>01:21.699</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_pytorch.html#sphx-glr-how-to-work-with-microtvm-micro-pytorch-py"><span class="std std-ref">4. microTVM PyTorch Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_pytorch.py</span></code>)</p></td>
-<td><p>01:18.650</p></td>
+<td><p>01:17.974</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="micro_aot.html#sphx-glr-how-to-work-with-microtvm-micro-aot-py"><span class="std std-ref">3. microTVM Ahead-of-Time (AOT) Compilation</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_aot.py</span></code>)</p></td>
-<td><p>00:10.409</p></td>
+<td><p>00:10.265</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_custom_ide.html#sphx-glr-how-to-work-with-microtvm-micro-custom-ide-py"><span class="std std-ref">9. Bring microTVM to your own development environment</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_custom_ide.py</span></code>)</p></td>
-<td><p>00:08.267</p></td>
+<td><p>00:08.078</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">2. microTVM TFLite Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></td>
-<td><p>00:07.416</p></td>
+<td><p>00:07.442</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">7. Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU with CMSIS-NN</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></td>
diff --git a/docs/how_to/work_with_relay/sg_execution_times.html b/docs/how_to/work_with_relay/sg_execution_times.html
index 8cc82f5ef6..afce3aaabb 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-relay-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:36.850</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:36.825</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -354,15 +354,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="using_pipeline_executor.html#sphx-glr-how-to-work-with-relay-using-pipeline-executor-py"><span class="std std-ref">Using Pipeline Executor in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_pipeline_executor.py</span></code>)</p></td>
-<td><p>00:32.167</p></td>
+<td><p>00:32.208</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>00:02.888</p></td>
+<td><p>00:02.876</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>00:01.789</p></td>
+<td><p>00:01.735</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 336f8aae03..a3ad3a0dee 100644
--- a/docs/how_to/work_with_schedules/intrin_math.html
+++ b/docs/how_to/work_with_schedules/intrin_math.html
@@ -540,7 +540,7 @@ The following example customizes CUDA lowering rule for <code class="code docuti
<a href="../../reference/api/python/ir.html#tvm.ir.register_intrin_lowering" title="tvm.ir.register_intrin_lowering" class="sphx-glr-backref-module-tvm-ir sphx-glr-backref-type-py-function"><span class="n">register_intrin_lowering</span></a><span class="p">(</span><span class="s2">"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 0x7f85b9eb20e0>
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span><function my_cuda_math_rule at 0x7fca2c73c290>
</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 b0a993dd85..702c6726da 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-schedules-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:09.169</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:08.131</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -354,15 +354,15 @@
</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>00:06.384</p></td>
+<td><p>00:05.312</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>00:01.280</p></td>
+<td><p>00:01.304</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>00:00.617</p></td>
+<td><p>00:00.623</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>
@@ -370,7 +370,7 @@
<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>00:00.130</p></td>
+<td><p>00:00.132</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><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>
@@ -378,7 +378,7 @@
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><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>00:00.056</p></td>
+<td><p>00:00.055</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="tuple_inputs.html#sphx-glr-how-to-work-with-schedules-tuple-inputs-py"><span class="std std-ref">Compute and Reduce with Tuple Inputs</span></a> (<code class="docutils literal notranslate"><span class="pre">tuple_inputs.py</span></code>)</p></td>
diff --git a/docs/objects.inv b/docs/objects.inv
index a4222e1ad8..835f544cf8 100644
Binary files a/docs/objects.inv and b/docs/objects.inv differ
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
index bf1d13435b..688044b416 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode-members.html
@@ -118,60 +118,61 @@ $(function() {
<tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac8855eae5db733dcf21542b3dbd06e15">HasBlock</a>(const BlockRV &block_rv) const =0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
<tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ac9e5eed7719e322117bde996a171e33a">IncRef</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">protected</span></td></tr>
<tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a90e90b3f4ba8a590baff78c75807bbc7">IsInstance</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12">mod</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a133436a9ec5c4a768b94102bf95a660b">Object</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ab7968feb6ad38ecaffc320e13819d826">Object</a>(const Object &other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#aa1612f69ea5b4225d4cda759cd517323">Object</a>(Object &&other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a69c32fbd96181f5c21d2c878ab285e4f">operator=</a>(const Object &other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ae341e561272ff43cdcbc927bc29ac50d">operator=</a>(Object &&other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a1ac39c82aee1f8de30d5871d5923fc24">PadEinsum</a>(const BlockRV &block_rv, const Array< Integer > &padding)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789">Parallel</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2f454daf29e582a65ffe361e958122df">ReadAt</a>(const LoopRV &loop_rv, const BlockRV &block_rv, int read_buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a0d492efee331e2239a093f4b2017c10f">ref_counter_</a></td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">protected</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a55549a6c23987890246248682560a03d">RefCounterType</a> typedef</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5">ReIndex</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">ReindexCacheRead</a>(const BlockRV &block_rv, int read_buffer_index, const String &storage_scope, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">ReindexCacheWrite</a>(const BlockRV &block_rv, int write_buffer_index, const String &storage_scope, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a70d353bb52f6fa29fedeb90a6ff872d5">RemoveRV</a>(const BlockRV &block_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a7c44d4f4ea662291ccb9d79383b6fefe">RemoveRV</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a00fcf343d2bc8f36f170c04e5e29d2dc">RemoveRV</a>(const ExprRV &expr_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a059229fe0e254961da406807a97f7a3d">Reorder</a>(const Array< LoopRV > &ordered_loop_rvs)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a3c3024de7f2da68069e593bb8ad64f7f">ReorderBlockIterVar</a>(const BlockRV &block_rv, const Array< Integer > new_order)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ad75e0424902b06dca23d46807a9a47d5">ReverseComputeAt</a>(const BlockRV &block_rv, const LoopRV &loop_rv, bool preserve_unit_loops, int index=-1)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a99c902d903680da14339842dd2fd29c7">ReverseComputeInline</a>(const BlockRV &block)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab185c8eac1065290d84d58e7f4617232">RFactor</a>(const LoopRV &loop_rv, int factor_axis)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab1d1f70230fa5f01d406fc122e62b190">RollingBuffer</a>(const BlockRV &block_rv, int write_buffer_index)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ad94d79729ac85aa7c976e23d39066383">RuntimeTypeIndex</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">static</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab9d2b3a98335b88f168b69deed49eb19">SampleCategorical</a>(const Array< Integer > &candidates, const Array< FloatImm > &probs, Optional< Integer > decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abf9fbec94271b7512c24b6eced230c39">SampleComputeLocation</a>(const BlockRV &block_rv, Optional< Integer > decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2c62b3f9486dd35714df50bc424d6698">SamplePerfectTile</a>(const LoopRV &loop_rv, int n, int max_innermost_factor, Optional< Array< Integer >> decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#aae5808dc2e987bf17ef42196457a654d">Schedule</a> class</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">friend</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a3cb60d6112fe5a443ef39bc005c9fbf1">Seed</a>(support::LinearCongruentialEngine::TRandState seed)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a025b5eef0c2516fc1f72eed9ced88807">SetAxisSeparator</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type, const Array< IntImm > &axis_separators)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#aa4760135d373af488a08aaeba7114c48">SetScope</a>(const BlockRV &block_rv, int buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac190a0ab76d8754a35209479bcc6dfa2">Split</a>(const LoopRV &loop_rv, const Array< Optional< ExprRV >> &factors, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26">state</a>() const =0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a93d1d23f24d903db844f75f51fe09a36">StorageAlign</a>(const BlockRV &block_rv, int buffer_index, int axis, int factor, int offset)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf">Tensorize</a>(const LoopRV &loop_rv, const String &intrin, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928">Tensorize</a>(const BlockRV &block_rv, const String &intrin, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a953bca4123b5a758adfdcd65634a5f3b">trace</a>() const =0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a998b22e37ef63a697a984c8ebcc39ca2">TransformBlockLayout</a>(const BlockRV &block_rv, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af4943cc242cec3064a5127515c22219b">TransformLayout</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type, const IndexMap &index_map, const Optional< IndexMap > &pad_value=NullOpt, bool assume_injective_transform=false)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mla [...]
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abc4294398b140f3ff13a33f94a2f9e5f">TVM_DECLARE_FINAL_OBJECT_INFO</a>(ScheduleNode, runtime::Object)</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a481f01923b14e1851ebd38506e9c66ea">type_index</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a4bfc2586cb55f2af47728187b3256255">type_index_</a></td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">protected</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a817ba6c23b7ee1821c48a75edf255a30">TypeIndex2Key</a>(uint32_t tindex)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a6ee32a02dd44257da105fbbe5d9c8622">TypeIndex2KeyHash</a>(uint32_t tindex)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a6841f97e06e6614dd7e82c6dd41b818a">TypeKey2Index</a>(const std::string &key)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a45cd553c09ec836dfcbff81379647f07">Unannotate</a>(const LoopRV &loop_rv, const String &ann_key)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a7c310bca5d1583e61a3f27052a1dd5d0">Unannotate</a>(const BlockRV &block_rv, const String &ann_key)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7">unique</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552">Unroll</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac797a00135c910d65da297038b930ed6">UnsafeSetDType</a>(const BlockRV &block_rv, int buffer_index, const String &dtype)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee">Vectorize</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#acb747d074e1f99477f7132e4614221a3">WorkOn</a>(const String &func_name)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ad66f22b795a1e34cb3c42e691e5864a7">WriteAt</a>(const LoopRV &loop_rv, const BlockRV &block_rv, int write_buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ae637f126412479ed9bec05fd55376f7f">~ScheduleNode</a>()=default</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec">Merge</a>(const Array< LoopRV > &loop_rvs)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a6dd7ec20629e09cd0be1aa49e5f57c12">mod</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a133436a9ec5c4a768b94102bf95a660b">Object</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ab7968feb6ad38ecaffc320e13819d826">Object</a>(const Object &other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#aa1612f69ea5b4225d4cda759cd517323">Object</a>(Object &&other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a69c32fbd96181f5c21d2c878ab285e4f">operator=</a>(const Object &other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ae341e561272ff43cdcbc927bc29ac50d">operator=</a>(Object &&other)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a1ac39c82aee1f8de30d5871d5923fc24">PadEinsum</a>(const BlockRV &block_rv, const Array< Integer > &padding)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a553dc17c0b49b175cd16881c81b6c789">Parallel</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2f454daf29e582a65ffe361e958122df">ReadAt</a>(const LoopRV &loop_rv, const BlockRV &block_rv, int read_buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a0d492efee331e2239a093f4b2017c10f">ref_counter_</a></td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">protected</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a55549a6c23987890246248682560a03d">RefCounterType</a> typedef</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a9e36a8a0e37a76e55068dd534e28c8c5">ReIndex</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2625f87c74fe0cbc95006c763bb172b3">ReindexCacheRead</a>(const BlockRV &block_rv, int read_buffer_index, const String &storage_scope, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af3983a7f5d094529049d83ce22d7b729">ReindexCacheWrite</a>(const BlockRV &block_rv, int write_buffer_index, const String &storage_scope, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a70d353bb52f6fa29fedeb90a6ff872d5">RemoveRV</a>(const BlockRV &block_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a7c44d4f4ea662291ccb9d79383b6fefe">RemoveRV</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a00fcf343d2bc8f36f170c04e5e29d2dc">RemoveRV</a>(const ExprRV &expr_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a059229fe0e254961da406807a97f7a3d">Reorder</a>(const Array< LoopRV > &ordered_loop_rvs)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a3c3024de7f2da68069e593bb8ad64f7f">ReorderBlockIterVar</a>(const BlockRV &block_rv, const Array< Integer > new_order)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ad75e0424902b06dca23d46807a9a47d5">ReverseComputeAt</a>(const BlockRV &block_rv, const LoopRV &loop_rv, bool preserve_unit_loops, int index=-1)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a99c902d903680da14339842dd2fd29c7">ReverseComputeInline</a>(const BlockRV &block)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab185c8eac1065290d84d58e7f4617232">RFactor</a>(const LoopRV &loop_rv, int factor_axis)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab1d1f70230fa5f01d406fc122e62b190">RollingBuffer</a>(const BlockRV &block_rv, int write_buffer_index)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#ad94d79729ac85aa7c976e23d39066383">RuntimeTypeIndex</a>()</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">static</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab9d2b3a98335b88f168b69deed49eb19">SampleCategorical</a>(const Array< Integer > &candidates, const Array< FloatImm > &probs, Optional< Integer > decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abf9fbec94271b7512c24b6eced230c39">SampleComputeLocation</a>(const BlockRV &block_rv, Optional< Integer > decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a2c62b3f9486dd35714df50bc424d6698">SamplePerfectTile</a>(const LoopRV &loop_rv, int n, int max_innermost_factor, Optional< Array< Integer >> decision=NullOpt)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#aae5808dc2e987bf17ef42196457a654d">Schedule</a> class</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">friend</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a3cb60d6112fe5a443ef39bc005c9fbf1">Seed</a>(support::LinearCongruentialEngine::TRandState seed)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a025b5eef0c2516fc1f72eed9ced88807">SetAxisSeparator</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type, const Array< IntImm > &axis_separators)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#aa4760135d373af488a08aaeba7114c48">SetScope</a>(const BlockRV &block_rv, int buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac190a0ab76d8754a35209479bcc6dfa2">Split</a>(const LoopRV &loop_rv, const Array< Optional< ExprRV >> &factors, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abb3612c2598fa2d3ee0e6e3fc3de8a26">state</a>() const =0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a93d1d23f24d903db844f75f51fe09a36">StorageAlign</a>(const BlockRV &block_rv, int buffer_index, int axis, int factor, int offset)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a66983e2dde6aeb18b443616398fff8bf">Tensorize</a>(const LoopRV &loop_rv, const String &intrin, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#afa67abdb51145a49d42cd1464429d928">Tensorize</a>(const BlockRV &block_rv, const String &intrin, bool preserve_unit_iters=true)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a953bca4123b5a758adfdcd65634a5f3b">trace</a>() const =0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a998b22e37ef63a697a984c8ebcc39ca2">TransformBlockLayout</a>(const BlockRV &block_rv, const IndexMap &index_map)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#af4943cc242cec3064a5127515c22219b">TransformLayout</a>(const BlockRV &block_rv, int buffer_index, BufferIndexType buffer_index_type, const IndexMap &index_map, const Optional< IndexMap > &pad_value=NullOpt, bool assume_injective_transform=false)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><sp [...]
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#abc4294398b140f3ff13a33f94a2f9e5f">TVM_DECLARE_FINAL_OBJECT_INFO</a>(ScheduleNode, runtime::Object)</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a481f01923b14e1851ebd38506e9c66ea">type_index</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a4bfc2586cb55f2af47728187b3256255">type_index_</a></td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">protected</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a817ba6c23b7ee1821c48a75edf255a30">TypeIndex2Key</a>(uint32_t tindex)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a6ee32a02dd44257da105fbbe5d9c8622">TypeIndex2KeyHash</a>(uint32_t tindex)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#a6841f97e06e6614dd7e82c6dd41b818a">TypeKey2Index</a>(const std::string &key)</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">static</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a45cd553c09ec836dfcbff81379647f07">Unannotate</a>(const LoopRV &loop_rv, const String &ann_key)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a7c310bca5d1583e61a3f27052a1dd5d0">Unannotate</a>(const BlockRV &block_rv, const String &ann_key)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html#afd548730a6139d19fe24473ad66026d7">unique</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a84ec742f6295f59390592a6d0d90a552">Unroll</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac797a00135c910d65da297038b930ed6">UnsafeSetDType</a>(const BlockRV &block_rv, int buffer_index, const String &dtype)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ab4a8cd91959ceab22855ec338978bcee">Vectorize</a>(const LoopRV &loop_rv)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#acb747d074e1f99477f7132e4614221a3">WorkOn</a>(const String &func_name)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ad66f22b795a1e34cb3c42e691e5864a7">WriteAt</a>(const LoopRV &loop_rv, const BlockRV &block_rv, int write_buffer_index, const String &storage_scope)=0</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">pure virtual</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ae637f126412479ed9bec05fd55376f7f">~ScheduleNode</a>()=default</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html">tvm::tir::ScheduleNode</a></td><td class="entry"><span class="mlabel">virtual</span></td></tr>
</table></div><!-- contents -->
<!-- start footer part -->
<hr class="footer"/><address class="footer"><small>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
index 65a22816e1..2caf2c09e3 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode.html
@@ -174,6 +174,9 @@ Public Member Functions</h2></td></tr>
<tr class="memitem:ac5e28e0b470d0e9299d953c2ee3f6261"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1runtime_1_1Array.html">Array</a>< <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> > </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#ac5e28e0b470d0e9299d953c2ee3f6261">GetConsumers</a> (const <a class="el" href="classtvm_1_1tir_1_1BlockRV.html">BlockRV</a> &blo [...]
<tr class="memdesc:ac5e28e0b470d0e9299d953c2ee3f6261"><td class="mdescLeft"> </td><td class="mdescRight">Get the consumers of a specific block, under the same block scope. <a href="#ac5e28e0b470d0e9299d953c2ee3f6261">More...</a><br /></td></tr>
<tr class="separator:ac5e28e0b470d0e9299d953c2ee3f6261"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a19b026c08dbd41a517544fac0866adec"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec">Merge</a> (const <a class="el" href="classtvm_1_1runtime_1_1Array.html">Array</a>< <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> > &loop_rvs)=0</td></tr>
+<tr class="memdesc:a19b026c08dbd41a517544fac0866adec"><td class="mdescLeft"> </td><td class="mdescRight">Merge a list of loops into one. The loops under their LCA requires: 1) Under the same scope 2) Can't have annotations or thread bindings 3) Start with 0 and have same extent and same nesting depth 4) From target loop to their LCA, the inner loop must be the only child of the outer loop. <a href="#a19b026c08dbd41a517544fac0866adec">More...</a><br /></td></tr>
+<tr class="separator:a19b026c08dbd41a517544fac0866adec"><td class="memSeparator" colspan="2"> </td></tr>
<tr class="memitem:a5b9a41d190be0f8d817b4936732bd0ef"><td class="memItemLeft" align="right" valign="top">virtual <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a5b9a41d190be0f8d817b4936732bd0ef">Fuse</a> (const <a class="el" href="classtvm_1_1runtime_1_1Array.html">Array</a>< <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> > &loop_rvs, bool [...]
<tr class="memdesc:a5b9a41d190be0f8d817b4936732bd0ef"><td class="mdescLeft"> </td><td class="mdescRight">Fuse a list of consecutive loops into one. It requires: 1) The loops can't have annotations or thread bindings. 2) The (i+1)-th loop must be the only child of the i-th loop. 3) All loops must start with 0. 4) The domain of a loop to be fused cannot depend on another loop to be fused. <a href="#a5b9a41d190be0f8d817b4936732bd0ef">More...</a><br /></td></tr>
<tr class="separator:a5b9a41d190be0f8d817b4936732bd0ef"><td class="memSeparator" colspan="2"> </td></tr>
@@ -1732,6 +1735,41 @@ Additional Inherited Members</h2></td></tr>
</dl>
<dl class="section return"><dt>Returns</dt><dd>Whether the corresponding block exists </dd></dl>
+</div>
+</div>
+<a id="a19b026c08dbd41a517544fac0866adec"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a19b026c08dbd41a517544fac0866adec">◆ </a></span>Merge()</h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+ <tr>
+ <td class="mlabels-left">
+ <table class="memname">
+ <tr>
+ <td class="memname">virtual <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> tvm::tir::ScheduleNode::Merge </td>
+ <td>(</td>
+ <td class="paramtype">const <a class="el" href="classtvm_1_1runtime_1_1Array.html">Array</a>< <a class="el" href="classtvm_1_1tir_1_1LoopRV.html">LoopRV</a> > & </td>
+ <td class="paramname"><em>loop_rvs</em></td><td>)</td>
+ <td></td>
+ </tr>
+ </table>
+ </td>
+ <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">pure virtual</span></span> </td>
+ </tr>
+</table>
+</div><div class="memdoc">
+
+<p>Merge a list of loops into one. The loops under their LCA requires: 1) Under the same scope 2) Can't have annotations or thread bindings 3) Start with 0 and have same extent and same nesting depth 4) From target loop to their LCA, the inner loop must be the only child of the outer loop. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+ <table class="params">
+ <tr><td class="paramname">loop_rvs</td><td>The loops to be merged </td></tr>
+ </table>
+ </dd>
+</dl>
+<dl class="section return"><dt>Returns</dt><dd>The new loop after merge </dd></dl>
+
</div>
</div>
<a id="a6dd7ec20629e09cd0be1aa49e5f57c12"></a>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
index 3eadfbabfe..311a4d996d 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__coll__graph.svg
@@ -27,7 +27,7 @@
<text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Seed()</text>
<text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ForkSeed()</text>
<text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Get()</text>
-<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 61 more...</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 62 more...</text>
</g>
<!-- Node3 -->
<g id="node2" class="node">
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
index a35a91c7f8..b4ea75427f 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1ScheduleNode__inherit__graph.svg
@@ -27,7 +27,7 @@
<text text-anchor="start" x="8" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Seed()</text>
<text text-anchor="start" x="8" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ForkSeed()</text>
<text text-anchor="start" x="8" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Get()</text>
-<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 61 more...</text>
+<text text-anchor="start" x="8" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 62 more...</text>
</g>
<!-- Node1 -->
<g id="node2" class="node">
diff --git a/docs/reference/api/doxygen/database_8h_source.html b/docs/reference/api/doxygen/database_8h_source.html
index 5aef87fa1c..69eb8eea51 100644
--- a/docs/reference/api/doxygen/database_8h_source.html
+++ b/docs/reference/api/doxygen/database_8h_source.html
@@ -85,7 +85,7 @@ $(function() {
<div class="ttc" id="structtvm_1_1meta__schedule_1_1WorkloadHash_html_a7cb09ddc6c76d9d00ddbeab8502d97cb"><div class="ttname"><a href="structtvm_1_1meta__schedule_1_1WorkloadHash.html#a7cb09ddc6c76d9d00ddbeab8502d97cb">tvm::meta_schedule::WorkloadHash::operator()</a></div><div class="ttdeci">size_t operator()(const Workload &a) const</div><div class="ttdef"><b>Definition:</b> database.h:96</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1PyDatabaseNode_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyDatabaseNode.html">tvm::meta_schedule::PyDatabaseNode</a></div><div class="ttdoc">The database with customized methods on the python-side. </div><div class="ttdef"><b>Definition:</b> database.h:277</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="structtvm_1_1meta__schedule_1_1WorkloadEqual_html_a49b0e137a01a278469ff63729afae804"><div class="ttname"><a href="structtvm_1_1meta__schedule_1_1WorkloadEqual.html#a49b0e137a01a278469ff63729afae804">tvm::meta_schedule::WorkloadEqual::WorkloadEqual</a></div><div class="ttdeci">WorkloadEqual(const ModuleEquality &mod_eq)</div><div class="ttdef"><b>Definition:</b> database.h:101</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1TuningRecordNode_html_a8cc2d64f796593a1a774eef259f17b29"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuningRecordNode.html#a8cc2d64f796593a1a774eef259f17b29">tvm::meta_schedule::TuningRecordNode::trace</a></div><div class="ttdeci">tir::Trace trace</div><div class="ttdoc">The trace tuned. </div><div class="ttdef"><b>Definition:</b> database.h:117</div></div>
<div class="ttc" id="arg__info_8h_html"><div class="ttname"><a href="arg__info_8h.html">arg_info.h</a></div></div>
diff --git a/docs/reference/api/doxygen/functions_func_m.html b/docs/reference/api/doxygen/functions_func_m.html
index b58f4fd202..d5b485d69f 100644
--- a/docs/reference/api/doxygen/functions_func_m.html
+++ b/docs/reference/api/doxygen/functions_func_m.html
@@ -141,6 +141,7 @@ $(function() {
</li>
<li>Merge()
: <a class="el" href="classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21">tvm::Span</a>
+, <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec">tvm::tir::ScheduleNode</a>
</li>
<li>Metadata()
: <a class="el" href="classtvm_1_1runtime_1_1metadata_1_1Metadata.html#a27735af93ebc35d49585dd7bba5ac099">tvm::runtime::metadata::Metadata</a>
@@ -192,7 +193,7 @@ $(function() {
: <a class="el" href="classtvm_1_1arith_1_1ModularSet.html#a9f54896d98169246c6a24cc338fde500">tvm::arith::ModularSet</a>
</li>
<li>Module()
-: <a class="el" href="classtvm_1_1runtime_1_1Module.html#abfbc619b3b3166d63ec52e399c24bed9">tvm::runtime::Module</a>
+: <a class="el" href="classtvm_1_1runtime_1_1Module.html#abd1380b3f813c2b6acefca3aaef425f4">tvm::runtime::Module</a>
</li>
<li>Move()
: <a class="el" href="structtvm_1_1runtime_1_1vm_1_1Instruction.html#a162dc8d73dc2306f066c3ee013ff096f">tvm::runtime::vm::Instruction</a>
diff --git a/docs/reference/api/doxygen/functions_m.html b/docs/reference/api/doxygen/functions_m.html
index 5b224f73b1..0c12d0e592 100644
--- a/docs/reference/api/doxygen/functions_m.html
+++ b/docs/reference/api/doxygen/functions_m.html
@@ -226,6 +226,7 @@ $(function() {
</li>
<li>Merge()
: <a class="el" href="classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21">tvm::Span</a>
+, <a class="el" href="classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec">tvm::tir::ScheduleNode</a>
</li>
<li>message
: <a class="el" href="classtvm_1_1DiagnosticNode.html#a6cb3da3e81e6d6dce39a046c18f55f04">tvm::DiagnosticNode</a>
diff --git a/docs/reference/api/doxygen/measure__candidate_8h_source.html b/docs/reference/api/doxygen/measure__candidate_8h_source.html
index 1a460da33d..66d1b80eeb 100644
--- a/docs/reference/api/doxygen/measure__candidate_8h_source.html
+++ b/docs/reference/api/doxygen/measure__candidate_8h_source.html
@@ -71,7 +71,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html_a99858dbe74082cc52938ac942523d792"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html#a99858dbe74082cc52938ac942523d792">tvm::meta_schedule::MeasureCandidateNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(tvm::AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> measure_candidate.h:40</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html_a6891e92cac8712bb690401ed121ae7e8"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html#a6891e92cac8712bb690401ed121ae7e8">tvm::meta_schedule::MeasureCandidateNode::args_info</a></div><div class="ttdeci">Array< ArgInfo > args_info</div><div class="ttdoc">The argument information, e.g., (shape, dtype) for tensors. </div><div class="ttdef"><b>Definition:</b> measure_candidate. [...]
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="arg__info_8h_html"><div class="ttname"><a href="arg__info_8h.html">arg_info.h</a></div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1MeasureCandidateNode_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1MeasureCandidateNode.html">tvm::meta_schedule::MeasureCandidateNode</a></div><div class="ttdoc">The schedule (with input shapes) to be measured. </div><div class="ttdef"><b>Definition:</b> measure_candidate.h:33</div></div>
<div class="ttc" id="array_8h_html"><div class="ttname"><a href="array_8h.html">array.h</a></div><div class="ttdoc">Runtime Array container types. </div></div>
diff --git a/docs/reference/api/doxygen/postproc_8h_source.html b/docs/reference/api/doxygen/postproc_8h_source.html
index f3bd3c8d53..f499c168e0 100644
--- a/docs/reference/api/doxygen/postproc_8h_source.html
+++ b/docs/reference/api/doxygen/postproc_8h_source.html
@@ -75,7 +75,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1PyPostprocNode_html_a3771e585727ef6dfecc502ffe57fd2a2"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyPostprocNode.html#a3771e585727ef6dfecc502ffe57fd2a2">tvm::meta_schedule::PyPostprocNode::f_apply</a></div><div class="ttdeci">FApply f_apply</div><div class="ttdoc">The packed function to the Apply function. </div><div class="ttdef"><b>Definition:</b> postproc.h:189</div></div>
<div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1TuneContext_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuneContext.html">tvm::meta_schedule::TuneContext</a></div><div class="ttdoc">Managed reference to TuneContextNode. </div><div class="ttdef"><b>Definition:</b> tune_context.h:95</div></div>
<div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1Postproc_html_a1b95aac48704d0c0740ede2040b942bb"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1Postproc.html#a1b95aac48704d0c0740ede2040b942bb">tvm::meta_schedule::Postproc::FAsString</a></div><div class="ttdeci">runtime::TypedPackedFunc< String()> FAsString</div><div class="ttdoc">Get the postprocessor function as string with name. </div><div class="ttdef"><b>Definition:</b> postproc.h:94</div></div>
diff --git a/docs/reference/api/doxygen/schedule__rule_8h_source.html b/docs/reference/api/doxygen/schedule__rule_8h_source.html
index 0ffeba1e7d..dac8c75381 100644
--- a/docs/reference/api/doxygen/schedule__rule_8h_source.html
+++ b/docs/reference/api/doxygen/schedule__rule_8h_source.html
@@ -80,7 +80,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
<div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1PyScheduleRuleNode_html_a752192bcb5385b1ba72b7c1856c6f360"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1PyScheduleRuleNode.html#a752192bcb5385b1ba72b7c1856c6f360">tvm::meta_schedule::PyScheduleRuleNode::f_apply</a></div><div class="ttdeci">FApply f_apply</div><div class="ttdoc">The packed function to the Apply function. </div><div class="ttdef"><b>Definition:</b> schedule_rule.h:320</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="classtvm_1_1meta__schedule_1_1TuneContext_html"><div class="ttname"><a href="classtvm_1_1meta__schedule_1_1TuneContext.html">tvm::meta_schedule::TuneContext</a></div><div class="ttdoc">Managed reference to TuneContextNode. </div><div class="ttdef"><b>Definition:</b> tune_context.h:95</div></div>
<div class="ttc" id="array_8h_html"><div class="ttname"><a href="array_8h.html">array.h</a></div><div class="ttdoc">Runtime Array container types. </div></div>
<div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
diff --git a/docs/reference/api/doxygen/search/all_e.js b/docs/reference/api/doxygen/search/all_e.js
index fc002ba3f4..52705e9d7f 100644
--- a/docs/reference/api/doxygen/search/all_e.js
+++ b/docs/reference/api/doxygen/search/all_e.js
@@ -124,7 +124,7 @@ var searchData=
['memorymanager',['MemoryManager',['../classtvm_1_1runtime_1_1vm_1_1MemoryManager.html',1,'tvm::runtime::vm']]],
['memorymanagerinterface',['MemoryManagerInterface',['../structMemoryManagerInterface.html',1,'MemoryManagerInterface'],['../page__allocator_8h.html#a6abc29ca90e4cfe4b5714cbfba0a3ca6',1,'MemoryManagerInterface(): page_allocator.h']]],
['memoryscope',['MemoryScope',['../namespacetvm.html#a8a6f73ead9c7ce068f6337ed16e8c53c',1,'tvm']]],
- ['merge',['Merge',['../classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21',1,'tvm::Span::Merge()'],['../namespacetvm_1_1runtime.html#aff337677f23f7d665960f553fb52ab86',1,'tvm::runtime::Merge()']]],
+ ['merge',['Merge',['../classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21',1,'tvm::Span::Merge()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec',1,'tvm::tir::ScheduleNode::Merge()'],['../namespacetvm_1_1runtime.html#aff337677f23f7d665960f553fb52ab86',1,'tvm::runtime::Merge()']]],
['mergedynamicsharedmemoryallocations',['MergeDynamicSharedMemoryAllocations',['../namespacetvm_1_1tir_1_1transform.html#a070137dbab05c63d4af6b77c25868a66',1,'tvm::tir::transform']]],
['meshgrid',['meshgrid',['../namespacetvm_1_1topi.html#a453f9f5a35a9852a086ce0228ee22164',1,'tvm::topi']]],
['meshgridattrs',['MeshgridAttrs',['../structtvm_1_1relay_1_1MeshgridAttrs.html',1,'tvm::relay']]],
diff --git a/docs/reference/api/doxygen/search/functions_d.js b/docs/reference/api/doxygen/search/functions_d.js
index ba58483ed7..bdc98e1453 100644
--- a/docs/reference/api/doxygen/search/functions_d.js
+++ b/docs/reference/api/doxygen/search/functions_d.js
@@ -48,7 +48,7 @@ var searchData=
['measureinput',['MeasureInput',['../classtvm_1_1auto__scheduler_1_1MeasureInput.html#aeba14979d93b0302ede5668b168098af',1,'tvm::auto_scheduler::MeasureInput']]],
['measureresult',['MeasureResult',['../classtvm_1_1auto__scheduler_1_1MeasureResult.html#a5023e3b9ad54d569fa7435aade73fd7d',1,'tvm::auto_scheduler::MeasureResult']]],
['memorydatabase',['MemoryDatabase',['../classtvm_1_1meta__schedule_1_1Database.html#af410a88a002698bc01161aae39abf84c',1,'tvm::meta_schedule::Database']]],
- ['merge',['Merge',['../classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21',1,'tvm::Span::Merge()'],['../namespacetvm_1_1runtime.html#aff337677f23f7d665960f553fb52ab86',1,'tvm::runtime::Merge()']]],
+ ['merge',['Merge',['../classtvm_1_1Span.html#ae8e7ed175f50096b7dd2bccd12998e21',1,'tvm::Span::Merge()'],['../classtvm_1_1tir_1_1ScheduleNode.html#a19b026c08dbd41a517544fac0866adec',1,'tvm::tir::ScheduleNode::Merge()'],['../namespacetvm_1_1runtime.html#aff337677f23f7d665960f553fb52ab86',1,'tvm::runtime::Merge()']]],
['mergedynamicsharedmemoryallocations',['MergeDynamicSharedMemoryAllocations',['../namespacetvm_1_1tir_1_1transform.html#a070137dbab05c63d4af6b77c25868a66',1,'tvm::tir::transform']]],
['meshgrid',['meshgrid',['../namespacetvm_1_1topi.html#a453f9f5a35a9852a086ce0228ee22164',1,'tvm::topi']]],
['meta_5fschedule_5flayout_5ftransform',['meta_schedule_layout_transform',['../namespacetvm_1_1topi.html#a3df09fe365656f953c8145a8cf0b3fa3',1,'tvm::topi']]],
diff --git a/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html b/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
index 167ae7aa31..a0507f6b01 100644
--- a/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
+++ b/docs/reference/api/doxygen/tir_2schedule_2schedule_8h_source.html
@@ -66,7 +66,7 @@ $(function() {
<div class="title">schedule.h</div> </div>
</div><!--header-->
<div class="contents">
-<a href="tir_2schedule_2schedule_8h.html">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno"> 1</span> <span class="comment">/*</span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span> <span class="comment"> * Licensed to the Apache Software Foundation (ASF) under one</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment [...]
+<a href="tir_2schedule_2schedule_8h.html">Go to the documentation of this file.</a><div class="fragment"><div class="line"><a name="l00001"></a><span class="lineno"> 1</span> <span class="comment">/*</span></div><div class="line"><a name="l00002"></a><span class="lineno"> 2</span> <span class="comment"> * Licensed to the Apache Software Foundation (ASF) under one</span></div><div class="line"><a name="l00003"></a><span class="lineno"> 3</span> <span class="comment [...]
<div class="ttc" id="namespacetvm_1_1script_1_1ir__builder_1_1tir_html_acd41556b0c4088d0f309ef5495aaebe3"><div class="ttname"><a href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acd41556b0c4088d0f309ef5495aaebe3">tvm::script::ir_builder::tir::Unroll</a></div><div class="ttdeci">ForFrame Unroll(PrimExpr start, PrimExpr stop, Optional< Map< String, ObjectRef >> annotations=NullOpt)</div><div class="ttdoc">The unrolled For statement. </div></div>
<div class="ttc" id="classtvm_1_1tir_1_1StmtNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1StmtNode.html">tvm::tir::StmtNode</a></div><div class="ttdoc">Base node of all statements. </div><div class="ttdef"><b>Definition:</b> stmt.h:38</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1BlockRVNode_html_af90b398c502892d19ff3bdf6463d32ab"><div class="ttname"><a href="classtvm_1_1tir_1_1BlockRVNode.html#af90b398c502892d19ff3bdf6463d32ab">tvm::tir::BlockRVNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(tvm::AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> schedule.h:53</div></div>
@@ -84,7 +84,7 @@ $(function() {
<div class="ttc" id="object_8h_html_aaaa3dc5b6dc33f84b2d28f9a81267212"><div class="ttname"><a href="object_8h.html#aaaa3dc5b6dc33f84b2d28f9a81267212">TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_MUTABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:744</div></div>
<div class="ttc" id="namespacetvm_1_1tir_html_a9ae244600a5e56c4adc9faf6d88f931e"><div class="ttname"><a href="namespacetvm_1_1tir.html#a9ae244600a5e56c4adc9faf6d88f931e">tvm::tir::ScheduleErrorRenderLevel</a></div><div class="ttdeci">ScheduleErrorRenderLevel</div><div class="ttdoc">The level of detailed error message rendering. </div><div class="ttdef"><b>Definition:</b> schedule.h:31</div></div>
<div class="ttc" id="namespacetvm_1_1tir_html_a9ae244600a5e56c4adc9faf6d88f931ead6733547bb237ce06cddf96357f1b66b"><div class="ttname"><a href="namespacetvm_1_1tir.html#a9ae244600a5e56c4adc9faf6d88f931ead6733547bb237ce06cddf96357f1b66b">tvm::tir::ScheduleErrorRenderLevel::kDetail</a></div><div class="ttdoc">Render a detailed error message. </div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="index__map_8h_html"><div class="ttname"><a href="index__map_8h.html">index_map.h</a></div><div class="ttdoc">Defines a remapping of buffer indices. </div></div>
<div class="ttc" id="classtvm_1_1support_1_1LinearCongruentialEngine_html_a4d3a3a94a3f3d2dfab4b5ccb1a7e97de"><div class="ttname"><a href="classtvm_1_1support_1_1LinearCongruentialEngine.html#a4d3a3a94a3f3d2dfab4b5ccb1a7e97de">tvm::support::LinearCongruentialEngine::TRandState</a></div><div class="ttdeci">int64_t TRandState</div><div class="ttdef"><b>Definition:</b> random_engine.h:46</div></div>
<div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
@@ -105,6 +105,7 @@ $(function() {
<div class="ttc" id="object_8h_html_a3aea9b3f65aeb9150c0fa7800e5573c6"><div class="ttname"><a href="object_8h.html#a3aea9b3f65aeb9150c0fa7800e5573c6">TVM_DECLARE_FINAL_OBJECT_INFO</a></div><div class="ttdeci">#define TVM_DECLARE_FINAL_OBJECT_INFO(TypeName, ParentType)</div><div class="ttdoc">helper macro to declare type information in a final class. </div><div class="ttdef"><b>Definition:</b> object.h:671</div></div>
<div class="ttc" id="classtvm_1_1IRModule_html"><div class="ttname"><a href="classtvm_1_1IRModule.html">tvm::IRModule</a></div><div class="ttdoc">Managed reference class to IRModuleNode. </div><div class="ttdef"><b>Definition:</b> module.h:348</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1LoopRVNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1LoopRVNode.html">tvm::tir::LoopRVNode</a></div><div class="ttdoc">A random variable that evaluates to a TensorIR for loop. </div><div class="ttdef"><b>Definition:</b> schedule.h:72</div></div>
+<div class="ttc" id="namespacetvm_1_1runtime_html_aff337677f23f7d665960f553fb52ab86"><div class="ttname"><a href="namespacetvm_1_1runtime.html#aff337677f23f7d665960f553fb52ab86">tvm::runtime::Merge</a></div><div class="ttdeci">Map< K, V > Merge(Map< K, V > lhs, const Map< K, V > &rhs)</div><div class="ttdoc">Merge two Maps. </div><div class="ttdef"><b>Definition:</b> map.h:1471</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1BlockRVNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1BlockRVNode.html">tvm::tir::BlockRVNode</a></div><div class="ttdoc">A random variable that evaluates to a TensorIR block. </div><div class="ttdef"><b>Definition:</b> schedule.h:51</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Optional_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Optional.html">tvm::runtime::Optional</a></div><div class="ttdoc">Optional container that to represent to a Nullable variant of T. </div><div class="ttdef"><b>Definition:</b> optional.h:51</div></div>
<div class="ttc" id="namespacetvm_1_1tir_html_a1c8232edeb2fcce8eb95477c5153237a"><div class="ttname"><a href="namespacetvm_1_1tir.html#a1c8232edeb2fcce8eb95477c5153237a">tvm::tir::BufferIndexType</a></div><div class="ttdeci">BufferIndexType</div><div class="ttdoc">Type of buffer index. </div><div class="ttdef"><b>Definition:</b> schedule.h:41</div></div>
diff --git a/docs/reference/api/doxygen/trace_8h_source.html b/docs/reference/api/doxygen/trace_8h_source.html
index 22344c98df..ca0369bcea 100644
--- a/docs/reference/api/doxygen/trace_8h_source.html
+++ b/docs/reference/api/doxygen/trace_8h_source.html
@@ -76,7 +76,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1tir_html_a75918aeef1136f9d6308556902d5bcae"><div class="ttname"><a href="namespacetvm_1_1tir.html#a75918aeef1136f9d6308556902d5bcae">tvm::tir::FTraceDecisionProvider</a></div><div class="ttdeci">runtime::TypedPackedFunc< ObjectRef(const Instruction &inst, const Array< ObjectRef > &inputs, const Array< ObjectRef > &attrs, const Optional< ObjectRef > &decision)> FTraceDecisionProvider</div><div class="ttdoc">A cal [...]
<div class="ttc" id="instruction_8h_html"><div class="ttname"><a href="instruction_8h.html">instruction.h</a></div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html">tvm::runtime::Object</a></div><div class="ttdoc">base class of all object containers. </div><div class="ttdef"><b>Definition:</b> object.h:167</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:794</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1Schedule_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Schedule.html">tvm::tir::Schedule</a></div><div class="ttdoc">Managed reference to ScheduleNode. </div><div class="ttdef"><b>Definition:</b> schedule.h:804</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1TraceNode_html_ad6c859ed32b1e2ae076355eda37df0a2"><div class="ttname"><a href="classtvm_1_1tir_1_1TraceNode.html#ad6c859ed32b1e2ae076355eda37df0a2">tvm::tir::TraceNode::insts</a></div><div class="ttdeci">Array< Instruction > insts</div><div class="ttdoc">The instructions invoked so far in the program execution. </div><div class="ttdef"><b>Definition:</b> trace.h:61</div></div>
<div class="ttc" id="classtvm_1_1AttrVisitor_html"><div class="ttname"><a href="classtvm_1_1AttrVisitor.html">tvm::AttrVisitor</a></div><div class="ttdoc">Visitor class to get the attributes of an AST/IR node. The content is going to be called for each fie...</div><div class="ttdef"><b>Definition:</b> reflection.h:52</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1TraceNode_html_a764346045e536fa26b56c9e140de8e7b"><div class="ttname"><a href="classtvm_1_1tir_1_1TraceNode.html#a764346045e536fa26b56c9e140de8e7b">tvm::tir::TraceNode::ApplyToSchedule</a></div><div class="ttdeci">void ApplyToSchedule(Schedule sch, bool remove_postproc, FTraceDecisionProvider decision_provider=nullptr) const</div><div class="ttdoc">Apply the trace to a TensorIR schedule. </div></div>
diff --git a/docs/reference/api/python/auto_scheduler.html b/docs/reference/api/python/auto_scheduler.html
index c281ad00cf..af3088cd82 100644
--- a/docs/reference/api/python/auto_scheduler.html
+++ b/docs/reference/api/python/auto_scheduler.html
@@ -1621,7 +1621,7 @@ history states as starting point to perform Evolutionary Search).</p></li>
<dl class="py class">
<dt class="sig sig-object py" id="tvm.auto_scheduler.SketchPolicy">
-<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
+<em class="property"><span class="pre">class</span> </em><span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">SketchPolicy</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">program_cost_model</span></span><span class="o"><span class="pre">=</span></span><span class="defau [...]
<dd><p>The search policy that searches in a hierarchical search space defined by sketches.
The policy randomly samples programs from the space defined by sketches and use evolutionary
search to fine-tune them.</p>
@@ -1905,7 +1905,7 @@ Candidates:
<dl class="py function">
<dt class="sig sig-object py" id="tvm.auto_scheduler.auto_schedule">
-<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
+<span class="sig-prename descclassname"><span class="pre">tvm.auto_scheduler.</span></span><span class="sig-name descname"><span class="pre">auto_schedule</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="n"><span class="pre">task</span></span></em>, <em class="sig-param"><span class="n"><span class="pre">search_policy</span></span><span class="o"><span class="pre">=</span></span><span class="default_value"><span class="pre">None</span></span></em>, <em clas [...]
<dd><p>THIS API IS DEPRECATED.</p>
<p>Run auto scheduling search for a task.</p>
<dl class="field-list simple">
diff --git a/docs/reference/api/python/tir.html b/docs/reference/api/python/tir.html
index 7cd084ea67..17a92ae8bd 100644
--- a/docs/reference/api/python/tir.html
+++ b/docs/reference/api/python/tir.html
@@ -5747,115 +5747,118 @@ preserve the semantics of computation. Some example of schedules:
<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.get_consumers" title="tvm.tir.Schedule.get_consumers"><code class="xref py py-obj docutils literal notranslate"><span class="pre">get_consumers</span></code></a>(block)</p></td>
<td><p>Get the consumers of a specific block</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.fuse" title="tvm.tir.Schedule.fuse"><code class="xref py py-obj docutils literal notranslate"><span class="pre">fuse</span></code></a>(*loops[, preserve_unit_iters])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.merge" title="tvm.tir.Schedule.merge"><code class="xref py py-obj docutils literal notranslate"><span class="pre">merge</span></code></a>(*loops)</p></td>
+<td><p>Merge a list of loops into one.</p></td>
+</tr>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.fuse" title="tvm.tir.Schedule.fuse"><code class="xref py py-obj docutils literal notranslate"><span class="pre">fuse</span></code></a>(*loops[, preserve_unit_iters])</p></td>
<td><p>Fuse a list of consecutive loops into one.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.split" title="tvm.tir.Schedule.split"><code class="xref py py-obj docutils literal notranslate"><span class="pre">split</span></code></a>(loop, factors[, preserve_unit_iters])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.split" title="tvm.tir.Schedule.split"><code class="xref py py-obj docutils literal notranslate"><span class="pre">split</span></code></a>(loop, factors[, preserve_unit_iters])</p></td>
<td><p>Split a loop into a list of consecutive loops.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reorder" title="tvm.tir.Schedule.reorder"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reorder</span></code></a>(*ordered_loops)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reorder" title="tvm.tir.Schedule.reorder"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reorder</span></code></a>(*ordered_loops)</p></td>
<td><p>Reorder a list of loops.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reorder_block_iter_var" title="tvm.tir.Schedule.reorder_block_iter_var"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reorder_block_iter_var</span></code></a>(block, new_order)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reorder_block_iter_var" title="tvm.tir.Schedule.reorder_block_iter_var"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reorder_block_iter_var</span></code></a>(block, new_order)</p></td>
<td><p>Reorder the itervars inside a given block.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.add_unit_loop" title="tvm.tir.Schedule.add_unit_loop"><code class="xref py py-obj docutils literal notranslate"><span class="pre">add_unit_loop</span></code></a>(block_or_loop)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.add_unit_loop" title="tvm.tir.Schedule.add_unit_loop"><code class="xref py py-obj docutils literal notranslate"><span class="pre">add_unit_loop</span></code></a>(block_or_loop)</p></td>
<td><p>Create a new unit loop on top of the specific block or loop.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.parallel" title="tvm.tir.Schedule.parallel"><code class="xref py py-obj docutils literal notranslate"><span class="pre">parallel</span></code></a>(loop)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.parallel" title="tvm.tir.Schedule.parallel"><code class="xref py py-obj docutils literal notranslate"><span class="pre">parallel</span></code></a>(loop)</p></td>
<td><p>Parallelize the input loop.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.vectorize" title="tvm.tir.Schedule.vectorize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">vectorize</span></code></a>(loop)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.vectorize" title="tvm.tir.Schedule.vectorize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">vectorize</span></code></a>(loop)</p></td>
<td><p>Vectorize the input loop.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.bind" title="tvm.tir.Schedule.bind"><code class="xref py py-obj docutils literal notranslate"><span class="pre">bind</span></code></a>(loop, thread_axis)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.bind" title="tvm.tir.Schedule.bind"><code class="xref py py-obj docutils literal notranslate"><span class="pre">bind</span></code></a>(loop, thread_axis)</p></td>
<td><p>Bind the input loop to the given thread axis.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unroll" title="tvm.tir.Schedule.unroll"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unroll</span></code></a>(loop)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unroll" title="tvm.tir.Schedule.unroll"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unroll</span></code></a>(loop)</p></td>
<td><p>Unroll the input loop.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_read" title="tvm.tir.Schedule.cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_read</span></code></a>(block, read_buffer_index, ...[, ...])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_read" title="tvm.tir.Schedule.cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_read</span></code></a>(block, read_buffer_index, ...[, ...])</p></td>
<td><p>Create a block that reads a buffer region into a read cache.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_write" title="tvm.tir.Schedule.cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_write</span></code></a>(block, write_buffer_index, ...)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_write" title="tvm.tir.Schedule.cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_write</span></code></a>(block, write_buffer_index, ...)</p></td>
<td><p>Create a block that reads a buffer region into a write cache.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_read" title="tvm.tir.Schedule.reindex_cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_read</span></code></a>(block, read_buffer_index, ...)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_read" title="tvm.tir.Schedule.reindex_cache_read"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_read</span></code></a>(block, read_buffer_index, ...)</p></td>
<td><p>Create a block that reads a buffer region into a read cache using customized indices specified by index map.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_write" title="tvm.tir.Schedule.reindex_cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_write</span></code></a>(block, ...)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex_cache_write" title="tvm.tir.Schedule.reindex_cache_write"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex_cache_write</span></code></a>(block, ...)</p></td>
<td><p>Create a block that reads a buffer region into a write cache using customized indices specified by index map.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_inplace" title="tvm.tir.Schedule.cache_inplace"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_inplace</span></code></a>(block, read_buffer_index, ...)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_inplace" title="tvm.tir.Schedule.cache_inplace"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_inplace</span></code></a>(block, read_buffer_index, ...)</p></td>
<td><p>Create blocks that reads & write a buffer region into a cache block.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_index" title="tvm.tir.Schedule.cache_index"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_index</span></code></a>(block, storage_scope[, cse_thresh])</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.cache_index" title="tvm.tir.Schedule.cache_index"><code class="xref py py-obj docutils literal notranslate"><span class="pre">cache_index</span></code></a>(block, storage_scope[, cse_thresh])</p></td>
<td><p>Create a block to cache precomputed index for later use.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex" title="tvm.tir.Schedule.reindex"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex</span></code></a>(block, buffer)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reindex" title="tvm.tir.Schedule.reindex"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reindex</span></code></a>(block, buffer)</p></td>
<td><p>Create a block that read/write a buffer region into a read/write cache with reindexing.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.compute_at" title="tvm.tir.Schedule.compute_at"><code class="xref py py-obj docutils literal notranslate"><span class="pre">compute_at</span></code></a>(block, loop[, ...])</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.compute_at" title="tvm.tir.Schedule.compute_at"><code class="xref py py-obj docutils literal notranslate"><span class="pre">compute_at</span></code></a>(block, loop[, ...])</p></td>
<td><p>Compute-At.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reverse_compute_at" title="tvm.tir.Schedule.reverse_compute_at"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reverse_compute_at</span></code></a>(block, loop[, ...])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reverse_compute_at" title="tvm.tir.Schedule.reverse_compute_at"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reverse_compute_at</span></code></a>(block, loop[, ...])</p></td>
<td><p>Reverse-Compute-At.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.compute_inline" title="tvm.tir.Schedule.compute_inline"><code class="xref py py-obj docutils literal notranslate"><span class="pre">compute_inline</span></code></a>(block)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.compute_inline" title="tvm.tir.Schedule.compute_inline"><code class="xref py py-obj docutils literal notranslate"><span class="pre">compute_inline</span></code></a>(block)</p></td>
<td><p>Inline a block into its consumer(s).</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reverse_compute_inline" title="tvm.tir.Schedule.reverse_compute_inline"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reverse_compute_inline</span></code></a>(block)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.reverse_compute_inline" title="tvm.tir.Schedule.reverse_compute_inline"><code class="xref py py-obj docutils literal notranslate"><span class="pre">reverse_compute_inline</span></code></a>(block)</p></td>
<td><p>Inline a block into its only producer.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.decompose_reduction" title="tvm.tir.Schedule.decompose_reduction"><code class="xref py py-obj docutils literal notranslate"><span class="pre">decompose_reduction</span></code></a>(block, loop)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.decompose_reduction" title="tvm.tir.Schedule.decompose_reduction"><code class="xref py py-obj docutils literal notranslate"><span class="pre">decompose_reduction</span></code></a>(block, loop)</p></td>
<td><p>Decompose a reduction block into two separate blocks.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.rfactor" title="tvm.tir.Schedule.rfactor"><code class="xref py py-obj docutils literal notranslate"><span class="pre">rfactor</span></code></a>(loop, factor_axis)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.rfactor" title="tvm.tir.Schedule.rfactor"><code class="xref py py-obj docutils literal notranslate"><span class="pre">rfactor</span></code></a>(loop, factor_axis)</p></td>
<td><p>Factorize an associative reduction block by the specified loop.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.storage_align" title="tvm.tir.Schedule.storage_align"><code class="xref py py-obj docutils literal notranslate"><span class="pre">storage_align</span></code></a>(block, buffer_index, axis, ...)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.storage_align" title="tvm.tir.Schedule.storage_align"><code class="xref py py-obj docutils literal notranslate"><span class="pre">storage_align</span></code></a>(block, buffer_index, axis, ...)</p></td>
<td><p>Set alignment requirement for specific dimension such that stride[axis] == k * factor + offset for some k.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.set_scope" title="tvm.tir.Schedule.set_scope"><code class="xref py py-obj docutils literal notranslate"><span class="pre">set_scope</span></code></a>(block, buffer_index, storage_scope)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.set_scope" title="tvm.tir.Schedule.set_scope"><code class="xref py py-obj docutils literal notranslate"><span class="pre">set_scope</span></code></a>(block, buffer_index, storage_scope)</p></td>
<td><p>Set the storage scope of a buffer, where the buffer is specified by the a block and a write-index.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unsafe_set_dtype" title="tvm.tir.Schedule.unsafe_set_dtype"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unsafe_set_dtype</span></code></a>(block, buffer_index, dtype)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unsafe_set_dtype" title="tvm.tir.Schedule.unsafe_set_dtype"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unsafe_set_dtype</span></code></a>(block, buffer_index, dtype)</p></td>
<td><p>Set the data type of a buffer, where the buffer is specified by the a block and write-index.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.blockize" title="tvm.tir.Schedule.blockize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">blockize</span></code></a>(loop[, preserve_unit_iters])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.blockize" title="tvm.tir.Schedule.blockize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">blockize</span></code></a>(loop[, preserve_unit_iters])</p></td>
<td><p>Convert the subtree rooted at a specific loop into a block.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.tensorize" title="tvm.tir.Schedule.tensorize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tensorize</span></code></a>(block_or_loop, tensor_intrin[, ...])</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.tensorize" title="tvm.tir.Schedule.tensorize"><code class="xref py py-obj docutils literal notranslate"><span class="pre">tensorize</span></code></a>(block_or_loop, tensor_intrin[, ...])</p></td>
<td><p>Tensorize the computation enclosed by loop with the tensor intrinsic.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.annotate" title="tvm.tir.Schedule.annotate"><code class="xref py py-obj docutils literal notranslate"><span class="pre">annotate</span></code></a>(block_or_loop, ann_key, ann_val)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.annotate" title="tvm.tir.Schedule.annotate"><code class="xref py py-obj docutils literal notranslate"><span class="pre">annotate</span></code></a>(block_or_loop, ann_key, ann_val)</p></td>
<td><p>Annotate a block/loop with a key value pair</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unannotate" title="tvm.tir.Schedule.unannotate"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unannotate</span></code></a>(block_or_loop, ann_key)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.unannotate" title="tvm.tir.Schedule.unannotate"><code class="xref py py-obj docutils literal notranslate"><span class="pre">unannotate</span></code></a>(block_or_loop, ann_key)</p></td>
<td><p>Unannotate a block/loop's annotation with key ann_key</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.transform_layout" title="tvm.tir.Schedule.transform_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_layout</span></code></a>(block, buffer, index_map[, ...])</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.transform_layout" title="tvm.tir.Schedule.transform_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_layout</span></code></a>(block, buffer, index_map[, ...])</p></td>
<td><p>Apply a transformation represented by IndexMap to buffer</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.transform_block_layout" title="tvm.tir.Schedule.transform_block_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_block_layout</span></code></a>(block, index_map)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.transform_block_layout" title="tvm.tir.Schedule.transform_block_layout"><code class="xref py py-obj docutils literal notranslate"><span class="pre">transform_block_layout</span></code></a>(block, index_map)</p></td>
<td><p>Apply a transformation represented by IndexMap to block</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.set_axis_separator" title="tvm.tir.Schedule.set_axis_separator"><code class="xref py py-obj docutils literal notranslate"><span class="pre">set_axis_separator</span></code></a>(block, buffer, ...)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.set_axis_separator" title="tvm.tir.Schedule.set_axis_separator"><code class="xref py py-obj docutils literal notranslate"><span class="pre">set_axis_separator</span></code></a>(block, buffer, ...)</p></td>
<td><p>Set the axis separator of a buffer, where the buffer is specified by a block and a read or write index.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.decompose_padding" title="tvm.tir.Schedule.decompose_padding"><code class="xref py py-obj docutils literal notranslate"><span class="pre">decompose_padding</span></code></a>(block, loop)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.decompose_padding" title="tvm.tir.Schedule.decompose_padding"><code class="xref py py-obj docutils literal notranslate"><span class="pre">decompose_padding</span></code></a>(block, loop)</p></td>
<td><p>Decompose a block of padding computation pattern into two separate blocks.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.can_decompose_padding" title="tvm.tir.Schedule.can_decompose_padding"><code class="xref py py-obj docutils literal notranslate"><span class="pre">can_decompose_padding</span></code></a>(block, loop)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.can_decompose_padding" title="tvm.tir.Schedule.can_decompose_padding"><code class="xref py py-obj docutils literal notranslate"><span class="pre">can_decompose_padding</span></code></a>(block, loop)</p></td>
<td><p>Check whether the block match padding pattern and can be decomposed.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.pad_einsum" title="tvm.tir.Schedule.pad_einsum"><code class="xref py py-obj docutils literal notranslate"><span class="pre">pad_einsum</span></code></a>(block, padding)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.pad_einsum" title="tvm.tir.Schedule.pad_einsum"><code class="xref py py-obj docutils literal notranslate"><span class="pre">pad_einsum</span></code></a>(block, padding)</p></td>
<td><p>Pad the computation of Einsum.</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.rolling_buffer" title="tvm.tir.Schedule.rolling_buffer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">rolling_buffer</span></code></a>(block, write_buffer_index)</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.rolling_buffer" title="tvm.tir.Schedule.rolling_buffer"><code class="xref py py-obj docutils literal notranslate"><span class="pre">rolling_buffer</span></code></a>(block, write_buffer_index)</p></td>
<td><p>Compute the target buffer via rolling buffering, select the outermost rollable axis with a positive bound overlap that appears in the block's ancestor loops as <cite>rolling axis</cite>, fold and circularize the buffer along the rolling dimension, append block predicate to avoid recomputing overlapping elements.</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="#tvm.tir.Schedule.enter_postproc" title="tvm.tir.Schedule.enter_postproc"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enter_postproc</span></code></a>()</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="#tvm.tir.Schedule.enter_postproc" title="tvm.tir.Schedule.enter_postproc"><code class="xref py py-obj docutils literal notranslate"><span class="pre">enter_postproc</span></code></a>()</p></td>
<td><p>A no-op that marks the start of postprocessing phase of scheduling</p></td>
</tr>
</tbody>
@@ -6162,6 +6165,74 @@ IndexError is raised if 0 or multiple blocks exist with the specific name.</p>
</dl>
</dd></dl>
+<dl class="py method">
+<dt class="sig sig-object py" id="tvm.tir.Schedule.merge">
+<span class="sig-name descname"><span class="pre">merge</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">loops</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.LoopRV</span><span class="p"><span class="pre">]</span></span></span></em><span [...]
+<dd><p>Merge a list of loops into one. The loops under their LCA requires:
+1) Under the same scope.
+2) Can’t have annotations or thread bindings.
+3) Start with 0 and have same extent and same nesting depth.
+4) From target loop to their LCA, The inner loop must be the only child of the outer loop.</p>
+<dl class="field-list simple">
+<dt class="field-odd">Parameters</dt>
+<dd class="field-odd"><p><strong>*loops</strong> (<em>List</em><em>[</em><em>LoopRV</em><em>]</em>) – The loops to be merged</p>
+</dd>
+<dt class="field-even">Returns</dt>
+<dd class="field-even"><p><strong>fused_loop</strong> – The new loop after merge</p>
+</dd>
+<dt class="field-odd">Return type</dt>
+<dd class="field-odd"><p>LoopRV</p>
+</dd>
+</dl>
+<p class="rubric">Examples</p>
+<p>Before applying merge, in TensorIR, the IR is:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">before_merge</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">c</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span [...]
+ <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="n">C</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">c</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+ <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">"B"</span><span class="p">):</span>
+ <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">"SS"</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+ <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
+ <span class="k">for</span> <span class="n">i</span><span class="p">,</span> <span class="n">j</span> <span class="ow">in</span> <span class="n">T</span><span class="o">.</span><span class="n">grid</span><span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">):</span>
+ <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">"C"</span><span class="p">):</span>
+ <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">"SS"</span><span class="p">,</span> <span class="p">[</span><span class="n">i</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+ <span class="n">C</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="mf">2.0</span>
+</pre></div>
+</div>
+<p>Create the schedule and do fuse:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="n">sch</span> <span class="o">=</span> <span class="n">tir</span><span class="o">.</span><span class="n">Schedule</span><span class="p">(</span><span class="n">before_fuse</span><span class="p">)</span>
+<span class="n">i1</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">sch</span><span class="o">.</span><span class="n">get_loops</span><span class="p">(</span><span class="n">sch</span><span class="o">.</span><span class="n">get_block</span><span class="p">(</span><span class="s2">"B"</span><span class="p">))</span>
+<span class="n">i2</span><span class="p">,</span> <span class="n">_</span> <span class="o">=</span> <span class="n">sch</span><span class="o">.</span><span class="n">get_loops</span><span class="p">(</span><span class="n">sch</span><span class="o">.</span><span class="n">get_block</span><span class="p">(</span><span class="s2">"C"</span><span class="p">))</span>
+<span class="n">sch</span><span class="o">.</span><span class="n">merge</span><span class="p">(</span><span class="n">i1</span><span class="p">,</span> <span class="n">i2</span><span class="p">)</span>
+<span class="nb">print</span><span class="p">(</span><span class="n">sch</span><span class="o">.</span><span class="n">mod</span><span class="p">[</span><span class="s2">"main"</span><span class="p">]</span><span class="o">.</span><span class="n">script</span><span class="p">())</span>
+</pre></div>
+</div>
+<p>After applying fuse, the IR becomes:</p>
+<div class="highlight-python notranslate"><div class="highlight"><pre><span></span><span class="nd">@T</span><span class="o">.</span><span class="n">prim_func</span>
+<span class="k">def</span> <span class="nf">after_fuse</span><span class="p">(</span><span class="n">a</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">b</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span class="n">handle</span><span class="p">,</span> <span class="n">c</span><span class="p">:</span> <span class="n">T</span><span class="o">.</span><span cl [...]
+ <span class="n">A</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">a</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="n">B</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">b</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="n">C</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">match_buffer</span><span class="p">(</span><span class="n">c</span><span class="p">,</span> <span class="p">(</span><span class="mi">128</span><span class="p">,</span> <span class="mi">128</span><span class="p">))</span>
+ <span class="c1"># the 2 loops are merged into 1</span>
+ <span class="k">for</span> <span class="n">i_m</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">128</span><span class="p">):</span>
+ <span class="k">for</span> <span class="n">j</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">128</span><span class="p">):</span>
+ <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">"B"</span><span class="p">):</span>
+ <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">"SS"</span><span class="p">,</span> <span class="p">[</span><span class="n">i_m</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+ <span class="n">T</span><span class="o">.</span><span class="n">reads</span><span class="p">(</span><span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">])</span>
+ <span class="n">T</span><span class="o">.</span><span class="n">writes</span><span class="p">(</span><span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">])</span>
+ <span class="n">B</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="n">T</span><span class="o">.</span><span class="n">float32</span><span class="p">(</span><span class="mi">2</span><sp [...]
+ <span class="k">for</span> <span class="n">j</span> <span class="ow">in</span> <span class="nb">range</span><span class="p">(</span><span class="mi">128</span><span class="p">):</span>
+ <span class="k">with</span> <span class="n">T</span><span class="o">.</span><span class="n">block</span><span class="p">(</span><span class="s2">"C"</span><span class="p">):</span>
+ <span class="n">vi</span><span class="p">,</span> <span class="n">vj</span> <span class="o">=</span> <span class="n">T</span><span class="o">.</span><span class="n">axis</span><span class="o">.</span><span class="n">remap</span><span class="p">(</span><span class="s2">"SS"</span><span class="p">,</span> <span class="p">[</span><span class="n">i_m</span><span class="p">,</span> <span class="n">j</span><span class="p">])</span>
+ <span class="n">T</span><span class="o">.</span><span class="n">reads</span><span class="p">(</span><span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">])</span>
+ <span class="n">T</span><span class="o">.</span><span class="n">writes</span><span class="p">(</span><span class="n">C</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">])</span>
+ <span class="n">C</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">=</span> <span class="n">A</span><span class="p">[</span><span class="n">vi</span><span class="p">,</span> <span class="n">vj</span><span class="p">]</span> <span class="o">*</span> <span class="n">T</span><span class="o">.</span><span class="n">float32</span><span class="p">(</span><span class="mi">2</span><sp [...]
+</pre></div>
+</div>
+</dd></dl>
+
<dl class="py method">
<dt class="sig sig-object py" id="tvm.tir.Schedule.fuse">
<span class="sig-name descname"><span class="pre">fuse</span></span><span class="sig-paren">(</span><em class="sig-param"><span class="o"><span class="pre">*</span></span><span class="n"><span class="pre">loops</span></span><span class="p"><span class="pre">:</span></span> <span class="n"><span class="pre">List</span><span class="p"><span class="pre">[</span></span><span class="pre">tvm.tir.schedule.schedule.LoopRV</span><span class="p"><span class="pre">]</span></span></span></em>, <em [...]
diff --git a/docs/reference/api/typedoc/classes/bytestreamreader.html b/docs/reference/api/typedoc/classes/bytestreamreader.html
index 72e73e95b2..1c39f9419d 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/2c052b206/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -141,7 +141,7 @@
<div class="tsd-signature tsd-kind-icon">bytes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">Uint8Array</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L45">rpc_server.ts:45</a></li>
</ul>
</aside>
</section>
@@ -151,7 +151,7 @@
<div class="tsd-signature tsd-kind-icon">offset<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span><span class="tsd-signature-symbol"> = 0</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/rpc_server.ts#L44">rpc_server.ts:44</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L44">rpc_server.ts:44</a></li>
</ul>
</aside>
</section>
@@ -168,7 +168,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/rpc_server.ts#L65">rpc_server.ts:65</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L65">rpc_server.ts:65</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">Uint8Array</span></h4>
@@ -185,7 +185,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/rpc_server.ts#L51">rpc_server.ts:51</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L51">rpc_server.ts:51</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -202,7 +202,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/rpc_server.ts#L59">rpc_server.ts:59</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/rpc_server.ts#L59">rpc_server.ts:59</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
diff --git a/docs/reference/api/typedoc/classes/cachedcallstack.html b/docs/reference/api/typedoc/classes/cachedcallstack.html
index 1c1187bba7..182e0f1de5 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/2c052b206/web/src/memory.ts#L223">memory.ts:223</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L208">memory.ts:208</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L312">memory.ts:312</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L284">memory.ts:284</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L388">memory.ts:388</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L376">memory.ts:376</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L267">memory.ts:267</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L243">memory.ts:243</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L321">memory.ts:321</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L252">memory.ts:252</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L359">memory.ts:359</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L342">memory.ts:342</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L350">memory.ts:350</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L326">memory.ts:326</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L363">memory.ts:363</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L346">memory.ts:346</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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/2c052b206/web/src/memory.ts#L334">memory.ts:334</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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 5553d9cef6..17ece251e6 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/2c052b206/web/src/runtime.ts#L359">runtime.ts:359</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L359">runtime.ts:359</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -147,7 +147,7 @@
<div class="tsd-signature tsd-kind-icon">bits<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L357">runtime.ts:357</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L357">runtime.ts:357</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -162,7 +162,7 @@
<div class="tsd-signature tsd-kind-icon">code<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L355">runtime.ts:355</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L355">runtime.ts:355</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -177,7 +177,7 @@
<div class="tsd-signature tsd-kind-icon">lanes<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L359">runtime.ts:359</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L359">runtime.ts:359</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -199,7 +199,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L376">runtime.ts:376</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L376">runtime.ts:376</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">number</span></h4>
@@ -216,7 +216,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L367">runtime.ts:367</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L367">runtime.ts:367</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/dldevice.html b/docs/reference/api/typedoc/classes/dldevice.html
index 388e07542a..7fffb98fef 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/2c052b206/web/src/runtime.ts#L299">runtime.ts:299</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L299">runtime.ts:299</a></li>
</ul>
</aside>
<h4 class="tsd-parameters-title">Parameters</h4>
@@ -146,7 +146,7 @@
<div class="tsd-signature tsd-kind-icon">device<wbr>Id<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L297">runtime.ts:297</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L297">runtime.ts:297</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -161,7 +161,7 @@
<div class="tsd-signature tsd-kind-icon">device<wbr>Type<span class="tsd-signature-symbol">:</span> <span class="tsd-signature-type">number</span></div>
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L295">runtime.ts:295</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L295">runtime.ts:295</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -183,7 +183,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L320">runtime.ts:320</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L320">runtime.ts:320</a></li>
</ul>
</aside>
<div class="tsd-comment tsd-typography">
@@ -205,7 +205,7 @@
<li class="tsd-description">
<aside class="tsd-sources">
<ul>
- <li>Defined in <a href="https://github.com/apache/tvm/blob/2c052b206/web/src/runtime.ts#L327">runtime.ts:327</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/web/src/runtime.ts#L327">runtime.ts:327</a></li>
</ul>
</aside>
<h4 class="tsd-returns-title">Returns <span class="tsd-signature-type">string</span></h4>
diff --git a/docs/reference/api/typedoc/classes/environment.html b/docs/reference/api/typedoc/classes/environment.html
index ec923f326a..588c112af7 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/2c052b206/web/src/environment.ts#L86">environment.ts:86</a></li>
+ <li>Defined in <a href="https://github.com/apache/tvm/blob/99a5734a9/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>
... 2888 lines suppressed ...