You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@tvm.apache.org by tq...@apache.org on 2023/03/08 11:57:08 UTC
[tvm-site] branch asf-site updated: deploying docs (apache/tvm@902c2e2db70b75c36f9bd8c253707b1e8761cc18)
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 638664a72b deploying docs (apache/tvm@902c2e2db70b75c36f9bd8c253707b1e8761cc18)
638664a72b is described below
commit 638664a72b7943d673b6da83e808d2ab93d18bf4
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Wed Mar 8 11:56:58 2023 +0000
deploying docs (apache/tvm@902c2e2db70b75c36f9bd8c253707b1e8761cc18)
---
docs/_images/sphx_glr_micro_train_001.png | Bin 333957 -> 319128 bytes
docs/_images/sphx_glr_micro_train_thumb.png | Bin 23827 -> 22867 bytes
.../how_to/compile_models/from_darknet.rst.txt | 2 +-
.../how_to/compile_models/from_keras.rst.txt | 2 +-
.../how_to/compile_models/from_mxnet.rst.txt | 2 +-
.../how_to/compile_models/from_oneflow.rst.txt | 2 +-
.../how_to/compile_models/from_pytorch.rst.txt | 2 +-
.../how_to/compile_models/from_tensorflow.rst.txt | 2 +-
.../compile_models/sg_execution_times.rst.txt | 22 +-
.../deploy_models/deploy_model_on_adreno.rst.txt | 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 | 20 +-
.../extend_tvm/bring_your_own_datatypes.rst.txt | 2 +-
.../how_to/extend_tvm/sg_execution_times.rst.txt | 10 +-
.../how_to/extend_tvm/use_pass_instrument.rst.txt | 16 +-
.../optimize_operators/opt_conv_cuda.rst.txt | 2 +-
.../optimize_operators/opt_conv_tensorcore.rst.txt | 2 +-
.../how_to/optimize_operators/opt_gemm.rst.txt | 16 +-
.../optimize_operators/sg_execution_times.rst.txt | 8 +-
.../sg_execution_times.rst.txt | 14 +-
.../tune_conv2d_layer_cuda.rst.txt | 1314 +++++++++++------
.../tune_network_cuda.rst.txt | 4 +-
.../tune_network_x86.rst.txt | 4 +-
.../tune_sparse_x86.rst.txt | 85 +-
.../tune_with_autotvm/sg_execution_times.rst.txt | 6 +-
.../tune_with_autotvm/tune_conv2d_cuda.rst.txt | 196 ++-
.../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 | 16 +-
.../work_with_relay/sg_execution_times.rst.txt | 8 +-
.../how_to/work_with_schedules/intrin_math.rst.txt | 2 +-
.../work_with_schedules/sg_execution_times.rst.txt | 16 +-
.../tutorials/autotvm/sg_execution_times.rst.txt | 6 +-
.../frontend/deploy_classification.rst.txt | 2 +-
.../tutorials/frontend/deploy_detection.rst.txt | 2 +-
.../tutorials/frontend/sg_execution_times.rst.txt | 6 +-
.../tutorials/optimize/sg_execution_times.rst.txt | 6 +-
.../topic/vta/tutorials/sg_execution_times.rst.txt | 6 +-
.../tutorial/auto_scheduler_matmul_x86.rst.txt | 6 +-
docs/_sources/tutorial/autotvm_matmul_x86.rst.txt | 20 +-
docs/_sources/tutorial/autotvm_relay_x86.rst.txt | 57 +-
.../tutorial/cross_compilation_and_rpc.rst.txt | 2 +-
docs/_sources/tutorial/intro_topi.rst.txt | 2 +-
docs/_sources/tutorial/sg_execution_times.rst.txt | 18 +-
.../tutorial/tensor_expr_get_started.rst.txt | 45 +-
docs/commit_hash | 2 +-
docs/how_to/compile_models/from_darknet.html | 2 +-
docs/how_to/compile_models/from_keras.html | 2 +-
docs/how_to/compile_models/from_mxnet.html | 2 +-
docs/how_to/compile_models/from_oneflow.html | 11 +-
docs/how_to/compile_models/from_pytorch.html | 10 +-
docs/how_to/compile_models/from_tensorflow.html | 2 +-
docs/how_to/compile_models/sg_execution_times.html | 26 +-
.../deploy_models/deploy_model_on_adreno.html | 2 +-
.../deploy_models/deploy_model_on_android.html | 2 +-
.../deploy_object_detection_pytorch.html | 37 +-
docs/how_to/deploy_models/deploy_prequantized.html | 9 +-
.../deploy_models/deploy_prequantized_tflite.html | 4 +-
docs/how_to/deploy_models/deploy_quantized.html | 2 +-
docs/how_to/deploy_models/deploy_ssd_gluoncv.html | 37 +-
docs/how_to/deploy_models/sg_execution_times.html | 20 +-
.../extend_tvm/bring_your_own_datatypes.html | 2 +-
docs/how_to/extend_tvm/sg_execution_times.html | 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 | 1314 +++++++++++------
.../tune_with_autoscheduler/tune_network_cuda.html | 4 +-
.../tune_with_autoscheduler/tune_network_x86.html | 4 +-
.../tune_with_autoscheduler/tune_sparse_x86.html | 85 +-
.../tune_with_autotvm/sg_execution_times.html | 6 +-
.../how_to/tune_with_autotvm/tune_conv2d_cuda.html | 196 ++-
docs/how_to/work_with_microtvm/micro_autotune.html | 18 +-
docs/how_to/work_with_microtvm/micro_pytorch.html | 4 +-
docs/how_to/work_with_microtvm/micro_train.html | 16 +-
.../work_with_microtvm/sg_execution_times.html | 16 +-
.../how_to/work_with_relay/sg_execution_times.html | 8 +-
docs/how_to/work_with_schedules/intrin_math.html | 2 +-
.../work_with_schedules/sg_execution_times.html | 20 +-
.../api/doxygen/block__scope_8h_source.html | 2 +-
docs/reference/api/doxygen/bound_8h_source.html | 2 +-
.../classtvm_1_1PrimExpr__inherit__graph.svg | 21 +-
.../classtvm_1_1tir_1_1SizeVar-members.html | 15 +-
.../api/doxygen/classtvm_1_1tir_1_1SizeVar.html | 63 +-
.../classtvm_1_1tir_1_1SizeVar__coll__graph.svg | 157 +-
.../classtvm_1_1tir_1_1SizeVar__inherit__graph.svg | 129 +-
.../api/doxygen/classtvm_1_1tir_1_1Var.html | 2 +-
.../classtvm_1_1tir_1_1Var__inherit__graph.svg | 129 +-
.../api/doxygen/cuda_2dense_8h_source.html | 2 +-
.../api/doxygen/cuda_2injective_8h_source.html | 2 +-
.../api/doxygen/cuda_2pooling_8h_source.html | 2 +-
.../api/doxygen/cuda_2reduction_8h_source.html | 2 +-
.../api/doxygen/cuda_2softmax_8h_source.html | 2 +-
.../api/doxygen/data__layout_8h_source.html | 2 +-
.../doxygen/data__type__rewriter_8h_source.html | 2 +-
docs/reference/api/doxygen/functions_func_s.html | 12 +-
docs/reference/api/doxygen/functions_s.html | 8 +-
docs/reference/api/doxygen/ir_2type_8h.html | 2 +-
docs/reference/api/doxygen/map_8h_source.html | 2 +-
docs/reference/api/doxygen/namespacemembers_b.html | 2 +-
docs/reference/api/doxygen/namespacemembers_f.html | 48 +-
.../api/doxygen/namespacemembers_func_b.html | 2 +-
.../api/doxygen/namespacemembers_func_f.html | 48 +-
.../api/doxygen/namespacemembers_func_h.html | 2 +-
.../api/doxygen/namespacemembers_func_i.html | 48 +-
.../api/doxygen/namespacemembers_func_u.html | 48 +-
.../api/doxygen/namespacemembers_func_v.html | 2 +-
docs/reference/api/doxygen/namespacemembers_h.html | 2 +-
docs/reference/api/doxygen/namespacemembers_i.html | 48 +-
docs/reference/api/doxygen/namespacemembers_u.html | 48 +-
docs/reference/api/doxygen/namespacemembers_v.html | 2 +-
...mespacetvm_1_1script_1_1ir__builder_1_1tir.html | 1507 +++++++++++++++-----
.../api/doxygen/nn_2softmax_8h_source.html | 2 +-
.../reference/api/doxygen/operation_8h_source.html | 2 +-
docs/reference/api/doxygen/search/all_14.js | 2 +-
docs/reference/api/doxygen/search/all_16.js | 48 +-
docs/reference/api/doxygen/search/all_17.js | 2 +-
docs/reference/api/doxygen/search/all_3.js | 2 +-
docs/reference/api/doxygen/search/all_7.js | 48 +-
docs/reference/api/doxygen/search/all_9.js | 2 +-
docs/reference/api/doxygen/search/all_a.js | 48 +-
docs/reference/api/doxygen/search/functions_13.js | 2 +-
docs/reference/api/doxygen/search/functions_15.js | 48 +-
docs/reference/api/doxygen/search/functions_16.js | 2 +-
docs/reference/api/doxygen/search/functions_2.js | 2 +-
docs/reference/api/doxygen/search/functions_6.js | 48 +-
docs/reference/api/doxygen/search/functions_8.js | 2 +-
docs/reference/api/doxygen/search/functions_9.js | 48 +-
docs/reference/api/doxygen/stmt_8h_source.html | 4 +-
.../api/doxygen/te_2schedule_8h_source.html | 6 +-
.../reference/api/doxygen/tir_2expr_8h_source.html | 2 +-
.../api/doxygen/tir_2frame_8h_source.html | 2 +-
docs/reference/api/doxygen/tir_2ir_8h.html | 311 ++--
docs/reference/api/doxygen/tir_2ir_8h_source.html | 20 +-
.../doxygen/tir_2op__attr__types_8h_source.html | 2 +-
.../api/doxygen/transform__step_8h_source.html | 2 +-
docs/reference/api/doxygen/var_8h_source.html | 46 +-
docs/reference/api/doxygen/x86_2bnn_8h_source.html | 2 +-
docs/reference/api/python/auto_scheduler.html | 4 +-
.../api/typedoc/classes/bytestreamreader.html | 12 +-
.../api/typedoc/classes/cachedcallstack.html | 34 +-
docs/reference/api/typedoc/classes/dldatatype.html | 12 +-
docs/reference/api/typedoc/classes/dldevice.html | 10 +-
.../reference/api/typedoc/classes/environment.html | 12 +-
docs/reference/api/typedoc/classes/ffilibrary.html | 20 +-
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 | 6 +-
.../tutorials/frontend/deploy_classification.html | 2 +-
.../vta/tutorials/frontend/deploy_detection.html | 2 +-
.../vta/tutorials/frontend/sg_execution_times.html | 6 +-
.../vta/tutorials/optimize/sg_execution_times.html | 6 +-
docs/topic/vta/tutorials/sg_execution_times.html | 6 +-
docs/tutorial/auto_scheduler_matmul_x86.html | 4 +-
docs/tutorial/autotvm_matmul_x86.html | 20 +-
docs/tutorial/autotvm_relay_x86.html | 263 ++--
docs/tutorial/cross_compilation_and_rpc.html | 2 +-
docs/tutorial/intro_topi.html | 2 +-
docs/tutorial/sg_execution_times.html | 18 +-
docs/tutorial/tensor_expr_get_started.html | 41 +-
188 files changed, 4897 insertions(+), 3027 deletions(-)
diff --git a/docs/_images/sphx_glr_micro_train_001.png b/docs/_images/sphx_glr_micro_train_001.png
index b0c19c3436..ce2ec0fa7a 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 475ea5ff8a..84ae33a9fa 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 ae53734820..275fbbb5be 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 18.519 seconds)
+ **Total running time of the script:** ( 1 minutes 18.997 seconds)
.. _sphx_glr_download_how_to_compile_models_from_darknet.py:
diff --git a/docs/_sources/how_to/compile_models/from_keras.rst.txt b/docs/_sources/how_to/compile_models/from_keras.rst.txt
index ef1b2f5575..29a2c63cd5 100644
--- a/docs/_sources/how_to/compile_models/from_keras.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_keras.rst.txt
@@ -232,7 +232,7 @@ Look up prediction top 1 index in 1000 class synset.
.. code-block:: none
Relay top-1 id: 285, class name: Egyptian cat
-
1/1 [==============================] - ETA: 0s
1/1 [==============================] - 1s 947ms/step
+
1/1 [==============================] - ETA: 0s
1/1 [==============================] - 1s 986ms/step
Keras top-1 id: 285, class name: Egyptian cat
diff --git a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
index 6477530f1b..6701a051ae 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.zipb927c167-6dd7-4bda-82fb-8245299da4bf from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+ Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.ziped3d270c-0a1e-4c3b-89bf-3b363ea686b2 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 fb4feb89cd..73f4e09883 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, 37.2MB/s]
39%|###8 | 16.0M/41.5M [00:00<00:00, 54.8MB/s]
58%|#####7 | 24.0M/41.5M [00:00<00:00, 50.7MB/s]
77%|#######7 | 32.0M/41.5M [00:00<00:00, 59.7MB/s]
96%|#########6| 40.0M/41.5M [00:00<00:00, 65.0MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 59.9MB/s]
+
0%| | 0.00/41.5M [00:00<?, ?B/s]
22%|##2 | 9.19M/41.5M [00:00<00:00, 96.3MB/s]
44%|####4 | 18.4M/41.5M [00:00<00:00, 76.6MB/s]
69%|######9 | 28.6M/41.5M [00:00<00:00, 89.0MB/s]
90%|######### | 37.4M/41.5M [00:00<00:00, 82.8MB/s]
100%|##########| 41.5M/41.5M [00:00<00:00, 78.5MB/s]
diff --git a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
index c51beea15a..0cce0c0d3f 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, 55.7MB/s]
36%|###5 | 16.0M/44.7M [00:00<00:00, 65.5MB/s]
60%|#####9 | 26.6M/44.7M [00:00<00:00, 77.1MB/s]
90%|########9 | 40.0M/44.7M [00:00<00:00, 95.6MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 86.6MB/s]
+
0%| | 0.00/44.7M [00:00<?, ?B/s]
23%|##2 | 10.1M/44.7M [00:00<00:00, 104MB/s]
45%|####4 | 20.0M/44.7M [00:00<00:00, 92.7MB/s]
67%|######6 | 29.8M/44.7M [00:00<00:00, 97.2MB/s]
90%|########9 | 40.0M/44.7M [00:00<00:00, 91.9MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 80.3MB/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 f4b86dd4ae..294671e4db 100644
--- a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
@@ -424,7 +424,7 @@ Run the corresponding model on tensorflow
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 22.917 seconds)
+ **Total running time of the script:** ( 1 minutes 24.787 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 e7939f9cbd..9e1b78cd0c 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
=================
-**06:34.510** total execution time for **how_to_compile_models** files:
+**06:40.909** total execution time for **how_to_compile_models** files:
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:22.917 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``) | 01:24.787 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:18.519 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``) | 01:18.997 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:54.985 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``) | 00:55.752 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:37.433 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_oneflow.py` (``from_oneflow.py``) | 00:38.385 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:31.631 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``) | 00:32.203 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:31.128 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``) | 00:31.879 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:27.247 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:27.035 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``) | 00:25.448 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``) | 00:26.668 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:22.479 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``) | 00:22.489 | 0.0 MB |
+-----------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.724 | 0.0 MB |
+| :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``) | 00:02.715 | 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 4dec228aae..6e2bfb9eb2 100644
--- a/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_model_on_adreno.rst.txt
@@ -727,7 +727,7 @@ well as provides information about the model's performance
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 2678.3957 2678.5500 2682.2350 2676.1978 1.5449
+ 2682.8694 2682.6587 2688.6528 2677.6245 2.7009
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 9e6265bb6f..beb0179575 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)
- 15.5596 15.5324 15.6770 15.4915 0.0597
+ 15.8619 15.6985 16.6804 15.4592 0.3908
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 a63705014a..46c24812d3 100644
--- a/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_object_detection_pytorch.rst.txt
@@ -130,7 +130,7 @@ Load pre-trained maskrcnn from torchvision and do tracing
/venv/apache-tvm-py3.7/lib/python3.7/site-packages/torchvision/models/_utils.py:223: UserWarning: Arguments other than a weight enum or `None` for 'weights' are deprecated since 0.13 and will be removed in 0.15. The current behavior is equivalent to passing `weights=MaskRCNN_ResNet50_FPN_Weights.COCO_V1`. You can also use `weights=MaskRCNN_ResNet50_FPN_Weights.DEFAULT` to get the most up-to-date weights.
warnings.warn(msg)
Downloading: "https://download.pytorch.org/models/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth" to /workspace/.cache/torch/hub/checkpoints/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth
-
0%| | 0.00/170M [00:00<?, ?B/s]
5%|4 | 7.99M/170M [00:00<00:02, 73.6MB/s]
9%|9 | 16.0M/170M [00:00<00:02, 72.3MB/s]
15%|#5 | 26.1M/170M [00:00<00:01, 80.3MB/s]
20%|#9 | 33.7M/170M [00:00<00:01, 72.4MB/s]
24%|##3 | 40.7M/170M [00:00<00:02, 67.6MB/s]
28%|##7 | 47.1M/170M [00:00<00:02, 63.7MB/s]
33%|###2 | 56.0M/170M [00:00<00:01, 61.2MB/s]
39%|###9 | 66.8M/170M [00:00<00:01, 74.6MB/s]
44%|####3 | 74.2M/170M [00:01<00:01, 73.1MB/s]
51%|##### | 86.3M/170M [00:01<00:01, 82.4MB/s]
56%|#####5 | 94.3M/170M [00:01<00:01, 72.7MB/s]
61%|######1 | 104M/170M [00:01<00:00, 71.6MB/s]
66%|######5 | 112M/170M [00:01<00:00, 74.6MB/s]
75%|#######5 | 128M/170M [00:01<00:00, 97.7MB/s]
81%|######## | 137M/170M [00:01<00:00, 80.9MB/s]
87%|########6 | 148M/170M [00:02<00:00, 86.9MB/s]
92%|#########2| 157M/170M [00:02<00:00, 87.0MB/s]
97%|#########7| 165M/170M [00:02<00:00, 69.6MB/s]
100%|##########| 170M/170M [00:02<00:00, 74.9MB/s]
+
0%| | 0.00/170M [00:00<?, ?B/s]
5%|4 | 7.99M/170M [00:00<00:02, 74.9MB/s]
12%|#1 | 19.5M/170M [00:00<00:01, 101MB/s]
17%|#7 | 29.7M/170M [00:00<00:01, 103MB/s]
24%|##3 | 40.0M/170M [00:00<00:01, 89.2MB/s]
32%|###1 | 54.1M/170M [00:00<00:01, 108MB/s]
38%|###8 | 64.8M/170M [00:00<00:01, 90.2MB/s]
44%|####4 | 75.0M/170M [00:00<00:01, 94.7MB/s]
52%|#####1 | 88.0M/170M [00:00<00:00, 94.6MB/s]
57%|#####7 | 97.4M/170M [00:01<00:00, 94.5MB/s]
63%|######3 | 107M/170M [00:01<00:00, 97.0MB/s]
69%|######8 | 117M/170M [00:01<00:00, 90.8MB/s]
74%|#######3 | 126M/170M [00:01<00:00, 86.9MB/s]
79%|#######9 | 135M/170M [00:01<00:00, 89.2MB/s]
85%|########4 | 144M/170M [00:01<00:00, 82.1MB/s]
94%|#########4| 160M/170M [00:01<00:00, 104MB/s]
100%|##########| 170M/170M [00:01<00:00, 94.5MB/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 31.461 seconds)
+ **Total running time of the script:** ( 3 minutes 34.920 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 f7a6832d0a..5655a34bc6 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, 72.8MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 90.3MB/s]
+
0%| | 0.00/13.6M [00:00<?, ?B/s]
47%|####6 | 6.30M/13.6M [00:00<00:00, 62.5MB/s]
90%|######### | 12.3M/13.6M [00:00<00:00, 52.6MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 58.4MB/s]
@@ -409,7 +409,7 @@ Here we give an example of how to measure performance of TVM compiled models.
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 90.2658 89.9287 95.1984 89.7850 1.0279
+ 90.2995 90.2016 93.5701 90.0290 0.3953
@@ -458,7 +458,7 @@ TODO
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 15.930 seconds)
+ **Total running time of the script:** ( 1 minutes 16.851 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 d2e84ff8f8..a7d4da2062 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)
- 116.2722 115.9485 121.2760 115.1441 1.0439
+ 120.1072 120.0018 125.6702 119.4399 0.6456
@@ -460,7 +460,7 @@ Here we give an example of how to measure performance of TVM compiled models.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 2 minutes 31.715 seconds)
+ **Total running time of the script:** ( 2 minutes 30.309 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 fe39927581..6093405493 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 31.874 seconds)
+ **Total running time of the script:** ( 1 minutes 32.348 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 d6aa34577a..8f2f8694ea 100644
--- a/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_ssd_gluoncv.rst.txt
@@ -170,7 +170,7 @@ Convert and compile model for CPU.
data: None
input_sym_arg_type = in_param.infer_type()[0]
Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
-
0%| | 0/132723 [00:00<?, ?KB/s]
4%|4 | 5801/132723 [00:00<00:02, 58003.56KB/s]
10%|# | 13817/132723 [00:00<00:01, 71034.81KB/s]
17%|#6 | 21921/132723 [00:00<00:01, 75601.70KB/s]
23%|##2 | 30088/132723 [00:00<00:01, 77994.48KB/s]
29%|##8 | 37962/132723 [00:00<00:01, 78260.17KB/s]
35%|###4 | 46148/132723 [00:00<00:01, 79482.79KB/s]
41%|#### | 54327/132723 [00:00<00:00, 80229.93KB/s]
47%|####6 | 62371/132723 [00:00<00:00, 80295.91KB/s]
53%|#####3 | 70535/132723 [00:00<00:00, 80712.90KB/s]
59%|#####9 | 78621/132723 [00:01<00:00, 80757.26KB/s]
65%|######5 | 86833/132723 [00:01<00:00, 81171.07KB/s]
72%|#######1 | 95037/132723 [00:01<00:00, 81434.23KB/s]
78%|#######7 | 103253/132723 [00:01<00:00, 81650.68KB/s]
84%|########3 | 111419/132723 [00:01<00:00, 81624.78KB/s]
90%|######### | 119582/132723 [00:01<00:00, 80186.97KB/s]
96%|########
#6| 127799/132723 [00:01<00:00, 80774.25KB/s]
100%|##########| 132723/132723 [00:01<00:00, 79648.48KB/s]
+
0%| | 0/132723 [00:00<?, ?KB/s]
4%|4 | 5951/132723 [00:00<00:02, 59495.73KB/s]
11%|# | 13941/132723 [00:00<00:01, 71489.98KB/s]
17%|#6 | 21976/132723 [00:00<00:01, 75532.45KB/s]
23%|##2 | 30017/132723 [00:00<00:01, 77447.42KB/s]
28%|##8 | 37762/132723 [00:00<00:01, 56359.24KB/s]
35%|###4 | 45845/132723 [00:00<00:01, 63029.02KB/s]
40%|###9 | 52766/132723 [00:00<00:01, 58941.31KB/s]
46%|####5 | 60748/132723 [00:00<00:01, 64516.85KB/s]
51%|#####1 | 68178/132723 [00:01<00:00, 67216.50KB/s]
57%|#####7 | 76228/132723 [00:01<00:00, 70973.02KB/s]
63%|######3 | 84255/132723 [00:01<00:00, 73647.00KB/s]
69%|######9 | 92177/132723 [00:01<00:00, 75267.31KB/s]
75%|#######5 | 100071/132723 [00:01<00:00, 76343.19KB/s]
81%|########1 | 107807/132723 [00:01<00:00, 75478.92KB/s]
87%|########7 | 115633/132723 [00:01<00:00, 76293.25KB/s]
93%|########
#3| 123606/132723 [00:01<00:00, 77303.12KB/s]
99%|#########9| 131637/132723 [00:01<00:00, 78193.00KB/s]
100%|##########| 132723/132723 [00:01<00:00, 71404.26KB/s]
@@ -246,7 +246,7 @@ Display result
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 3 minutes 46.232 seconds)
+ **Total running time of the script:** ( 3 minutes 54.095 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 8e28abadeb..170b9b7b24 100644
--- a/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/deploy_models/sg_execution_times.rst.txt
@@ -5,26 +5,26 @@
Computation times
=================
-**15:10.721** total execution time for **how_to_deploy_models** files:
+**15:23.240** total execution time for **how_to_deploy_models** files:
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:46.232 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``) | 03:54.095 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:31.461 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``) | 03:34.920 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:31.715 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``) | 02:30.309 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:31.874 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``) | 01:32.348 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:15.930 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``) | 01:16.851 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 00:56.162 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_adreno.py` (``deploy_model_on_adreno.py``) | 00:56.338 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:42.182 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``) | 00:42.547 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:27.740 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_nano.py` (``deploy_model_on_nano.py``) | 00:28.089 | 0.0 MB |
+------------------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:27.419 | 0.0 MB |
+| :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``) | 00:27.738 | 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 2907449ddd..4e3f4a6cc9 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.zip176d3132-d7ac-4e94-8d7f-8b6c12b275a6 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+ Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip4ed6713c-af6a-4491-98d8-b627d6fa4630 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 2ba20c8cd7..283db86b20 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
=================
-**00:54.557** total execution time for **how_to_extend_tvm** files:
+**00:54.631** 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``) | 00:50.593 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``) | 00:50.749 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.842 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``) | 00:02.784 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.112 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``) | 00:01.091 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``) | 00:00.009 | 0.0 MB |
+| :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``) | 00:00.008 | 0.0 MB |
+-------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt b/docs/_sources/how_to/extend_tvm/use_pass_instrument.rst.txt
index feb9e4295d..2b3bc2e1b5 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: 22538us [22538us] (48.58%; 48.58%)
- FoldScaleAxis: 23853us [8us] (51.42%; 51.42%)
- FoldConstant: 23845us [1752us] (51.40%; 99.97%)
- InferType: 22093us [22093us] (47.62%; 92.65%)
+ InferType: 22432us [22432us] (48.66%; 48.66%)
+ FoldScaleAxis: 23666us [9us] (51.34%; 51.34%)
+ FoldConstant: 23657us [1804us] (51.32%; 99.96%)
+ InferType: 21853us [21853us] (47.41%; 92.38%)
@@ -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: 21907us [21907us] (48.39%; 48.39%)
- FoldScaleAxis: 23369us [6us] (51.61%; 51.61%)
- FoldConstant: 23363us [1766us] (51.60%; 99.97%)
- InferType: 21597us [21597us] (47.70%; 92.44%)
+ InferType: 21736us [21736us] (47.66%; 47.66%)
+ FoldScaleAxis: 23874us [5us] (52.34%; 52.34%)
+ FoldConstant: 23869us [1750us] (52.33%; 99.98%)
+ InferType: 22119us [22119us] (48.50%; 92.67%)
diff --git a/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt b/docs/_sources/how_to/optimize_operators/opt_conv_cuda.rst.txt
index bea62650b7..bcaa92d2d5 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: 34.640190 ms
+ Convolution: 54.244159 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 e9f5039fa9..bc38a4f514 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: 13.358934 ms
+ conv2d with tensor core: 6.854656 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 854687c8e7..50c9436667 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: 0.019040
- Baseline: 3.425819
+ Numpy running time: 0.019124
+ Baseline: 3.231377
@@ -227,7 +227,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
.. code-block:: none
- Opt1: 0.298916
+ Opt1: 0.308924
@@ -318,7 +318,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
.. code-block:: none
- Opt2: 0.336510
+ Opt2: 0.332410
@@ -406,7 +406,7 @@ the access pattern for A matrix is more cache friendly.
.. code-block:: none
- Opt3: 0.116727
+ Opt3: 0.116921
@@ -523,7 +523,7 @@ flattening.
.. code-block:: none
- Opt4: 0.110358
+ Opt4: 0.110576
@@ -635,7 +635,7 @@ write to C when all the block results are ready.
.. code-block:: none
- Opt5: 0.111632
+ Opt5: 0.112377
@@ -748,7 +748,7 @@ Furthermore, we can also utilize multi-core processors to do the thread-level pa
.. code-block:: none
- Opt6: 0.148690
+ Opt6: 0.148110
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 ec750567bc..b6eb7942c9 100644
--- a/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/sg_execution_times.rst.txt
@@ -5,12 +5,12 @@
Computation times
=================
-**00:35.224** total execution time for **how_to_optimize_operators** files:
+**00:34.601** total execution time for **how_to_optimize_operators** files:
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:32.531 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``) | 00:32.066 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.597 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``) | 00:01.493 | 0.0 MB |
+-----------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.097 | 0.0 MB |
+| :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``) | 00:01.042 | 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 56b30c13b0..9b02b87d77 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
=================
-**10:04.125** total execution time for **how_to_tune_with_autoscheduler** files:
+**09:57.707** 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``) | 06:14.904 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``) | 06:08.709 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:41.717 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``) | 01:42.014 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:07.384 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``) | 01:07.519 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:32.581 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``) | 00:31.804 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:14.054 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``) | 00:14.138 | 0.0 MB |
+----------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:13.486 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``) | 00:13.521 | 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 0a9ab61d2f..2bac3498cb 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
@@ -243,12 +243,12 @@ 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": True, "global_symbol": "main", "tir.noalias": 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([4032], "float32", "shared")
- kernel_shared = T.allocate([6144], "float32", "shared")
- threadIdx_x = T.launch_thread("threadIdx.x", 112)
- conv2d_nchw_1 = T.Buffer((1,), data=conv2d_nchw, scope="local", align=4)
+ pad_temp_shared = T.allocate([72], "float32", "shared")
+ kernel_shared = T.allocate([3072], "float32", "shared")
+ threadIdx_x = T.launch_thread("threadIdx.x", 64)
+ conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
conv2d_nchw_1[2] = T.float32(0)
@@ -263,257 +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, rx_outer_outer in T.grid(8, 3):
- cse_var_2: T.int32 = rc_outer_outer * 3136
- cse_var_1: T.int32 = rc_outer_outer * 576
+ 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((4032,), 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(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 112] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 112) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 224) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 336) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 448) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 560] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 560) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 672] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 672) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], 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 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 784) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 896] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 896) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1008] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 776], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1120] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1120) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1232] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1232) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1344] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1344) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1456] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1456) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1568] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1568) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1680] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1680) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1792] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1792) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1904] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1904) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2016] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 1560], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2128] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2128) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2240] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2240) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2352] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2352) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2464] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2464) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2576] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2576) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2688] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2688) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2800] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2800) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2912] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2912) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3024] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 2344], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3136] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3136) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3248] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3248) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3360] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3360) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3472] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3472) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3584] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3584) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3696] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3696) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3808] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3808) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3920] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3920) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- threadIdx_x_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.Buffer((6144,), 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 + cse_var_1 + threadIdx_x_2 * 3 + 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) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + 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 + 224) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 336) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + 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 + 448) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 560) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + 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 + 672) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 784) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 896) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1008) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1120] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1120) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1232] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1232) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 32256]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1456] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1456) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1568] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1568) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1680] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1680) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1792] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1792) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1904] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1904) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2016] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2016) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2128] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2128) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2240] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2240) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2352] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2352) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2464] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2464) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2576] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2576) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2688] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2800] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2800) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2912] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2912) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3024] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3024) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3136] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3136) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3248] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3248) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3360] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3360) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3472] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3472) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3584] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3584) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3696] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3696) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3808] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3808) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3920] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3920) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4032] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96768]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4144] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4144) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4256] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4256) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4368] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4368) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4480] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4480) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4592] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4592) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4704] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4704) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4816] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4816) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4928] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4928) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5040] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5040) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5152] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5152) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5264] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5264) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5376] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5488] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5488) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5600] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5600) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5712] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5712) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5824] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5824) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5936] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5936) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 96):
- kernel_shared_1[threadIdx_x_2 + 6048] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 6048) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 288]
- for rc_outer_inner in range(64):
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 56] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 56] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- compute_1 = T.Buffer((25088,), data=compute.data)
- bias_1 = T.Buffer((512,), data=bias.data)
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7] = T.max(conv2d_nchw_1[0] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 7] = T.max(conv2d_nchw_1[1] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 14] = T.max(conv2d_nchw_1[2] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 21] = T.max(conv2d_nchw_1[3] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 28] = T.max(conv2d_nchw_1[4] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 35] = T.max(conv2d_nchw_1[5] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 42] = T.max(conv2d_nchw_1[6] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 784] = T.max(conv2d_nchw_1[7] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 791] = T.max(conv2d_nchw_1[8] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 798] = T.max(conv2d_nchw_1[9] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 805] = T.max(conv2d_nchw_1[10] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 812] = T.max(conv2d_nchw_1[11] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 819] = T.max(conv2d_nchw_1[12] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 826] = T.max(conv2d_nchw_1[13] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
+ for i1_inner, i3_inner in T.grid(2, 7):
+ compute_1 = T.Buffer((25088,), data=compute.data)
+ bias_1 = T.Buffer((512,), data=bias.data)
+ compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 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))
@@ -563,7 +765,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 0.271 ms
+ Execution time of this operator: 0.354 ms
@@ -612,35 +814,35 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
- conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
- conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=16)
- 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=2)
+ 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=1)
- conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=7)
+ conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
- conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
- conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=7)
+ conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
+ conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=1)
conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
- conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=1)
- conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=64)
- conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=3)
+ conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=2)
+ conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=4)
+ conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=1)
conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=1)
conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=1)
- conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+ conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2 [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
- compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
- compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
- compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=2)
+ compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
+ compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=64)
+ compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
- compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=7)
- compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
- compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=7)
+ compute_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)
compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
s[compute].reorder(compute_i0_o_o_o, compute_i1_o_o_o, compute_i2_o_o_o, compute_i3_o_o_o, compute_i0_o_o_i, compute_i1_o_o_i, compute_i2_o_o_i, compute_i3_o_o_i, compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i, compute_i0_i, compute_i1_i, compute_i2_i, compute_i3_i)
s[conv2d_nchw].compute_at(s[compute], compute_i3_o_i)
@@ -660,14 +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", 64)
+ s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "auto_unroll_max_step", 512)
s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "unroll_explicit", True)
CUDA source code:
@@ -685,10 +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[4032];
- __shared__ float kernel_shared[6144];
+ __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;
@@ -703,163 +905,413 @@ 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 < 8; ++rc_outer_outer) {
- for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
+ for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
+ for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
__syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 112) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 224) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 336) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 448) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 560)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 560) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 672)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 672) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 784)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 784) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 896)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 896) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1008)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 776)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1120)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1120) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1232)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1232) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1344)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1344) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1456)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1456) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1568)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1568) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1680)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1680) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1792)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1792) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1904)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1904) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2016)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 1560)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2128)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2128) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2240)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2240) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2352)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2352) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2464)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2464) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2576)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2576) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2688)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2688) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2800)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2800) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2912)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2912) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3024)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 2344)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3136)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3136) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3248)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3248) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3360)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3360) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3472)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3472) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3584)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3584) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3696)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3696) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3808)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3808) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3920)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3920) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- kernel_shared[((int)threadIdx.x)] = kernel[((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 112)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 112) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 224) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 336) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 448) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 560)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 560) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 672)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 672) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 784) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 896) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1008)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1008) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1120) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1232)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1232) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 32256)];
- kernel_shared[(((int)threadIdx.x) + 1456)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1456) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1568)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1568) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 1680)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1680) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1792) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 1904)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1904) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2016)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2016) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2128)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2128) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2240) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2352)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2352) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 2464)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2464) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2576)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2576) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 2800)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2800) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2912)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2912) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 3024)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3024) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3136)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3136) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 3248)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3248) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3360)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3360) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3472)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3472) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 3584)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3584) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3696)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3696) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 3808)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3808) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3920)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3920) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 4032)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96768)];
- kernel_shared[(((int)threadIdx.x) + 4144)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4144) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4256)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4256) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 4368)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4368) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4480)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4480) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 4592)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4592) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4704)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4704) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4816)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4816) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 4928)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4928) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5040)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5040) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 5152)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5152) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5264)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5264) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 5376)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 129024)];
- kernel_shared[(((int)threadIdx.x) + 5488)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5488) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5600)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5600) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 5712)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5712) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5824)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5824) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 5936)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5936) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- if (((int)threadIdx.x) < 96) {
- kernel_shared[(((int)threadIdx.x) + 6048)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 6048) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 288)];
+ 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);
}
- __syncthreads();
- for (int rc_outer_inner = 0; rc_outer_inner < 64; ++rc_outer_inner) {
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
+ 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) / 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);
}
}
- compute[(((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[0] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 7)] = max((conv2d_nchw[1] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 14)] = max((conv2d_nchw[2] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 21)] = max((conv2d_nchw[3] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 28)] = max((conv2d_nchw[4] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 35)] = max((conv2d_nchw[5] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 42)] = max((conv2d_nchw[6] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 784)] = max((conv2d_nchw[7] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 791)] = max((conv2d_nchw[8] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 798)] = max((conv2d_nchw[9] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 805)] = max((conv2d_nchw[10] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 812)] = max((conv2d_nchw[11] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 819)] = max((conv2d_nchw[12] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 826)] = max((conv2d_nchw[13] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
}
@@ -918,7 +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 14.904 seconds)
+ **Total running time of the script:** ( 6 minutes 8.709 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 b324eb4968..d108e57e97 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_network_cuda.rst.txt
@@ -647,7 +647,7 @@ so we can read the log file and load the best schedules.
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 7.8994 7.9014 7.9033 7.8934 0.0043
+ 7.8820 7.8852 7.8878 7.8730 0.0064
@@ -675,7 +675,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 7.384 seconds)
+ **Total running time of the script:** ( 1 minutes 7.519 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 b5d2fec746..b541529127 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)
- 758.6970 758.7265 761.0138 756.3507 1.9038
+ 762.1932 762.1844 762.2230 762.1721 0.0217
@@ -694,7 +694,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 41.717 seconds)
+ **Total running time of the script:** ( 1 minutes 42.014 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 16cfa2c380..6ff273fa48 100644
--- a/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
+++ b/docs/_sources/how_to/tune_with_autoscheduler/tune_sparse_x86.rst.txt
@@ -389,74 +389,27 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
@T.prim_func
def main(placeholder: T.Buffer((128, 256), "float32"), placeholder_1: T.Buffer((4916, 16, 1), "float32"), placeholder_2: T.Buffer((4916,), "int32"), placeholder_3: T.Buffer((33,), "int32"), placeholder_4: T.Buffer((128, 512), "float32"), compute: T.Buffer((128, 512), "float32")):
T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
- for i0_outer_i1_outer_fused in T.parallel(32):
- compute_1 = T.allocate([2048], "float32", "global")
- compute_2 = T.Buffer((2048,), data=compute_1)
- for nb_j_inner in range(2):
- for i_inner_init in range(64):
- cse_var_1: T.int32 = i_inner_init * 32 + nb_j_inner * 16
- compute_2[cse_var_1] = T.float32(0)
- compute_2[cse_var_1 + 1] = T.float32(0)
- compute_2[cse_var_1 + 2] = T.float32(0)
- compute_2[cse_var_1 + 3] = T.float32(0)
- compute_2[cse_var_1 + 4] = T.float32(0)
- compute_2[cse_var_1 + 5] = T.float32(0)
- compute_2[cse_var_1 + 6] = T.float32(0)
- compute_2[cse_var_1 + 7] = T.float32(0)
- compute_2[cse_var_1 + 8] = T.float32(0)
- compute_2[cse_var_1 + 9] = T.float32(0)
- compute_2[cse_var_1 + 10] = T.float32(0)
- compute_2[cse_var_1 + 11] = T.float32(0)
- compute_2[cse_var_1 + 12] = T.float32(0)
- compute_2[cse_var_1 + 13] = T.float32(0)
- compute_2[cse_var_1 + 14] = T.float32(0)
- compute_2[cse_var_1 + 15] = T.float32(0)
- for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner}), 64):
- cse_var_2 = T.int32()
+ for i0_outer_i1_outer_fused in T.parallel(128):
+ compute_1 = T.allocate([512], "float32", "global")
+ compute_2 = T.Buffer((512,), data=compute_1)
+ for i_outer_inner in range(64):
+ for i_inner_init, j_init in T.grid(2, 4):
+ compute_2[i_outer_inner * 8 + i_inner_init * 4 + 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: i0_outer_i1_outer_fused // 4}), 2, 4):
+ cse_var_1 = T.int32()
placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
- cse_var_21: T.int32 = elem_idx * 16
- cse_var_20: T.int32 = i_inner * 32 + nb_j_inner * 16
- cse_var_19: T.int32 = i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner
- cse_var_18: T.int32 = i0_outer_i1_outer_fused // 16 * 16384 + i_inner * 256
- cse_var_17: T.int32 = cse_var_20 + 9
- cse_var_16: T.int32 = cse_var_20 + 8
- cse_var_15: T.int32 = cse_var_20 + 7
- cse_var_14: T.int32 = cse_var_20 + 6
- cse_var_13: T.int32 = cse_var_20 + 5
- cse_var_12: T.int32 = cse_var_20 + 4
- cse_var_11: T.int32 = cse_var_20 + 3
- cse_var_10: T.int32 = cse_var_20 + 2
- cse_var_9: T.int32 = cse_var_20 + 15
- cse_var_8: T.int32 = cse_var_20 + 14
- cse_var_7: T.int32 = cse_var_20 + 13
- cse_var_6: T.int32 = cse_var_20 + 12
- cse_var_5: T.int32 = cse_var_20 + 11
- cse_var_4: T.int32 = cse_var_20 + 10
- cse_var_3: T.int32 = cse_var_20 + 1
- placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
- placeholder_7 = T.Buffer((32768,), data=placeholder.data)
- placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_20] = compute_2[cse_var_20] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(64, 32):
- cse_var_22: T.int32 = i0_outer_i1_outer_fused // 16 * 32768 + i0_inner * 512 + i0_outer_i1_outer_fused % 16 * 32 + i1_inner
+ cse_var_2: T.int32 = i0_outer_i1_outer_fused // 4
+ if T.likely(elem_idx < placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]):
+ 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)
+ cse_var_3: T.int32 = i_outer_inner * 8 + i_inner * 4 + j
+ compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_2] * 16 + elem_idx * 16 + i0_outer_i1_outer_fused % 4 * 4 + j] * T.max(placeholder_7[i_outer_inner * 512 + i_inner * 256 + placeholder_8[placeholder_5[cse_var_2] + elem_idx]], T.float32(0))
+ for i0_inner in range(128):
+ cse_var_4: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 4
compute_3 = T.Buffer((65536,), data=compute.data)
placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
- compute_3[cse_var_22] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_22], T.float32(0))
+ compute_3[cse_var_4:cse_var_4 + 4] = T.max(compute_2[i0_inner * 4:i0_inner * 4 + 4] + placeholder_5[cse_var_4:cse_var_4 + 4], T.Broadcast(T.float32(0), 4))
@@ -506,7 +459,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 1.850 ms
+ Execution time of this operator: 3.125 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 289c808f57..9f1b84aea1 100644
--- a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
@@ -5,12 +5,12 @@
Computation times
=================
-**00:31.133** total execution time for **how_to_tune_with_autotvm** files:
+**00:35.099** 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``) | 00:31.100 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``) | 00:35.060 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``) | 00:00.020 | 0.0 MB |
+| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``) | 00:00.025 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``) | 00:00.005 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
index 6e11aa6b5b..9b2723bb79 100644
--- a/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/tune_conv2d_cuda.rst.txt
@@ -390,8 +390,9 @@ 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, 64, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9074474
- No: 2 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 256, 2, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5959158
+ No: 2 GFLOPS: 12.04/12.04 result: MeasureResult(costs=(0.01922388833333333,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.596773386001587, timestamp=1678274198.0724711) [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3687096
+ No: 3 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -513,8 +514,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 8, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4858385
- No: 3 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 64, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1556414
+ No: 4 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -636,8 +637,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 1, 64]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4084063
- No: 4 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 32, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7305670
+ No: 5 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -759,8 +760,8 @@ 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, 128, 1, 4]), ('tile_y', [-1, 1, 7, 1]), ('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', 1500), ('unroll_explicit', 1)],None,10386307
- No: 5 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4246113
+ No: 6 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -882,8 +883,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 2, 32]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8031292
- No: 6 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 2, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8048670
+ No: 7 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1005,8 +1006,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 64, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5852536
- No: 7 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1348496
+ No: 8 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1128,10 +1129,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 16, 2, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6032634
- No: 8 GFLOPS: 5.46/5.46 result: MeasureResult(costs=(0.04242529675,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.1453189849853516, timestamp=1678271999.5114067) [('tile_f', [-1, 2, 1, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4702857
- No: 9 GFLOPS: 234.60/234.60 result: MeasureResult(costs=(0.0009867967423312883,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5571238994598389, timestamp=1678272001.4171991) [('tile_f', [-1, 4, 16, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4714196
- No: 10 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 8, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4195579
+ No: 9 GFLOPS: 0.00/12.04 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
@@ -1253,8 +1252,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 128, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,677869
- No: 11 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 128, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 128, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8579146
+ No: 10 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1376,9 +1375,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9650809
- No: 12 GFLOPS: 1.28/234.60 result: MeasureResult(costs=(0.18103141775,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.474294662475586, timestamp=1678272004.214991) [('tile_f', [-1, 4, 2, 64]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3329246
- No: 13 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 32, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7210152
+ No: 11 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1500,8 +1498,8 @@ 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, 1, 4]), ('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, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4983106
- No: 14 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 2, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1499693
+ No: 12 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1623,8 +1621,9 @@ 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, 16, 4, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8529993
- No: 15 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 32, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1891602
+ No: 13 GFLOPS: 7.71/12.04 result: MeasureResult(costs=(0.03004289625,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.9099907875061035, timestamp=1678274204.5159023) [('tile_f', [-1, 32, 4, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3581237
+ No: 14 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1746,8 +1745,8 @@ 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, 1, 256]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4212776
- No: 16 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ 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, 7, 1, 1]), ('tile_rc', [-1, 8, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6154661
+ No: 15 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1869,8 +1868,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 64, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8609747
- No: 17 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1175758
+ No: 16 GFLOPS: 0.00/12.04 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 592, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 544, in _build_func_common
@@ -1992,9 +1991,8 @@ for this template
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
- tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 64, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3078403
- No: 18 GFLOPS: 171.84/234.60 result: MeasureResult(costs=(0.0013472277733333335,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2511568069458008, timestamp=1678272005.8511157) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1457959
- No: 19 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2672315
+ No: 17 GFLOPS: 0.00/12.04 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
@@ -2116,8 +2114,9 @@ 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, 8]), ('tile_y', [-1, 1, 1, 7]), ('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,9656180
- No: 20 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 128, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3882829
+ No: 18 GFLOPS: 9.93/12.04 result: MeasureResult(costs=(0.023322064666666666,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7776870727539062, timestamp=1678274206.5398247) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2281400
+ No: 19 GFLOPS: 0.00/12.04 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
@@ -2239,7 +2238,130 @@ 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, 4, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10230560
+ tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 16, 2]), ('tile_y', [-1, 1, 7, 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', 512), ('unroll_explicit', 1)],None,7913488
+ No: 20 GFLOPS: 0.00/12.04 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, 1, 32, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8245030
@@ -2294,9 +2416,9 @@ and measure running time.
Finish loading 20 records
Best config:
- [('tile_f', [-1, 4, 16, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4714196
+ [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3687096
Finish loading 20 records
- Time cost of this operator: 0.001362
+ Time cost of this operator: 0.011204
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 8fdd7a444a..4081f2aab6 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 324.8 98.785 (1, 2, 10, 10, 3) 2 1 [324.8]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.042 0.925 (1, 6, 10, 10) 1 1 [3.042]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.953 0.29 (1, 1, 10, 10, 3) 1 1 [0.953]
- Total_time - 328.795 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 315.0 98.742 (1, 2, 10, 10, 3) 2 1 [315.0]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.057 0.958 (1, 6, 10, 10) 1 1 [3.057]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.958 0.3 (1, 1, 10, 10, 3) 1 1 [0.958]
+ Total_time - 319.014 - - - - -
@@ -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 100.3 97.348 (1, 6, 10, 10, 1) 2 1 [100.3]
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.757 1.705 (1, 6, 10, 10) 1 1 [1.757]
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.976 0.947 (1, 1, 10, 10, 3) 1 1 [0.976]
- Total_time - 103.033 - - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 108.1 97.604 (1, 6, 10, 10, 1) 2 1 [108.1]
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.795 1.621 (1, 6, 10, 10) 1 1 [1.795]
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.858 0.775 (1, 3, 10, 10, 1) 1 1 [0.858]
+ Total_time - 110.754 - - - - -
@@ -439,7 +439,7 @@ Timing the tuned program
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 22.752 seconds)
+ **Total running time of the script:** ( 1 minutes 20.648 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 e484752843..86487ac5cc 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_pytorch.rst.txt
@@ -118,7 +118,7 @@ download a cat image and preprocess it to use as the model input.
/venv/apache-tvm-py3.7/lib/python3.7/site-packages/torch/ao/quantization/utils.py:281: UserWarning: must run observer before calling calculate_qparams. Returning default values.
"must run observer before calling calculate_qparams. " +
Downloading: "https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2_qnnpack_37f702c5.pth
-
0%| | 0.00/3.42M [00:00<?, ?B/s]
100%|##########| 3.42M/3.42M [00:00<00:00, 105MB/s]
+
0%| | 0.00/3.42M [00:00<?, ?B/s]
100%|##########| 3.42M/3.42M [00:00<00:00, 124MB/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 17.022 seconds)
+ **Total running time of the script:** ( 1 minutes 19.092 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 cfd6a93fae..ffe4df9be8 100644
--- a/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt
+++ b/docs/_sources/how_to/work_with_microtvm/micro_train.rst.txt
@@ -218,7 +218,7 @@ take about **2 minutes** to download the Stanford Cars, while COCO 2017 validati
.. code-block:: none
- '/tmp/tmpmkfhz42r/images/random'
+ '/tmp/tmp96e_gj6m/images/random'
@@ -309,7 +309,7 @@ objects to other stuff? We can display some examples from our datasets using ``m
.. image-sg:: /how_to/work_with_microtvm/images/sphx_glr_micro_train_001.png
- :alt: [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0]
+ :alt: [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0]
:srcset: /how_to/work_with_microtvm/images/sphx_glr_micro_train_001.png
:class: sphx-glr-single-img
@@ -318,8 +318,8 @@ objects to other stuff? We can display some examples from our datasets using ``m
.. code-block:: none
- /tmp/tmpmkfhz42r/images/target contains 8144 images
- /tmp/tmpmkfhz42r/images/random contains 5000 images
+ /tmp/tmp96e_gj6m/images/target contains 8144 images
+ /tmp/tmp96e_gj6m/images/random contains 5000 images
@@ -494,13 +494,13 @@ the time on our validation set).
.. code-block:: none
Epoch 1/3
- 328/328 - 48s - loss: 0.2043 - accuracy: 0.9280 - val_loss: 0.1360 - val_accuracy: 0.9562 - 48s/epoch - 145ms/step
+ 328/328 - 48s - loss: 0.2257 - accuracy: 0.9211 - val_loss: 0.1038 - val_accuracy: 0.9622 - 48s/epoch - 148ms/step
Epoch 2/3
- 328/328 - 44s - loss: 0.1009 - accuracy: 0.9624 - val_loss: 0.1234 - val_accuracy: 0.9585 - 44s/epoch - 134ms/step
+ 328/328 - 43s - loss: 0.1021 - accuracy: 0.9618 - val_loss: 0.0971 - val_accuracy: 0.9649 - 43s/epoch - 132ms/step
Epoch 3/3
- 328/328 - 44s - loss: 0.0591 - accuracy: 0.9779 - val_loss: 0.1900 - val_accuracy: 0.9415 - 44s/epoch - 133ms/step
+ 328/328 - 43s - loss: 0.0728 - accuracy: 0.9731 - val_loss: 0.1309 - val_accuracy: 0.9558 - 43s/epoch - 132ms/step
- <keras.callbacks.History object at 0x7f7bc84ed950>
+ <keras.callbacks.History object at 0x7f149e420b90>
@@ -861,7 +861,7 @@ Arduino tutorial for how to do that `on GitHub <https://github.com/guberti/tvm-a
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 4 minutes 24.406 seconds)
+ **Total running time of the script:** ( 4 minutes 26.784 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 7c5febefd9..970da6f943 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
@@ -5,22 +5,22 @@
Computation times
=================
-**07:21.426** total execution time for **how_to_work_with_microtvm** files:
+**07:24.526** total execution time for **how_to_work_with_microtvm** files:
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:24.406 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_train.py` (``micro_train.py``) | 04:26.784 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 01:22.752 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``) | 01:20.648 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:17.022 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_pytorch.py` (``micro_pytorch.py``) | 01:19.092 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:10.087 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_aot.py` (``micro_aot.py``) | 00:10.165 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:07.159 | 0.0 MB |
-+-----------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``) | 00:00.000 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``) | 00:07.838 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``) | 00:00.000 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
+| :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``) | 00:00.000 | 0.0 MB |
++-----------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_microtvm_micro_mlperftiny.py` (``micro_mlperftiny.py``) | 00:00.000 | 0.0 MB |
+-----------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt b/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
index df23ad4128..38ab0c9919 100644
--- a/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/work_with_relay/sg_execution_times.rst.txt
@@ -5,14 +5,14 @@
Computation times
=================
-**00:46.439** total execution time for **how_to_work_with_relay** files:
+**00:50.749** 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``) | 00:34.165 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_pipeline_executor.py` (``using_pipeline_executor.py``) | 00:36.555 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:10.412 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``) | 00:12.092 | 0.0 MB |
+----------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:01.856 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``) | 00:02.095 | 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 43ac9331e3..317e39e48d 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 0x7f7a7598b200>
+ <function my_cuda_math_rule at 0x7f135015d200>
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 4d1267686d..9b5aeb8a21 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
=================
-**00:06.702** total execution time for **how_to_work_with_schedules** files:
+**00:07.645** total execution time for **how_to_work_with_schedules** files:
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:04.170 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``) | 00:04.633 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.135 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``) | 00:01.587 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.589 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``) | 00:00.594 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``) | 00:00.572 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``) | 00:00.589 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``) | 00:00.118 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``) | 00:00.117 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.053 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``) | 00:00.065 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``) | 00:00.038 | 0.0 MB |
+| :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``) | 00:00.033 | 0.0 MB |
+------------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``) | 00:00.027 | 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 1a461894a0..8a5d5c4d71 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
=================
-**00:31.039** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:31.138** total execution time for **topic_vta_tutorials_autotvm** files:
+---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.033 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``) | 00:31.132 | 0.0 MB |
+---------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``) | 00:00.007 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``) | 00:00.006 | 0.0 MB |
+---------------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
index a62924c80c..b2fada776b 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 32.99s!
+ resnet18_v1 inference graph built in 33.31s!
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 23554a1d2d..1c44e5e3d3 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 22.58s!
+ yolov3-tiny inference graph built in 22.66s!
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 2034c75cca..b700e2e20d 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
=================
-**01:39.229** total execution time for **topic_vta_tutorials_frontend** files:
+**01:39.564** total execution time for **topic_vta_tutorials_frontend** files:
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:49.730 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``) | 00:50.228 | 0.0 MB |
+------------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:49.499 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``) | 00:49.336 | 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 054ec1ee56..45c1b0f9ad 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
=================
-**00:03.175** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.113** total execution time for **topic_vta_tutorials_optimize** files:
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.719 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``) | 00:02.652 | 0.0 MB |
+--------------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.456 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``) | 00:00.461 | 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 742b803792..2e19c50b26 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
=================
-**00:00.790** total execution time for **topic_vta_tutorials** files:
+**00:00.771** total execution time for **topic_vta_tutorials** files:
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.409 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``) | 00:00.399 | 0.0 MB |
+---------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.381 | 0.0 MB |
+| :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``) | 00:00.372 | 0.0 MB |
+---------------------------------------------------------------------------------+-----------+--------+
diff --git a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
index c3bbab20a7..04941aa22f 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -211,12 +211,12 @@ trials, we can load the best schedule from the log file and apply it.
.. code-block:: none
-
*E
+
.. GENERATED FROM PYTHON SOURCE LINES 133-139
Inspecting the Optimized Schedule
@@ -325,7 +325,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 98.445 ms
+ Execution time of this operator: 94.709 ms
@@ -441,7 +441,7 @@ operations.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 33.908 seconds)
+ **Total running time of the script:** ( 1 minutes 38.775 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 81041b7777..9a0c107b14 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: 11.22/11.22 result: MeasureResult(costs=(0.0239279706,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.623093843460083, timestamp=1678270415.0959404) [('tile_y', [-1, 256]), ('tile_x', [-1, 512])],None,98
- No: 2 GFLOPS: 0.89/11.22 result: MeasureResult(costs=(0.30232326360000006,), error_no=MeasureErrorNo.NO_ERROR, all_cost=5.066333055496216, timestamp=1678270420.1778522) [('tile_y', [-1, 256]), ('tile_x', [-1, 2])],None,18
- No: 3 GFLOPS: 11.39/11.39 result: MeasureResult(costs=(0.0235573528,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6353390216827393, timestamp=1678270422.0577884) [('tile_y', [-1, 256]), ('tile_x', [-1, 256])],None,88
- No: 4 GFLOPS: 0.95/11.39 result: MeasureResult(costs=(0.28289550119999995,), error_no=MeasureErrorNo.NO_ERROR, all_cost=4.773803472518921, timestamp=1678270428.0979815) [('tile_y', [-1, 32]), ('tile_x', [-1, 2])],None,15
- No: 5 GFLOPS: 11.49/11.49 result: MeasureResult(costs=(0.0233721852,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6218712329864502, timestamp=1678270428.8542662) [('tile_y', [-1, 32]), ('tile_x', [-1, 512])],None,95
- No: 6 GFLOPS: 1.92/11.49 result: MeasureResult(costs=(0.1397680108,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.467949628829956, timestamp=1678270431.342784) [('tile_y', [-1, 4]), ('tile_x', [-1, 1])],None,2
- No: 7 GFLOPS: 11.79/11.79 result: MeasureResult(costs=(0.0227654784,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6303389072418213, timestamp=1678270433.2135043) [('tile_y', [-1, 64]), ('tile_x', [-1, 512])],None,96
- No: 8 GFLOPS: 0.49/11.79 result: MeasureResult(costs=(0.5501215777999999,), error_no=MeasureErrorNo.NO_ERROR, all_cost=9.04657530784607, timestamp=1678270442.2935045) [('tile_y', [-1, 512]), ('tile_x', [-1, 1])],None,9
- No: 9 GFLOPS: 2.90/11.79 result: MeasureResult(costs=(0.0925959972,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.715602159500122, timestamp=1678270444.1215656) [('tile_y', [-1, 8]), ('tile_x', [-1, 4])],None,23
- No: 10 GFLOPS: 2.55/11.79 result: MeasureResult(costs=(0.10513146420000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.914182424545288, timestamp=1678270446.0786948) [('tile_y', [-1, 512]), ('tile_x', [-1, 8])],None,39
+ No: 1 GFLOPS: 14.60/14.60 result: MeasureResult(costs=(0.0183803462,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6169836521148682, timestamp=1678272587.7717292) [('tile_y', [-1, 32]), ('tile_x', [-1, 64])],None,65
+ No: 2 GFLOPS: 7.74/14.60 result: MeasureResult(costs=(0.034695930800000004,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.7929935455322266, timestamp=1678272589.8186598) [('tile_y', [-1, 512]), ('tile_x', [-1, 32])],None,59
+ No: 3 GFLOPS: 11.38/14.60 result: MeasureResult(costs=(0.0235958738,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6478281021118164, timestamp=1678272590.45551) [('tile_y', [-1, 128]), ('tile_x', [-1, 256])],None,87
+ No: 4 GFLOPS: 12.65/14.60 result: MeasureResult(costs=(0.0212239796,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6940491199493408, timestamp=1678272592.3151495) [('tile_y', [-1, 64]), ('tile_x', [-1, 128])],None,76
+ No: 5 GFLOPS: 1.99/14.60 result: MeasureResult(costs=(0.13455430940000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.3972744941711426, timestamp=1678272594.9107897) [('tile_y', [-1, 8]), ('tile_x', [-1, 1])],None,3
+ No: 6 GFLOPS: 10.55/14.60 result: MeasureResult(costs=(0.0254377706,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.7333240509033203, timestamp=1678272596.8148508) [('tile_y', [-1, 1]), ('tile_x', [-1, 128])],None,70
+ No: 7 GFLOPS: 10.49/14.60 result: MeasureResult(costs=(0.0255893816,), error_no=MeasureErrorNo.NO_ERROR, all_cost=0.6714065074920654, timestamp=1678272597.4824355) [('tile_y', [-1, 4]), ('tile_x', [-1, 64])],None,62
+ No: 8 GFLOPS: 0.50/14.60 result: MeasureResult(costs=(0.539256833,), error_no=MeasureErrorNo.NO_ERROR, all_cost=8.892446279525757, timestamp=1678272606.3906024) [('tile_y', [-1, 32]), ('tile_x', [-1, 1])],None,5
+ No: 9 GFLOPS: 0.48/14.60 result: MeasureResult(costs=(0.561035428,), error_no=MeasureErrorNo.NO_ERROR, all_cost=9.184801578521729, timestamp=1678272615.7025707) [('tile_y', [-1, 512]), ('tile_x', [-1, 1])],None,9
+ No: 10 GFLOPS: 2.37/14.60 result: MeasureResult(costs=(0.11314761800000002,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.021230697631836, timestamp=1678272617.7673893) [('tile_y', [-1, 8]), ('tile_x', [-1, 2])],None,13
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index 5d1eb351c5..83f3dda9b3 100644
--- a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
@@ -311,7 +311,7 @@ standard deviation.
.. code-block:: none
- {'mean': 515.0090213199996, 'median': 515.2850962500054, 'std': 2.0018207248826223}
+ {'mean': 512.6563362900002, 'median': 512.4201752500028, 'std': 1.346805710381719}
@@ -545,30 +545,31 @@ the tuning data to.
.. code-block:: none
-
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 1/25] Current/Best: 8.03/ 16.69 GFLOPS | Progress: (4/20) | 14.48 s
[Task 1/25] Current/Best: 18.02/ 18.02 GFLOPS | Progress: (8/20) | 18.61 s
[Task 1/25] Current/Best: 16.02/ 21.80 GFLOPS | Progress: (12/20) | 20.64 s
[Task 1/25] Current/Best: 15.21/ 21.80 GFLOPS | Progress: (16/20) | 24.42 s
[Task 1/25] Current/Best: 14.87/ 21.80 GFLOPS | Progress: (20/20) | 27.02 s Done.
-
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 15.13/ 20.18 GFLOPS | Progress: (4/20) | 4.79 s
[Task 2/25] Current/Best: 16.20/ 20.18 GFLOPS | Progress: (8/20) | 6.45 s
[Task 2/25] Current/Best: 11.68/ 20.18 GFLOPS | Progress: (12/20) | 8.30 s
[Task 2/25] Current/Best: 3.38/ 20.18 GFLOPS | Progress: (16/20) | 9.91 s
[Task 2/25] Current/Best: 11.77/ 20.18 GFLOPS | Progress: (20/20) | 11.67 s Done.
-
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 6.20/ 19.19 GFLOPS | Progress: (4/20) | 5.23 s
[Task 3/25] Current/Best: 9.43/ 22.74 GFLOPS | Progress: (8/20) | 7.49 s
[Task 3/25] Current/Best: 5.69/ 22.74 GFLOPS | Progress: (12/20) | 10.44 s
[Task 3/25] Current/Best: 16.12/ 23.69 GFLOPS | Progress: (16/20) | 12.80 s
[Task 3/25] Current/Best: 14.45/ 23.69 GFLOPS | Progress: (20/20) | 14.85 s Done.
-
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 17.60/ 21.27 GFLOPS | Progress: (4/20) | 10.74 s
[Task 4/25] Current/Best: 12.93/ 21.27 GFLOPS | Progress: (8/20) | 12.25 s
[Task 4/25] Current/Best: 13.71/ 21.27 GFLOPS | Progress: (12/20) | 15.33 s
[Task 4/25] Current/Best: 12.19/ 21.27 GFLOPS | Progress: (16/20) | 17.41 s
[Task 4/25] Current/Best: 9.74/ 21.27 GFLOPS | Progress: (20/20) | 19.43 s Done.
-
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 13.10/ 13.10 GFLOPS | Progress: (4/20) | 4.97 s
[Task 5/25] Current/Best: 14.91/ 15.05 GFLOPS | Progress: (8/20) | 7.66 s
[Task 5/25] Current/Best: 18.45/ 18.45 GFLOPS | Progress: (12/20) | 9.58 s
[Task 5/25] Current/Best: 8.18/ 19.11 GFLOPS | Progress: (16/20) | 11.63 s
[Task 5/25] Current/Best: 10.94/ 21.61 GFLOPS | Progress: (20/20) | 13.48 s Done.
-
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 22.97/ 22.97 GFLOPS | Progress: (4/20) | 5.31 s
[Task 6/25] Current/Best: 14.76/ 23.16 GFLOPS | Progress: (8/20) | 7.66 s
[Task 6/25] Current/Best: 7.63/ 23.16 GFLOPS | Progress: (12/20) | 11.86 s
[Task 6/25] Current/Best: 6.23/ 23.16 GFLOPS | Progress: (16/20) | 14.80 s
[Task 6/25] Current/Best: 5.85/ 23.16 GFLOPS | Progress: (20/20) | 18.30 s Done.
-
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 14.65/ 16.86 GFLOPS | Progress: (4/20) | 5.15 s
[Task 7/25] Current/Best: 13.87/ 16.86 GFLOPS | Progress: (8/20) | 7.40 s
[Task 7/25] Current/Best: 19.39/ 19.39 GFLOPS | Progress: (12/20) | 9.92 s
[Task 7/25] Current/Best: 5.97/ 19.39 GFLOPS | Progress: (16/20) | 13.28 s
[Task 7/25] Current/Best: 19.94/ 19.94 GFLOPS | Progress: (20/20) | 15.69 s Done.
-
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 11.63/ 20.95 GFLOPS | Progress: (4/20) | 7.92 s
[Task 8/25] Current/Best: 11.98/ 20.95 GFLOPS | Progress: (8/20) | 13.24 s
[Task 8/25] Current/Best: 10.75/ 20.95 GFLOPS | Progress: (12/20) | 16.44 s
[Task 8/25] Current/Best: 2.90/ 20.95 GFLOPS | Progress: (16/20) | 20.95 s
[Task 8/25] Current/Best: 13.82/ 20.95 GFLOPS | Progress: (20/20) | 24.24 s Done.
-
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 11.11/ 17.78 GFLOPS | Progress: (4/20) | 5.00 s
[Task 9/25] Current/Best: 16.02/ 17.78 GFLOPS | Progress: (8/20) | 7.01 s
[Task 9/25] Current/Best: 11.27/ 17.78 GFLOPS | Progress: (12/20) | 9.75 s
[Task 9/25] Current/Best: 13.91/ 22.33 GFLOPS | Progress: (16/20) | 12.35 s
[Task 9/25] Current/Best: 7.14/ 22.33 GFLOPS | Progress: (20/20) | 15.61 s Done.
-
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 17.74/ 17.74 GFLOPS | Progress: (4/20) | 4.98 s
[Task 10/25] Current/Best: 3.77/ 18.26 GFLOPS | Progress: (8/20) | 6.96 s
[Task 10/25] Current/Best: 2.92/ 18.26 GFLOPS | Progress: (12/20) | 9.09 s
[Task 10/25] Current/Best: 5.58/ 18.26 GFLOPS | Progress: (16/20) | 11.17 s
[Task 10/25] Current/Best: 8.27/ 18.26 GFLOPS | Progress: (20/20) | 14.71 s Done.
-
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 3.10/ 17.66 GFLOPS | Progress: (4/20) | 6.09 s
[Task 11/25] Current/Best: 13.32/ 17.66 GFLOPS | Progress: (8/20) | 8.33 s
[Task 11/25] Current/Best: 6.18/ 18.25 GFLOPS | Progress: (12/20) | 10.91 s
[Task 11/25] Current/Best: 15.93/ 18.25 GFLOPS | Progress: (16/20) | 13.14 s
[Task 11/25] Current/Best: 18.23/ 19.61 GFLOPS | Progress: (20/20) | 15.71 s Done.
-
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 5.24/ 16.05 GFLOPS | Progress: (4/20) | 5.97 s
[Task 12/25] Current/Best: 19.28/ 19.28 GFLOPS | Progress: (8/20) | 8.19 s
[Task 12/25] Current/Best: 15.50/ 19.28 GFLOPS | Progress: (12/20) | 11.19 s
[Task 12/25] Current/Best: 11.50/ 19.28 GFLOPS | Progress: (16/20) | 14.85 s
[Task 12/25] Current/Best: 12.74/ 19.28 GFLOPS | Progress: (20/20) | 20.93 s Done.
-
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 6.21/ 19.45 GFLOPS | Progress: (4/20) | 6.29 s
[Task 13/25] Current/Best: 17.59/ 20.64 GFLOPS | Progress: (8/20) | 9.70 s
[Task 13/25] Current/Best: 21.04/ 21.04 GFLOPS | Progress: (12/20) | 14.22 s
[Task 13/25] Current/Best: 12.52/ 21.36 GFLOPS | Progress: (16/20) | 16.49 s
[Task 13/25] Current/Best: 11.62/ 21.36 GFLOPS | Progress: (20/20) | 20.31 s Done.
-
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 12.61/ 17.38 GFLOPS | Progress: (4/20) | 5.75 s
[Task 14/25] Current/Best: 12.08/ 17.38 GFLOPS | Progress: (8/20) | 9.23 s
[Task 14/25] Current/Best: 18.71/ 18.71 GFLOPS | Progress: (12/20) | 15.09 s
[Task 14/25] Current/Best: 11.94/ 18.71 GFLOPS | Progress: (16/20) | 17.20 s
[Task 14/25] Current/Best: 5.20/ 18.71 GFLOPS | Progress: (20/20) | 23.16 s Done.
-
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 13.56/ 14.85 GFLOPS | Progress: (4/20) | 4.81 s
[Task 15/25] Current/Best: 15.99/ 15.99 GFLOPS | Progress: (8/20) | 7.31 s
[Task 15/25] Current/Best: 12.78/ 18.22 GFLOPS | Progress: (12/20) | 10.14 s
[Task 15/25] Current/Best: 10.92/ 18.22 GFLOPS | Progress: (16/20) | 13.34 s
[Task 15/25] Current/Best: 10.19/ 18.22 GFLOPS | Progress: (20/20) | 17.53 s Done.
-
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 15.76/ 18.51 GFLOPS | Progress: (4/20) | 4.69 s
[Task 16/25] Current/Best: 9.20/ 18.51 GFLOPS | Progress: (8/20) | 6.45 s
[Task 16/25] Current/Best: 15.85/ 18.51 GFLOPS | Progress: (12/20) | 8.20 s
[Task 16/25] Current/Best: 15.08/ 18.51 GFLOPS | Progress: (16/20) | 10.13 s
[Task 16/25] Current/Best: 15.75/ 20.16 GFLOPS | Progress: (20/20) | 12.90 s Done.
-
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 9.65/ 18.96 GFLOPS | Progress: (4/20) | 5.29 s
[Task 17/25] Current/Best: 17.19/ 18.96 GFLOPS | Progress: (8/20) | 8.45 s
[Task 17/25] Current/Best: 12.02/ 18.96 GFLOPS | Progress: (12/20) | 11.15 s
[Task 17/25] Current/Best: 11.79/ 20.72 GFLOPS | Progress: (16/20) | 13.85 s
[Task 17/25] Current/Best: 10.06/ 23.42 GFLOPS | Progress: (20/20) | 16.47 s Done.
-
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 18.58/ 18.58 GFLOPS | Progress: (4/20) | 5.19 s
[Task 18/25] Current/Best: 6.13/ 18.58 GFLOPS | Progress: (8/20) | 8.55 s
[Task 18/25] Current/Best: 5.96/ 18.58 GFLOPS | Progress: (12/20) | 11.15 s
[Task 18/25] Current/Best: 11.68/ 18.58 GFLOPS | Progress: (16/20) | 14.44 s
[Task 18/25] Current/Best: 12.20/ 18.58 GFLOPS | Progress: (20/20) | 17.44 s Done.
-
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 5.96/ 18.60 GFLOPS | Progress: (4/20) | 6.60 s
[Task 19/25] Current/Best: 10.55/ 22.40 GFLOPS | Progress: (8/20) | 9.60 s
[Task 19/25] Current/Best: 6.93/ 22.40 GFLOPS | Progress: (12/20) | 13.37 s
[Task 19/25] Current/Best: 6.49/ 22.40 GFLOPS | Progress: (16/20) | 16.89 s
[Task 19/25] Current/Best: 20.23/ 22.40 GFLOPS | Progress: (20/20) | 21.19 s Done.
-
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 5.06/ 9.64 GFLOPS | Progress: (4/20) | 5.19 s
[Task 20/25] Current/Best: 4.75/ 15.55 GFLOPS | Progress: (8/20) | 8.40 s
[Task 20/25] Current/Best: 16.91/ 16.91 GFLOPS | Progress: (12/20) | 12.15 s
[Task 20/25] Current/Best: 7.68/ 17.34 GFLOPS | Progress: (16/20) | 15.10 s
[Task 20/25] Current/Best: 8.20/ 17.37 GFLOPS | Progress: (20/20) | 18.94 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 19.38/ 19.38 GFLOPS | Progress: (4/20) | 5.73 s
[Task 21/25] Current/Best: 20.33/ 20.33 GFLOPS | Progress: (8/20) | 7.55 s
[Task 21/25] Current/Best: 19.97/ 20.33 GFLOPS | Progress: (12/20) | 10.59 s Done.
-
[Task 21/25] Current/Best: 2.74/ 20.33 GFLOPS | Progress: (16/20) | 14.49 s
[Task 21/25] Current/Best: 2.71/ 20.33 GFLOPS | Progress: (20/20) | 17.35 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 10.44/ 10.99 GFLOPS | Progress: (4/20) | 7.10 s
[Task 22/25] Current/Best: 10.68/ 11.81 GFLOPS | Progress: (8/20) | 9.09 s
[Task 22/25] Current/Best: 15.97/ 18.41 GFLOPS | Progress: (12/20) | 11.19 s
[Task 22/25] Current/Best: 9.86/ 19.61 GFLOPS | Progress: (16/20) | 13.42 s
[Task 22/25] Current/Best: 7.42/ 19.61 GFLOPS | Progress: (20/20) | 15.22 s Done.
-
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 5.35/ 18.45 GFLOPS | Progress: (4/20) | 6.49 s
[Task 23/25] Current/Best: 10.09/ 19.34 GFLOPS | Progress: (8/20) | 10.92 s
[Task 23/25] Current/Best: 12.20/ 19.34 GFLOPS | Progress: (12/20) | 13.96 s
[Task 23/25] Current/Best: 19.75/ 19.75 GFLOPS | Progress: (16/20) | 19.02 s
[Task 23/25] Current/Best: 13.81/ 19.75 GFLOPS | Progress: (20/20) | 21.82 s Done.
-
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 6.60/ 8.73 GFLOPS | Progress: (4/20) | 13.72 s
[Task 24/25] Current/Best: 1.03/ 8.73 GFLOPS | Progress: (8/20) | 24.41 s
[Task 24/25] Current/Best: 6.14/ 8.73 GFLOPS | Progress: (12/20) | 37.30 s
[Task 24/25] Current/Best: 1.82/ 8.73 GFLOPS | Progress: (16/20) | 40.04 s
[Task 24/25] Current/Best: 9.31/ 9.31 GFLOPS | Progress: (20/20) | 45.43 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
-
[Task 25/25] Current/Best: 3.50/ 8.21 GFLOPS | Progress: (4/20) | 13.76 s
[Task 25/25] Current/Best: 7.86/ 9.44 GFLOPS | Progress: (8/20) | 24.69 s
[Task 25/25] Current/Best: 8.71/ 9.44 GFLOPS | Progress: (12/20) | 27.33 s
[Task 25/25] Current/Best: 1.54/ 9.44 GFLOPS | Progress: (16/20) | 38.01 s
[Task 25/25] Current/Best: 3.37/ 9.44 GFLOPS | Progress: (20/20) | 50.29 s
+
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 1/25] Current/Best: 12.76/ 14.93 GFLOPS | Progress: (4/20) | 9.80 s
[Task 1/25] Current/Best: 6.92/ 15.43 GFLOPS | Progress: (8/20) | 14.91 s
[Task 1/25] Current/Best: 10.03/ 15.43 GFLOPS | Progress: (12/20) | 18.75 s
[Task 1/25] Current/Best: 8.89/ 22.89 GFLOPS | Progress: (16/20) | 22.14 s
[Task 1/25] Current/Best: 11.16/ 22.89 GFLOPS | Progress: (20/20) | 27.71 s Done.
+
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 2/25] Current/Best: 14.17/ 19.29 GFLOPS | Progress: (4/20) | 4.59 s
[Task 2/25] Current/Best: 14.13/ 19.29 GFLOPS | Progress: (8/20) | 7.52 s
[Task 2/25] Current/Best: 8.35/ 19.29 GFLOPS | Progress: (12/20) | 9.90 s
[Task 2/25] Current/Best: 13.30/ 19.29 GFLOPS | Progress: (16/20) | 11.49 s
[Task 2/25] Current/Best: 17.41/ 19.29 GFLOPS | Progress: (20/20) | 12.93 s Done.
+
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 3/25] Current/Best: 6.20/ 20.15 GFLOPS | Progress: (4/20) | 4.96 s
[Task 3/25] Current/Best: 12.63/ 21.66 GFLOPS | Progress: (8/20) | 7.51 s
[Task 3/25] Current/Best: 7.59/ 21.66 GFLOPS | Progress: (12/20) | 10.19 s
[Task 3/25] Current/Best: 19.29/ 21.66 GFLOPS | Progress: (16/20) | 13.21 s
[Task 3/25] Current/Best: 12.49/ 23.40 GFLOPS | Progress: (20/20) | 15.19 s Done.
+
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 4/25] Current/Best: 17.47/ 17.47 GFLOPS | Progress: (4/20) | 7.54 s
[Task 4/25] Current/Best: 13.26/ 18.18 GFLOPS | Progress: (8/20) | 9.47 s
[Task 4/25] Current/Best: 11.85/ 18.18 GFLOPS | Progress: (12/20) | 11.70 s
[Task 4/25] Current/Best: 5.72/ 18.18 GFLOPS | Progress: (16/20) | 13.98 s
[Task 4/25] Current/Best: 14.04/ 18.18 GFLOPS | Progress: (20/20) | 16.79 s Done.
+
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 5/25] Current/Best: 10.87/ 17.63 GFLOPS | Progress: (4/20) | 5.05 s
[Task 5/25] Current/Best: 5.23/ 17.93 GFLOPS | Progress: (8/20) | 7.35 s
[Task 5/25] Current/Best: 11.46/ 17.93 GFLOPS | Progress: (12/20) | 10.17 s
[Task 5/25] Current/Best: 17.84/ 17.93 GFLOPS | Progress: (16/20) | 12.52 s
[Task 5/25] Current/Best: 16.15/ 20.04 GFLOPS | Progress: (20/20) | 14.38 s Done.
+
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 6/25] Current/Best: 3.00/ 14.81 GFLOPS | Progress: (4/20) | 5.83 s
[Task 6/25] Current/Best: 5.09/ 14.81 GFLOPS | Progress: (8/20) | 9.16 s
[Task 6/25] Current/Best: 20.63/ 20.63 GFLOPS | Progress: (12/20) | 11.53 s
[Task 6/25] Current/Best: 16.32/ 20.63 GFLOPS | Progress: (16/20) | 13.67 s
[Task 6/25] Current/Best: 14.73/ 20.63 GFLOPS | Progress: (20/20) | 16.80 s Done.
+
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 7/25] Current/Best: 23.12/ 23.12 GFLOPS | Progress: (4/20) | 5.32 s
[Task 7/25] Current/Best: 7.03/ 23.12 GFLOPS | Progress: (8/20) | 7.68 s
[Task 7/25] Current/Best: 12.18/ 23.12 GFLOPS | Progress: (12/20) | 9.95 s
[Task 7/25] Current/Best: 7.11/ 23.12 GFLOPS | Progress: (16/20) | 12.59 s
[Task 7/25] Current/Best: 12.49/ 23.12 GFLOPS | Progress: (20/20) | 16.03 s Done.
+
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 8/25] Current/Best: 18.88/ 18.88 GFLOPS | Progress: (4/20) | 5.18 s
[Task 8/25] Current/Best: 13.70/ 18.88 GFLOPS | Progress: (8/20) | 8.15 s
[Task 8/25] Current/Best: 20.22/ 20.22 GFLOPS | Progress: (12/20) | 10.64 s
[Task 8/25] Current/Best: 10.34/ 20.22 GFLOPS | Progress: (16/20) | 14.45 s
[Task 8/25] Current/Best: 3.38/ 20.22 GFLOPS | Progress: (20/20) | 17.64 s Done.
+
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 9/25] Current/Best: 20.31/ 20.31 GFLOPS | Progress: (4/20) | 8.79 s
[Task 9/25] Current/Best: 22.58/ 22.58 GFLOPS | Progress: (8/20) | 11.44 s
[Task 9/25] Current/Best: 9.10/ 22.58 GFLOPS | Progress: (12/20) | 15.00 s
[Task 9/25] Current/Best: 7.16/ 22.58 GFLOPS | Progress: (16/20) | 26.22 s
[Task 9/25] Current/Best: 14.84/ 22.58 GFLOPS | Progress: (20/20) | 29.65 s
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 10/25] Current/Best: 4.04/ 16.99 GFLOPS | Progress: (4/20) | 4.79 s
[Task 10/25] Current/Best: 10.09/ 16.99 GFLOPS | Progress: (8/20) | 7.74 s
[Task 10/25] Current/Best: 14.18/ 16.99 GFLOPS | Progress: (12/20) | 9.76 s
[Task 10/25] Current/Best: 17.83/ 21.38 GFLOPS | Progress: (16/20) | 11.64 s
[Task 10/25] Current/Best: 10.68/ 21.38 GFLOPS | Progress: (20/20
) | 13.37 s Done.
+
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 11/25] Current/Best: 11.83/ 18.25 GFLOPS | Progress: (4/20) | 5.89 s Done.
+
[Task 11/25] Current/Best: 12.35/ 21.20 GFLOPS | Progress: (8/20) | 8.77 s
[Task 11/25] Current/Best: 8.42/ 21.56 GFLOPS | Progress: (12/20) | 11.21 s
[Task 11/25] Current/Best: 16.18/ 24.11 GFLOPS | Progress: (16/20) | 14.11 s
[Task 11/25] Current/Best: 19.08/ 24.11 GFLOPS | Progress: (20/20) | 16.23 s Done.
+
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 12/25] Current/Best: 4.49/ 18.36 GFLOPS | Progress: (4/20) | 5.59 s
[Task 12/25] Current/Best: 16.31/ 18.36 GFLOPS | Progress: (8/20) | 9.26 s
[Task 12/25] Current/Best: 19.08/ 19.08 GFLOPS | Progress: (12/20) | 12.71 s
[Task 12/25] Current/Best: 10.83/ 19.08 GFLOPS | Progress: (16/20) | 16.19 s
[Task 12/25] Current/Best: 9.54/ 19.08 GFLOPS | Progress: (20/20) | 19.03 s Done.
+
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 13/25] Current/Best: 15.89/ 18.55 GFLOPS | Progress: (4/20) | 5.84 s
[Task 13/25] Current/Best: 11.35/ 18.55 GFLOPS | Progress: (8/20) | 9.79 s
[Task 13/25] Current/Best: 3.10/ 18.55 GFLOPS | Progress: (12/20) | 13.29 s
[Task 13/25] Current/Best: 9.54/ 18.55 GFLOPS | Progress: (16/20) | 15.89 s
[Task 13/25] Current/Best: 21.35/ 21.35 GFLOPS | Progress: (20/20) | 18.28 s Done.
+
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 14/25] Current/Best: 10.97/ 10.97 GFLOPS | Progress: (4/20) | 9.26 s
[Task 14/25] Current/Best: 13.41/ 13.41 GFLOPS | Progress: (8/20) | 13.88 s
[Task 14/25] Current/Best: 16.80/ 16.80 GFLOPS | Progress: (12/20) | 17.83 s
[Task 14/25] Current/Best: 11.37/ 16.80 GFLOPS | Progress: (16/20) | 23.85 s
[Task 14/25] Current/Best: 8.97/ 16.93 GFLOPS | Progress: (20/20) | 27.07 s
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 15/25] Current/Best: 5.10/ 22.42 GFLOPS | Progress: (4/20) | 8.47 s
[Task 15/25] Current/Best: 22.74/ 22.74 GFLOPS | Progress: (8/20) | 12.38 s
[Task 15/25] Current/Best: 9.20/ 22.74 GFLOPS | Progress: (12/20) | 15.50 s Done.
+
[Task 15/25] Current/Best: 14.19/ 22.74 GFLOPS | Progress: (16/20) | 21.97 s
[Task 15/25] Current/Best: 11.59/ 22.74 GFLOPS | Progress: (20/20) | 24.22 s Done.
+
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 16/25] Current/Best: 12.44/ 15.08 GFLOPS | Progress: (4/20) | 4.72 s
[Task 16/25] Current/Best: 11.12/ 18.56 GFLOPS | Progress: (8/20) | 7.23 s
[Task 16/25] Current/Best: 10.64/ 19.50 GFLOPS | Progress: (12/20) | 10.32 s
[Task 16/25] Current/Best: 5.81/ 19.50 GFLOPS | Progress: (16/20) | 14.36 s
[Task 16/25] Current/Best: 5.41/ 19.50 GFLOPS | Progress: (20/20) | 17.57 s Done.
+
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 17/25] Current/Best: 12.20/ 12.20 GFLOPS | Progress: (4/20) | 6.81 s
[Task 17/25] Current/Best: 9.86/ 21.14 GFLOPS | Progress: (8/20) | 9.13 s
[Task 17/25] Current/Best: 12.18/ 21.14 GFLOPS | Progress: (12/20) | 11.83 s
[Task 17/25] Current/Best: 7.65/ 21.14 GFLOPS | Progress: (16/20) | 14.67 s
[Task 17/25] Current/Best: 15.92/ 21.14 GFLOPS | Progress: (20/20) | 18.36 s Done.
+
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 18/25] Current/Best: 3.09/ 16.34 GFLOPS | Progress: (4/20) | 6.77 s
[Task 18/25] Current/Best: 14.90/ 18.33 GFLOPS | Progress: (8/20) | 8.83 s
[Task 18/25] Current/Best: 10.09/ 18.33 GFLOPS | Progress: (12/20) | 13.01 s
[Task 18/25] Current/Best: 18.50/ 18.50 GFLOPS | Progress: (16/20) | 18.83 s
[Task 18/25] Current/Best: 19.35/ 20.65 GFLOPS | Progress: (20/20) | 22.49 s Done.
+
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 19/25] Current/Best: 19.70/ 19.70 GFLOPS | Progress: (4/20) | 5.08 s
[Task 19/25] Current/Best: 6.14/ 19.70 GFLOPS | Progress: (8/20) | 8.12 s
[Task 19/25] Current/Best: 16.82/ 19.70 GFLOPS | Progress: (12/20) | 13.31 s
[Task 19/25] Current/Best: 17.36/ 19.70 GFLOPS | Progress: (16/20) | 16.56 s
[Task 19/25] Current/Best: 7.62/ 19.70 GFLOPS | Progress: (20/20) | 20.62 s Done.
+
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 20/25] Current/Best: 8.28/ 14.97 GFLOPS | Progress: (4/20) | 4.98 s
[Task 20/25] Current/Best: 11.43/ 16.56 GFLOPS | Progress: (8/20) | 7.13 s
[Task 20/25] Current/Best: 14.40/ 16.56 GFLOPS | Progress: (12/20) | 10.63 s
[Task 20/25] Current/Best: 12.33/ 18.74 GFLOPS | Progress: (16/20) | 13.21 s
[Task 20/25] Current/Best: 17.88/ 18.74 GFLOPS | Progress: (20/20) | 15.07 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 21/25] Current/Best: 12.09/ 12.09 GFLOPS | Progress: (4/20) | 6.58 s
[Task 21/25] Current/Best: 7.21/ 20.24 GFLOPS | Progress: (8/20) | 9.13 s
[Task 21/25] Current/Best: 8.88/ 22.66 GFLOPS | Progress: (12/20) | 11.76 s
[Task 21/25] Current/Best: 6.72/ 22.66 GFLOPS | Progress: (16/20) | 13.76 s
[Task 21/25] Current/Best: 14.55/ 22.66 GFLOPS | Progress: (20/20
) | 17.55 s Done.
+
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 22/25] Current/Best: 2.69/ 16.00 GFLOPS | Progress: (4/20) | 4.85 s
[Task 22/25] Current/Best: 9.87/ 16.86 GFLOPS | Progress: (8/20) | 7.66 s
[Task 22/25] Current/Best: 10.04/ 18.05 GFLOPS | Progress: (12/20) | 10.65 s
[Task 22/25] Current/Best: 5.20/ 20.69 GFLOPS | Progress: (16/20) | 12.51 s
[Task 22/25] Current/Best: 15.96/ 20.69 GFLOPS | Progress: (20/20) | 14.34 s Done.
+
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 23/25] Current/Best: 6.48/ 6.48 GFLOPS | Progress: (4/20) | 6.85 s
[Task 23/25] Current/Best: 10.48/ 13.45 GFLOPS | Progress: (8/20) | 10.20 s
[Task 23/25] Current/Best: 12.40/ 13.45 GFLOPS | Progress: (12/20) | 14.00 s
[Task 23/25] Current/Best: 19.43/ 19.43 GFLOPS | Progress: (16/20) | 16.83 s
[Task 23/25] Current/Best: 6.13/ 19.43 GFLOPS | Progress: (20/20) | 20.31 s Done.
+
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s
[Task 24/25] Current/Best: 6.95/ 9.59 GFLOPS | Progress: (4/20) | 4.86 s
[Task 24/25] Current/Best: 3.56/ 9.59 GFLOPS | Progress: (8/20) | 15.85 s
[Task 24/25] Current/Best: 0.82/ 9.59 GFLOPS | Progress: (12/20) | 18.91 s
[Task 24/25] Current/Best: 0.00/ 9.59 GFLOPS | Progress: (16/20) | 28.99 s
[Task 24/25] Current/Best: 0.85/ 9.59 GFLOPS | Progress: (20/20) | 41.73 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/20) | 0.00 s Done.
+ Done.
+
[Task 25/25] Current/Best: 3.54/ 4.17 GFLOPS | Progress: (4/20) | 14.98 s
[Task 25/25] Current/Best: 3.63/ 5.59 GFLOPS | Progress: (8/20) | 25.66 s
[Task 25/25] Current/Best: 8.35/ 8.35 GFLOPS | Progress: (12/20) | 28.48 s
[Task 25/25] Current/Best: 7.64/ 8.35 GFLOPS | Progress: (16/20) | 33.55 s
[Task 25/25] Current/Best: 1.53/ 8.35 GFLOPS | Progress: (20/20) | 44.57 s
@@ -722,8 +723,8 @@ improvement in comparing the optimized model to the unoptimized model.
.. code-block:: none
- optimized: {'mean': 415.0788926600035, 'median': 415.16708340001287, 'std': 1.6494581347165918}
- unoptimized: {'mean': 515.0090213199996, 'median': 515.2850962500054, 'std': 2.0018207248826223}
+ optimized: {'mean': 420.73439764999875, 'median': 419.8400248500093, 'std': 4.782772813666293}
+ unoptimized: {'mean': 512.6563362900002, 'median': 512.4201752500028, 'std': 1.346805710381719}
@@ -746,7 +747,7 @@ profiling/benchmarking.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 12 minutes 26.486 seconds)
+ **Total running time of the script:** ( 12 minutes 27.534 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 8eff72f5fd..9157735847 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
- 1.252e-07 secs/op
+ 1.273e-07 secs/op
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index 4abdcea5a6..28ce799227 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, 0xe887ce0)), stage(b, placeholder(b, 0x1c14f920)), stage(T_add, compute(T_add, body=[a[ax0, ax1, ax2] + b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T.Range(0, 10), "DataPar", ""), T.iter_var(ax2, T.Range(0, 10), "DataPar", "")], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[a[ax0, ax1, ax2] * b[ax1, ax2]], axis=[T.iter_var(ax0, T.Range(0, 100), "DataPar", ""), T.iter_var(ax1, T. [...]
+ [stage(a, placeholder(a, 0x22742c80)), stage(b, placeholder(b, 0x67d66b0)), 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 5c25d83252..410a64f280 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
=================
-**16:18.814** total execution time for **tutorial** files:
+**16:21.108** total execution time for **tutorial** files:
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 12:26.486 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``) | 12:27.534 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:33.908 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``) | 01:38.775 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 01:01.330 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``) | 00:58.101 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:38.090 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``) | 00:37.113 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:36.506 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``) | 00:36.561 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:01.486 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``) | 00:02.006 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``) | 00:00.846 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``) | 00:00.853 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
-| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.162 | 0.0 MB |
+| :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``) | 00:00.166 | 0.0 MB |
+------------------------------------------------------------------------------------------+-----------+--------+
| :ref:`sphx_glr_tutorial_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 66790b654e..52c78a1f03 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -389,7 +389,7 @@ compile and run this new schedule with the parallel operation applied:
.. code-block:: none
- parallel: 0.000008
+ parallel: 0.000007
@@ -498,10 +498,10 @@ We can now compare the different schedules
.. code-block:: none
Operator Timing Performance
- numpy 6.989260000409558e-06 1.0
- naive 6.6807e-06 0.9558522647044928
- parallel 7.879e-06 1.1273010303720716
- vector 2.46777e-05 3.5308029746430853
+ numpy 7.310850000976643e-06 1.0
+ naive 6.681799999999999e-06 0.9139566533450132
+ parallel 6.9723999999999995e-06 0.9537057933165869
+ vector 2.46222e-05 3.3678983971372363
@@ -922,7 +922,7 @@ matrix multiplication.
.. code-block:: none
- Numpy running time: 0.017893
+ Numpy running time: 0.018142
@@ -980,7 +980,7 @@ optimizations.
.. code-block:: none
- none: 3.438976
+ none: 3.183008
@@ -1080,7 +1080,7 @@ schedule.
.. code-block:: none
- blocking: 0.303537
+ blocking: 0.300140
@@ -1164,7 +1164,7 @@ already cache friendly from our previous optimizations.
.. code-block:: none
- vectorization: 0.335426
+ vectorization: 0.339756
# 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: 0.117224
+ loop permutation: 0.117529
# 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: 0.109691
+ array packing: 0.107828
# 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: 0.111505
+ block caching: 0.110102
# 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: 0.146021
+ parallelization: 0.145577
# 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 3.4389756333999997 1.0
- blocking 0.30353692769999996 0.08826376225291925
- vectorization 0.33542634050000003 0.09753670169752708
- loop permutation 0.11722409649999999 0.03408692267589709
- array packing 0.109690763 0.03189634783528618
- block caching 0.11150506980000001 0.03242391970360043
- parallelization 0.1460211104 0.04246064118100012
+ none 3.1830083366 1.0
+ blocking 0.3001401027 0.09429447584187016
+ vectorization 0.3397556593 0.10674042395469108
+ loop permutation 0.1175288184 0.03692381733613082
+ array packing 0.10782795860000001 0.033876115673381745
+ block caching 0.11010203970000002 0.034590559639440945
+ parallelization 0.14557671600000002 0.04573557484159812
@@ -1594,11 +1594,6 @@ operations with tunable parameters that allows you to automatically optimize
the computation for specific platforms.
-.. rst-class:: sphx-glr-timing
-
- **Total running time of the script:** ( 1 minutes 1.330 seconds)
-
-
.. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
.. only:: html
diff --git a/docs/commit_hash b/docs/commit_hash
index 0454dde131..67c21ff1e7 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-e8576db88bad824a9997b8d499a9730059d0348c
+902c2e2db70b75c36f9bd8c253707b1e8761cc18
diff --git a/docs/how_to/compile_models/from_darknet.html b/docs/how_to/compile_models/from_darknet.html
index e41da66701..cd9881819a 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 18.519 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 18.997 seconds)</p>
<div class="sphx-glr-footer sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-darknet-py">
<div class="sphx-glr-download sphx-glr-download-python docutils container">
<p><a class="reference download internal" download="" href="../../_downloads/7716f96385bd5abb6e822041e285be54/from_darknet.py"><code class="xref download docutils literal notranslate"><span class="pre">Download</span> <span class="pre">Python</span> <span class="pre">source</span> <span class="pre">code:</span> <span class="pre">from_darknet.py</span></code></a></p>
diff --git a/docs/how_to/compile_models/from_keras.html b/docs/how_to/compile_models/from_keras.html
index 1eeb99b618..7be7b6397a 100644
--- a/docs/how_to/compile_models/from_keras.html
+++ b/docs/how_to/compile_models/from_keras.html
@@ -511,7 +511,7 @@ Tensorflow is also required since it’s used as the default backend of keras.</
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Relay top-1 id: 285, class name: Egyptian cat
1/1 [==============================] - ETA: 0s
-1/1 [==============================] - 1s 947ms/step
+1/1 [==============================] - 1s 986ms/step
Keras top-1 id: 285, class name: Egyptian cat
</pre></div>
</div>
diff --git a/docs/how_to/compile_models/from_mxnet.html b/docs/how_to/compile_models/from_mxnet.html
index dc95b0bdd6..437ceafbf8 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.zipb927c167-6dd7-4bda-82fb-8245299da4bf 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.ziped3d270c-0a1e-4c3b-89bf-3b363ea686b2 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 8ff30525c8..d8ac8e679a 100644
--- a/docs/how_to/compile_models/from_oneflow.html
+++ b/docs/how_to/compile_models/from_oneflow.html
@@ -454,12 +454,11 @@ 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, 37.2MB/s]
- 39%|###8 | 16.0M/41.5M [00:00<00:00, 54.8MB/s]
- 58%|#####7 | 24.0M/41.5M [00:00<00:00, 50.7MB/s]
- 77%|#######7 | 32.0M/41.5M [00:00<00:00, 59.7MB/s]
- 96%|#########6| 40.0M/41.5M [00:00<00:00, 65.0MB/s]
-100%|##########| 41.5M/41.5M [00:00<00:00, 59.9MB/s]
+ 22%|##2 | 9.19M/41.5M [00:00<00:00, 96.3MB/s]
+ 44%|####4 | 18.4M/41.5M [00:00<00:00, 76.6MB/s]
+ 69%|######9 | 28.6M/41.5M [00:00<00:00, 89.0MB/s]
+ 90%|######### | 37.4M/41.5M [00:00<00:00, 82.8MB/s]
+100%|##########| 41.5M/41.5M [00:00<00:00, 78.5MB/s]
</pre></div>
</div>
</div>
diff --git a/docs/how_to/compile_models/from_pytorch.html b/docs/how_to/compile_models/from_pytorch.html
index adad65e438..7be501396e 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -437,11 +437,11 @@ 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, 55.7MB/s]
- 36%|###5 | 16.0M/44.7M [00:00<00:00, 65.5MB/s]
- 60%|#####9 | 26.6M/44.7M [00:00<00:00, 77.1MB/s]
- 90%|########9 | 40.0M/44.7M [00:00<00:00, 95.6MB/s]
-100%|##########| 44.7M/44.7M [00:00<00:00, 86.6MB/s]
+ 23%|##2 | 10.1M/44.7M [00:00<00:00, 104MB/s]
+ 45%|####4 | 20.0M/44.7M [00:00<00:00, 92.7MB/s]
+ 67%|######6 | 29.8M/44.7M [00:00<00:00, 97.2MB/s]
+ 90%|########9 | 40.0M/44.7M [00:00<00:00, 91.9MB/s]
+100%|##########| 44.7M/44.7M [00:00<00:00, 80.3MB/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 c2d37bc97d..575c02c5f1 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -654,7 +654,7 @@ banana (score = 0.00022)
desk (score = 0.00019)
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 22.917 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 24.787 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 1b360b5b87..b3cac55976 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>06:34.510</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>06:40.909</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>01:22.917</p></td>
+<td><p>01:24.787</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>01:18.519</p></td>
+<td><p>01:18.997</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>00:54.985</p></td>
+<td><p>00:55.752</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>00:37.433</p></td>
+<td><p>00:38.385</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>00:31.631</p></td>
+<td><p>00:32.203</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="from_mxnet.html#sphx-glr-how-to-compile-models-from-mxnet-py"><span class="std std-ref">Compile MXNet Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_mxnet.py</span></code>)</p></td>
-<td><p>00:31.128</p></td>
+<td><p>00:31.879</p></td>
<td><p>0.0 MB</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="from_tflite.html#sphx-glr-how-to-compile-models-from-tflite-py"><span class="std std-ref">Compile TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tflite.py</span></code>)</p></td>
-<td><p>00:27.247</p></td>
+<tr class="row-odd"><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>00:27.035</p></td>
<td><p>0.0 MB</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="from_pytorch.html#sphx-glr-how-to-compile-models-from-pytorch-py"><span class="std std-ref">Compile PyTorch Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_pytorch.py</span></code>)</p></td>
-<td><p>00:25.448</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="from_tflite.html#sphx-glr-how-to-compile-models-from-tflite-py"><span class="std std-ref">Compile TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">from_tflite.py</span></code>)</p></td>
+<td><p>00:26.668</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>00:22.479</p></td>
+<td><p>00:22.489</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>00:02.724</p></td>
+<td><p>00:02.715</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 627e411b3f..9f2a727d7a 100644
--- a/docs/how_to/deploy_models/deploy_model_on_adreno.html
+++ b/docs/how_to/deploy_models/deploy_model_on_adreno.html
@@ -925,7 +925,7 @@ Top5 predictions:
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 2678.3957 2678.5500 2682.2350 2676.1978 1.5449
+ 2682.8694 2682.6587 2688.6528 2677.6245 2.7009
</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_android.html b/docs/how_to/deploy_models/deploy_model_on_android.html
index 5cfb22e71c..670f916c26 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)
- 15.5596 15.5324 15.6770 15.4915 0.0597
+ 15.8619 15.6985 16.6804 15.4592 0.3908
</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 45289a6dc2..0802a82c94 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -459,25 +459,22 @@ be unstable.</p>
Downloading: "https://download.pytorch.org/models/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth" to /workspace/.cache/torch/hub/checkpoints/maskrcnn_resnet50_fpn_coco-bf2d0c1e.pth
0%| | 0.00/170M [00:00<?, ?B/s]
- 5%|4 | 7.99M/170M [00:00<00:02, 73.6MB/s]
- 9%|9 | 16.0M/170M [00:00<00:02, 72.3MB/s]
- 15%|#5 | 26.1M/170M [00:00<00:01, 80.3MB/s]
- 20%|#9 | 33.7M/170M [00:00<00:01, 72.4MB/s]
- 24%|##3 | 40.7M/170M [00:00<00:02, 67.6MB/s]
- 28%|##7 | 47.1M/170M [00:00<00:02, 63.7MB/s]
- 33%|###2 | 56.0M/170M [00:00<00:01, 61.2MB/s]
- 39%|###9 | 66.8M/170M [00:00<00:01, 74.6MB/s]
- 44%|####3 | 74.2M/170M [00:01<00:01, 73.1MB/s]
- 51%|##### | 86.3M/170M [00:01<00:01, 82.4MB/s]
- 56%|#####5 | 94.3M/170M [00:01<00:01, 72.7MB/s]
- 61%|######1 | 104M/170M [00:01<00:00, 71.6MB/s]
- 66%|######5 | 112M/170M [00:01<00:00, 74.6MB/s]
- 75%|#######5 | 128M/170M [00:01<00:00, 97.7MB/s]
- 81%|######## | 137M/170M [00:01<00:00, 80.9MB/s]
- 87%|########6 | 148M/170M [00:02<00:00, 86.9MB/s]
- 92%|#########2| 157M/170M [00:02<00:00, 87.0MB/s]
- 97%|#########7| 165M/170M [00:02<00:00, 69.6MB/s]
-100%|##########| 170M/170M [00:02<00:00, 74.9MB/s]
+ 5%|4 | 7.99M/170M [00:00<00:02, 74.9MB/s]
+ 12%|#1 | 19.5M/170M [00:00<00:01, 101MB/s]
+ 17%|#7 | 29.7M/170M [00:00<00:01, 103MB/s]
+ 24%|##3 | 40.0M/170M [00:00<00:01, 89.2MB/s]
+ 32%|###1 | 54.1M/170M [00:00<00:01, 108MB/s]
+ 38%|###8 | 64.8M/170M [00:00<00:01, 90.2MB/s]
+ 44%|####4 | 75.0M/170M [00:00<00:01, 94.7MB/s]
+ 52%|#####1 | 88.0M/170M [00:00<00:00, 94.6MB/s]
+ 57%|#####7 | 97.4M/170M [00:01<00:00, 94.5MB/s]
+ 63%|######3 | 107M/170M [00:01<00:00, 97.0MB/s]
+ 69%|######8 | 117M/170M [00:01<00:00, 90.8MB/s]
+ 74%|#######3 | 126M/170M [00:01<00:00, 86.9MB/s]
+ 79%|#######9 | 135M/170M [00:01<00:00, 89.2MB/s]
+ 85%|########4 | 144M/170M [00:01<00:00, 82.1MB/s]
+ 94%|#########4| 160M/170M [00:01<00:00, 104MB/s]
+100%|##########| 170M/170M [00:01<00:00, 94.5MB/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=& [...]
@@ -575,7 +572,7 @@ 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 31.461 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes 34.920 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 edd946c62a..ca7dad2744 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -500,8 +500,9 @@ 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, 72.8MB/s]
-100%|##########| 13.6M/13.6M [00:00<00:00, 90.3MB/s]
+ 47%|####6 | 6.30M/13.6M [00:00<00:00, 62.5MB/s]
+ 90%|######### | 12.3M/13.6M [00:00<00:00, 52.6MB/s]
+100%|##########| 13.6M/13.6M [00:00<00:00, 58.4MB/s]
</pre></div>
</div>
</div>
@@ -592,7 +593,7 @@ 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)
- 90.2658 89.9287 95.1984 89.7850 1.0279
+ 90.2995 90.2016 93.5701 90.0290 0.3953
</pre></div>
</div>
<div class="admonition note">
@@ -631,7 +632,7 @@ 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 15.930 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 16.851 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 5ffcb3673e..6ef4ceb70f 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)
- 116.2722 115.9485 121.2760 115.1441 1.0439
+ 120.1072 120.0018 125.6702 119.4399 0.6456
</pre></div>
</div>
<div class="admonition note">
@@ -613,7 +613,7 @@ network for ARM CPU</span></a>.</p></li>
</ul>
</div></blockquote>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 31.715 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 30.309 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 6f517f094b..a423b7ee8a 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 31.874 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 32.348 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 6c6c93e21d..6aea28099b 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -468,23 +468,24 @@ to your device.</p>
Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
0%| | 0/132723 [00:00<?, ?KB/s]
- 4%|4 | 5801/132723 [00:00<00:02, 58003.56KB/s]
- 10%|# | 13817/132723 [00:00<00:01, 71034.81KB/s]
- 17%|#6 | 21921/132723 [00:00<00:01, 75601.70KB/s]
- 23%|##2 | 30088/132723 [00:00<00:01, 77994.48KB/s]
- 29%|##8 | 37962/132723 [00:00<00:01, 78260.17KB/s]
- 35%|###4 | 46148/132723 [00:00<00:01, 79482.79KB/s]
- 41%|#### | 54327/132723 [00:00<00:00, 80229.93KB/s]
- 47%|####6 | 62371/132723 [00:00<00:00, 80295.91KB/s]
- 53%|#####3 | 70535/132723 [00:00<00:00, 80712.90KB/s]
- 59%|#####9 | 78621/132723 [00:01<00:00, 80757.26KB/s]
- 65%|######5 | 86833/132723 [00:01<00:00, 81171.07KB/s]
- 72%|#######1 | 95037/132723 [00:01<00:00, 81434.23KB/s]
- 78%|#######7 | 103253/132723 [00:01<00:00, 81650.68KB/s]
- 84%|########3 | 111419/132723 [00:01<00:00, 81624.78KB/s]
- 90%|######### | 119582/132723 [00:01<00:00, 80186.97KB/s]
- 96%|#########6| 127799/132723 [00:01<00:00, 80774.25KB/s]
-100%|##########| 132723/132723 [00:01<00:00, 79648.48KB/s]
+ 4%|4 | 5951/132723 [00:00<00:02, 59495.73KB/s]
+ 11%|# | 13941/132723 [00:00<00:01, 71489.98KB/s]
+ 17%|#6 | 21976/132723 [00:00<00:01, 75532.45KB/s]
+ 23%|##2 | 30017/132723 [00:00<00:01, 77447.42KB/s]
+ 28%|##8 | 37762/132723 [00:00<00:01, 56359.24KB/s]
+ 35%|###4 | 45845/132723 [00:00<00:01, 63029.02KB/s]
+ 40%|###9 | 52766/132723 [00:00<00:01, 58941.31KB/s]
+ 46%|####5 | 60748/132723 [00:00<00:01, 64516.85KB/s]
+ 51%|#####1 | 68178/132723 [00:01<00:00, 67216.50KB/s]
+ 57%|#####7 | 76228/132723 [00:01<00:00, 70973.02KB/s]
+ 63%|######3 | 84255/132723 [00:01<00:00, 73647.00KB/s]
+ 69%|######9 | 92177/132723 [00:01<00:00, 75267.31KB/s]
+ 75%|#######5 | 100071/132723 [00:01<00:00, 76343.19KB/s]
+ 81%|########1 | 107807/132723 [00:01<00:00, 75478.92KB/s]
+ 87%|########7 | 115633/132723 [00:01<00:00, 76293.25KB/s]
+ 93%|#########3| 123606/132723 [00:01<00:00, 77303.12KB/s]
+ 99%|#########9| 131637/132723 [00:01<00:00, 78193.00KB/s]
+100%|##########| 132723/132723 [00:01<00:00, 71404.26KB/s]
</pre></div>
</div>
<p>Create TVM runtime and do inference
@@ -523,7 +524,7 @@ 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 46.232 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 54.095 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 4bfd144761..1828366167 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-deploy-models-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>15:10.721</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>15:23.240</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 86%" />
@@ -354,39 +354,39 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_ssd_gluoncv.html#sphx-glr-how-to-deploy-models-deploy-ssd-gluoncv-py"><span class="std std-ref">Deploy Single Shot Multibox Detector(SSD) model</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_ssd_gluoncv.py</span></code>)</p></td>
-<td><p>03:46.232</p></td>
+<td><p>03:54.095</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>03:31.461</p></td>
+<td><p>03:34.920</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_prequantized_tflite.html#sphx-glr-how-to-deploy-models-deploy-prequantized-tflite-py"><span class="std std-ref">Deploy a Framework-prequantized Model with TVM - Part 3 (TFLite)</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_prequantized_tflite.py</span></code>)</p></td>
-<td><p>02:31.715</p></td>
+<td><p>02:30.309</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>01:31.874</p></td>
+<td><p>01:32.348</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>01:15.930</p></td>
+<td><p>01:16.851</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>00:56.162</p></td>
+<td><p>00:56.338</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_model_on_android.html#sphx-glr-how-to-deploy-models-deploy-model-on-android-py"><span class="std std-ref">Deploy the Pretrained Model on Android</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_android.py</span></code>)</p></td>
-<td><p>00:42.182</p></td>
+<td><p>00:42.547</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="deploy_model_on_nano.html#sphx-glr-how-to-deploy-models-deploy-model-on-nano-py"><span class="std std-ref">Deploy the Pretrained Model on Jetson Nano</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_nano.py</span></code>)</p></td>
-<td><p>00:27.740</p></td>
+<td><p>00:28.089</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="deploy_model_on_rasp.html#sphx-glr-how-to-deploy-models-deploy-model-on-rasp-py"><span class="std std-ref">Deploy the Pretrained Model on Raspberry Pi</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_model_on_rasp.py</span></code>)</p></td>
-<td><p>00:27.419</p></td>
+<td><p>00:27.738</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="deploy_sparse.html#sphx-glr-how-to-deploy-models-deploy-sparse-py"><span class="std std-ref">Deploy a Hugging Face Pruned Model on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">deploy_sparse.py</span></code>)</p></td>
diff --git a/docs/how_to/extend_tvm/bring_your_own_datatypes.html b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
index 82c6e8a5eb..fb05a67314 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.zip176d3132-d7ac-4e94-8d7f-8b6c12b275a6 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.zip4ed6713c-af6a-4491-98d8-b627d6fa4630 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 0099955805..fc1f1d4a22 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>00:54.557</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:54.631</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>00:50.593</p></td>
+<td><p>00:50.749</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>00:02.842</p></td>
+<td><p>00:02.784</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>00:01.112</p></td>
+<td><p>00:01.091</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="low_level_custom_pass.html#sphx-glr-how-to-extend-tvm-low-level-custom-pass-py"><span class="std std-ref">Writing a Customized Pass</span></a> (<code class="docutils literal notranslate"><span class="pre">low_level_custom_pass.py</span></code>)</p></td>
-<td><p>00:00.009</p></td>
+<td><p>00:00.008</p></td>
<td><p>0.0 MB</p></td>
</tr>
</tbody>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index cc7e497638..9ea9e5dd22 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: 22538us [22538us] (48.58%; 48.58%)
-FoldScaleAxis: 23853us [8us] (51.42%; 51.42%)
- FoldConstant: 23845us [1752us] (51.40%; 99.97%)
- InferType: 22093us [22093us] (47.62%; 92.65%)
+InferType: 22432us [22432us] (48.66%; 48.66%)
+FoldScaleAxis: 23666us [9us] (51.34%; 51.34%)
+ FoldConstant: 23657us [1804us] (51.32%; 99.96%)
+ InferType: 21853us [21853us] (47.41%; 92.38%)
</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: 21907us [21907us] (48.39%; 48.39%)
-FoldScaleAxis: 23369us [6us] (51.61%; 51.61%)
- FoldConstant: 23363us [1766us] (51.60%; 99.97%)
- InferType: 21597us [21597us] (47.70%; 92.44%)
+InferType: 21736us [21736us] (47.66%; 47.66%)
+FoldScaleAxis: 23874us [5us] (52.34%; 52.34%)
+ FoldConstant: 23869us [1750us] (52.33%; 99.98%)
+ InferType: 22119us [22119us] (48.50%; 92.67%)
</pre></div>
</div>
<p>Register empty list to clear existing instruments.</p>
diff --git a/docs/how_to/optimize_operators/opt_conv_cuda.html b/docs/how_to/optimize_operators/opt_conv_cuda.html
index 7d48a35f5a..3c4163338f 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: 34.640190 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.244159 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 1e2c8462fa..018422e622 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: 13.358934 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 6.854656 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 53200baee2..8efd456309 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: 0.019040
-Baseline: 3.425819
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019124
+Baseline: 3.231377
</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: 0.298916
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.308924
</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: 0.336510
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.332410
</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: 0.116727
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.116921
</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: 0.110358
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.110576
</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: 0.111632
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.112377
</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: 0.148690
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.148110
</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 f456f68498..16266b387f 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-optimize-operators-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:35.224</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:34.601</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -354,15 +354,15 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="opt_gemm.html#sphx-glr-how-to-optimize-operators-opt-gemm-py"><span class="std std-ref">How to optimize GEMM on CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">opt_gemm.py</span></code>)</p></td>
-<td><p>00:32.531</p></td>
+<td><p>00:32.066</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>00:01.597</p></td>
+<td><p>00:01.493</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>00:01.097</p></td>
+<td><p>00:01.042</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 f8d2089070..a0e6cabe13 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>10:04.125</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>09:57.707</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>06:14.904</p></td>
+<td><p>06:08.709</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>01:41.717</p></td>
+<td><p>01:42.014</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_network_cuda.html#sphx-glr-how-to-tune-with-autoscheduler-tune-network-cuda-py"><span class="std std-ref">Auto-scheduling a Neural Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_network_cuda.py</span></code>)</p></td>
-<td><p>01:07.384</p></td>
+<td><p>01:07.519</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>00:32.581</p></td>
+<td><p>00:31.804</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>00:14.054</p></td>
+<td><p>00:14.138</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>00:13.486</p></td>
+<td><p>00:13.521</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 55a50dd497..5b277edd21 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
@@ -510,12 +510,12 @@ 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": True, "global_symbol": "main", "tir.noalias": 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([4032], "float32", "shared")
- kernel_shared = T.allocate([6144], "float32", "shared")
- threadIdx_x = T.launch_thread("threadIdx.x", 112)
- conv2d_nchw_1 = T.Buffer((1,), data=conv2d_nchw, scope="local", align=4)
+ pad_temp_shared = T.allocate([72], "float32", "shared")
+ kernel_shared = T.allocate([3072], "float32", "shared")
+ threadIdx_x = T.launch_thread("threadIdx.x", 64)
+ conv2d_nchw_1 = T.Buffer((14,), data=conv2d_nchw, scope="local", align=32)
conv2d_nchw_1[0] = T.float32(0)
conv2d_nchw_1[1] = T.float32(0)
conv2d_nchw_1[2] = T.float32(0)
@@ -530,257 +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, rx_outer_outer in T.grid(8, 3):
- cse_var_2: T.int32 = rc_outer_outer * 3136
- cse_var_1: T.int32 = rc_outer_outer * 576
+ 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((4032,), 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(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 112] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 112) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 224] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 224) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 336] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 336) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 448] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 448) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 560] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 560) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 672] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 672) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], 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 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 784) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 896] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 896) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1008] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 776], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1120] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1120) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1232] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1232) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1344] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1344) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1456] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1456) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1568] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1568) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1680] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1680) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1792] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1792) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 1904] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 1904) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2016] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 1560], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2128] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2128) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2240] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2240) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2352] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2352) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2464] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2464) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2576] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2576) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2688] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2688) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2800] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2800) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 2912] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 2912) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3024] = T.if_then_else(7 <= threadIdx_x_1 % 63 and threadIdx_x_1 % 63 < 56 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + threadIdx_x_1 // 63 * 49 + rx_outer_outer + threadIdx_x_1 % 63 + 2344], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3136] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 7) % 9 and (threadIdx_x_1 // 7 + 7) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3136) // 63 * 49 + (threadIdx_x_1 // 7 + 7) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3248] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 5) % 9 and (threadIdx_x_1 // 7 + 5) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3248) // 63 * 49 + (threadIdx_x_1 // 7 + 5) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3360] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 3) % 9 and (threadIdx_x_1 // 7 + 3) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3360) // 63 * 49 + (threadIdx_x_1 // 7 + 3) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3472] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 1) % 9 and (threadIdx_x_1 // 7 + 1) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3472) // 63 * 49 + (threadIdx_x_1 // 7 + 1) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3584] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 8) % 9 and (threadIdx_x_1 // 7 + 8) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3584) // 63 * 49 + (threadIdx_x_1 // 7 + 8) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3696] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 6) % 9 and (threadIdx_x_1 // 7 + 6) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3696) // 63 * 49 + (threadIdx_x_1 // 7 + 6) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3808] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 4) % 9 and (threadIdx_x_1 // 7 + 4) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3808) // 63 * 49 + (threadIdx_x_1 // 7 + 4) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- with T.launch_thread(threadIdx_x_1, 112):
- pad_temp_shared_1[threadIdx_x_1 + 3920] = T.if_then_else(1 <= (threadIdx_x_1 // 7 + 2) % 9 and (threadIdx_x_1 // 7 + 2) % 9 < 8 and 1 <= rx_outer_outer + threadIdx_x_1 % 7 and rx_outer_outer + threadIdx_x_1 % 7 < 8, data_1[cse_var_2 + (threadIdx_x_1 + 3920) // 63 * 49 + (threadIdx_x_1 // 7 + 2) % 9 * 7 + rx_outer_outer + threadIdx_x_1 % 7 - 8], T.float32(0))
- threadIdx_x_2 = T.env_thread("threadIdx.x")
- kernel_shared_1 = T.Buffer((6144,), 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 + cse_var_1 + threadIdx_x_2 * 3 + 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) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + 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 + 224) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 336] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 336) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + 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 + 448) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 560] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 560) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + 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 + 672) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 784] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 784) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 896] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 896) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1008] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1008) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1120] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1120) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1232] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1232) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1344] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 32256]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1456] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1456) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1568] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1568) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1680] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1680) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1792] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1792) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 1904] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 1904) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2016] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2016) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2128] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2128) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2240] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2240) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2352] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2352) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2464] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2464) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2576] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2576) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2688] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 64512]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2800] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2800) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 2912] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 2912) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3024] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3024) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3136] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3136) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3248] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3248) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3360] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3360) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3472] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3472) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3584] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3584) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3696] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3696) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3808] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3808) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 3920] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 3920) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4032] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96768]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4144] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4144) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4256] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4256) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4368] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4368) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4480] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4480) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4592] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4592) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4704] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4704) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 96) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4816] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4816) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 48]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 4928] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 4928) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 128) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5040] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5040) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 144]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5152] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5152) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 160) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5264] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5264) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 240]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5376] = kernel_1[blockIdx_x * 147456 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 129024]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5488] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5488) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 112) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5600] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5600) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 96]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5712] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5712) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 144) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5824] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5824) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 192]
- with T.launch_thread(threadIdx_x_2, 112):
- kernel_shared_1[threadIdx_x_2 + 5936] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 5936) // 192 * 4608 + cse_var_1 + (threadIdx_x_2 + 176) % 192 * 3 + rx_outer_outer]
- with T.launch_thread(threadIdx_x_2, 112):
- if T.likely(threadIdx_x_2 < 96):
- kernel_shared_1[threadIdx_x_2 + 6048] = kernel_1[blockIdx_x * 147456 + (threadIdx_x_2 + 6048) // 192 * 4608 + cse_var_1 + threadIdx_x_2 * 3 + rx_outer_outer + 288]
- for rc_outer_inner in range(64):
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3072]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 1]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 7] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3073]
- conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 56] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 2]
- conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 14] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 21] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 28] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 35] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 42] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 49] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[rc_outer_inner * 63 + threadIdx_x % 7 + 56] * kernel_shared_1[threadIdx_x // 7 * 192 + rc_outer_inner * 3 + 3074]
- compute_1 = T.Buffer((25088,), data=compute.data)
- bias_1 = T.Buffer((512,), data=bias.data)
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7] = T.max(conv2d_nchw_1[0] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 7] = T.max(conv2d_nchw_1[1] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 14] = T.max(conv2d_nchw_1[2] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 21] = T.max(conv2d_nchw_1[3] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 28] = T.max(conv2d_nchw_1[4] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 35] = T.max(conv2d_nchw_1[5] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 42] = T.max(conv2d_nchw_1[6] + bias_1[blockIdx_x * 32 + threadIdx_x // 7], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 784] = T.max(conv2d_nchw_1[7] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 791] = T.max(conv2d_nchw_1[8] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 798] = T.max(conv2d_nchw_1[9] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 805] = T.max(conv2d_nchw_1[10] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 812] = T.max(conv2d_nchw_1[11] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 819] = T.max(conv2d_nchw_1[12] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
- compute_1[blockIdx_x * 1568 + threadIdx_x // 7 * 49 + threadIdx_x % 7 + 826] = T.max(conv2d_nchw_1[13] + bias_1[blockIdx_x * 32 + threadIdx_x // 7 + 16], T.float32(0))
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 64) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 64) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 128) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 128) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 192] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 36864]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 256) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 256) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 320) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 320) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 384] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 73728]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 448) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 448) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 512) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 512) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 576] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 110592]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 640) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 640) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 704) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 704) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 768] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 147456]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 832) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 832) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 896) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 896) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 960] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 184320]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1024) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1024) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1088) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1088) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1152] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 221184]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1216) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1216) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1280) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1280) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1344] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 258048]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1408) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1408) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1472) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1472) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1536] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 294912]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1600) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1600) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1664) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1664) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1728] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 331776]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1792) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1792) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1856) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1856) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 1920] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 368640]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 1984) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 1984) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2048) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2048) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2112] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 405504]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2176) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2176) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2240) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2240) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2304] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 442368]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2368) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2368) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2432) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2432) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2496] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 479232]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2560) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2560) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2624) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2624) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2688] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 516096]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2752) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2752) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2816) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2816) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[threadIdx_x_1 + 2880] = kernel_1[blockIdx_x // 7 * 589824 + threadIdx_x_1 // 24 * 4608 + cse_var_2 + threadIdx_x_1 % 24 // 3 * 9 + cse_var_1 + threadIdx_x_1 % 3 + 552960]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 2944) // 24 * 24 + (threadIdx_x_1 + 16) % 24 // 3 * 3 + (threadIdx_x_1 + 1) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 2944) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 16) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 1) % 3]
+ with T.launch_thread(threadIdx_x_1, 64):
+ kernel_shared_1[(threadIdx_x_1 + 3008) // 24 * 24 + (threadIdx_x_1 + 8) % 24 // 3 * 3 + (threadIdx_x_1 + 2) % 3] = kernel_1[blockIdx_x // 7 * 589824 + (threadIdx_x_1 + 3008) // 24 * 4608 + cse_var_2 + (threadIdx_x_1 + 8) % 24 // 3 * 9 + cse_var_1 + (threadIdx_x_1 + 2) % 3]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 3]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[0] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[9] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 24]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 27]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 1]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 4]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[1] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[10] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 25]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 28]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 2]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 5]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[2] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[11] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[3] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[12] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[4] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[13] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[5] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[14] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[6] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[15] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[7] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[16] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[8] * kernel_shared_1[threadIdx_x * 48 + 26]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[17] * kernel_shared_1[threadIdx_x * 48 + 29]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 6]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 9]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[18] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[27] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 30]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 33]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 7]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 10]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[19] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[28] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 31]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 34]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 8]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 11]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[20] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[29] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[21] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[30] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[22] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[31] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[23] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[32] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[24] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[33] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[25] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[34] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[26] * kernel_shared_1[threadIdx_x * 48 + 32]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[35] * kernel_shared_1[threadIdx_x * 48 + 35]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 12]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 15]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[36] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[45] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 36]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 39]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 13]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 16]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[37] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[46] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 37]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 40]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 14]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 17]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[38] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[47] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[39] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[48] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[40] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[49] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[41] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[50] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[42] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[51] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[43] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[52] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[44] * kernel_shared_1[threadIdx_x * 48 + 38]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[53] * kernel_shared_1[threadIdx_x * 48 + 41]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 18]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 21]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[54] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[63] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 42]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 45]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 19]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 22]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[55] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[64] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 43]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 46]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[0] = conv2d_nchw_1[0] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[1] = conv2d_nchw_1[1] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[2] = conv2d_nchw_1[2] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[3] = conv2d_nchw_1[3] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[4] = conv2d_nchw_1[4] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[5] = conv2d_nchw_1[5] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 20]
+ conv2d_nchw_1[6] = conv2d_nchw_1[6] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 23]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[56] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[7] = conv2d_nchw_1[7] + pad_temp_shared_1[65] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[57] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[8] = conv2d_nchw_1[8] + pad_temp_shared_1[66] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[58] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[9] = conv2d_nchw_1[9] + pad_temp_shared_1[67] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[59] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[10] = conv2d_nchw_1[10] + pad_temp_shared_1[68] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[60] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[11] = conv2d_nchw_1[11] + pad_temp_shared_1[69] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[61] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[12] = conv2d_nchw_1[12] + pad_temp_shared_1[70] * kernel_shared_1[threadIdx_x * 48 + 47]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[62] * kernel_shared_1[threadIdx_x * 48 + 44]
+ conv2d_nchw_1[13] = conv2d_nchw_1[13] + pad_temp_shared_1[71] * kernel_shared_1[threadIdx_x * 48 + 47]
+ for i1_inner, i3_inner in T.grid(2, 7):
+ compute_1 = T.Buffer((25088,), data=compute.data)
+ bias_1 = T.Buffer((512,), data=bias.data)
+ compute_1[blockIdx_x // 7 * 6272 + threadIdx_x * 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>
@@ -814,7 +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: 0.271 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.354 ms
</pre></div>
</div>
</div>
@@ -844,35 +1046,35 @@ conv2d_nchw_nn_o_o_i, conv2d_nchw_nn_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o
conv2d_nchw_nn_o_o_o_i, conv2d_nchw_nn_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_i, factor=1)
conv2d_nchw_nn_o_o_o_o, conv2d_nchw_nn_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_nn_o_o_o_i, factor=1)
conv2d_nchw_ff_o_i, conv2d_nchw_ff_i = s[conv2d_nchw].split(conv2d_nchw_ff, factor=1)
-conv2d_nchw_ff_o_o_i, conv2d_nchw_ff_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_i, factor=1)
-conv2d_nchw_ff_o_o_o_i, conv2d_nchw_ff_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_i, factor=16)
-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=2)
+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=1)
-conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=7)
+conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
-conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=1)
-conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=7)
+conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
+conv2d_nchw_xx_o_o_o_i, conv2d_nchw_xx_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_i, factor=1)
conv2d_nchw_xx_o_o_o_o, conv2d_nchw_xx_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_o_o_i, factor=1)
-conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=1)
-conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=64)
-conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=3)
+conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=2)
+conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=4)
+conv2d_nchw_ry_o_i, conv2d_nchw_ry_i = s[conv2d_nchw].split(conv2d_nchw_ry, factor=1)
conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=1)
conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=1)
-conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
+conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=3)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2d_nc [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=1)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=16)
-compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=2)
+compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
+compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=64)
+compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
-compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=7)
-compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=1)
-compute_i3_o_o_i, compute_i3_o_i = s[compute].split(compute_i3_o_i, factor=7)
+compute_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)
compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
s[compute].reorder(compute_i0_o_o_o, compute_i1_o_o_o, compute_i2_o_o_o, compute_i3_o_o_o, compute_i0_o_o_i, compute_i1_o_o_i, compute_i2_o_o_i, compute_i3_o_o_i, compute_i0_o_i, compute_i1_o_i, compute_i2_o_i, compute_i3_o_i, compute_i0_i, compute_i1_i, compute_i2_i, compute_i3_i)
s[conv2d_nchw].compute_at(s[compute], compute_i3_o_i)
@@ -892,14 +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", 64)
+s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "auto_unroll_max_step", 512)
s[conv2d_nchw].pragma(conv2d_nchw_nn_o_o_o_o, "unroll_explicit", True)
CUDA source code:
@@ -917,10 +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[4032];
- __shared__ float kernel_shared[6144];
+ __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;
@@ -935,163 +1137,413 @@ 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 < 8; ++rc_outer_outer) {
- for (int rx_outer_outer = 0; rx_outer_outer < 3; ++rx_outer_outer) {
+ for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
+ for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
__syncthreads();
- pad_temp_shared[((int)threadIdx.x)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 112)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 112) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 224) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 336)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 336) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 448)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 448) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 560)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 560) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 672)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 672) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 784)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 784) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 896)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 896) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1008)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 776)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1120)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1120) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1232)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1232) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1344)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1344) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1456)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1456) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1568)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1568) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1680)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1680) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1792)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1792) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 1904)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 1904) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2016)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 1560)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2128)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2128) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2240)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2240) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2352)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2352) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2464)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2464) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2576)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2576) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2688)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2688) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2800)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2800) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 2912)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 2912) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3024)] = (((((7 <= (((int)threadIdx.x) % 63)) && ((((int)threadIdx.x) % 63) < 56)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[(((((rc_outer_outer * 3136) + ((((int)threadIdx.x) / 63) * 49)) + rx_outer_outer) + (((int)threadIdx.x) % 63)) + 2344)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3136)] = (((((1 <= (((((int)threadIdx.x) / 7) + 7) % 9)) && ((((((int)threadIdx.x) / 7) + 7) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3136) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 7) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3248)] = (((((1 <= (((((int)threadIdx.x) / 7) + 5) % 9)) && ((((((int)threadIdx.x) / 7) + 5) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3248) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 5) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3360)] = (((((1 <= (((((int)threadIdx.x) / 7) + 3) % 9)) && ((((((int)threadIdx.x) / 7) + 3) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3360) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 3) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3472)] = (((((1 <= (((((int)threadIdx.x) / 7) + 1) % 9)) && ((((((int)threadIdx.x) / 7) + 1) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3472) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 1) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3584)] = (((((1 <= (((((int)threadIdx.x) / 7) + 8) % 9)) && ((((((int)threadIdx.x) / 7) + 8) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3584) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 8) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3696)] = (((((1 <= (((((int)threadIdx.x) / 7) + 6) % 9)) && ((((((int)threadIdx.x) / 7) + 6) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3696) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 6) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3808)] = (((((1 <= (((((int)threadIdx.x) / 7) + 4) % 9)) && ((((((int)threadIdx.x) / 7) + 4) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3808) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 4) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- pad_temp_shared[(((int)threadIdx.x) + 3920)] = (((((1 <= (((((int)threadIdx.x) / 7) + 2) % 9)) && ((((((int)threadIdx.x) / 7) + 2) % 9) < 8)) && (1 <= (rx_outer_outer + (((int)threadIdx.x) % 7)))) && ((rx_outer_outer + (((int)threadIdx.x) % 7)) < 8)) ? data[((((((rc_outer_outer * 3136) + (((((int)threadIdx.x) + 3920) / 63) * 49)) + ((((((int)threadIdx.x) / 7) + 2) % 9) * 7)) + rx_outer_outer) + (((int)threadIdx.x) % 7)) - 8)] : 0.000000e+00f);
- kernel_shared[((int)threadIdx.x)] = kernel[((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 112)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 112) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 224)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 224) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 336)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 336) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 448) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 560)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 560) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 672)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 672) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 784)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 784) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 896) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1008)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1008) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 1120)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1120) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1232)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1232) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 32256)];
- kernel_shared[(((int)threadIdx.x) + 1456)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1456) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1568)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1568) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 1680)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1680) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1792) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 1904)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 1904) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2016)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2016) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2128)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2128) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2240) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2352)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2352) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 2464)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2464) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2576)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2576) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 64512)];
- kernel_shared[(((int)threadIdx.x) + 2800)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2800) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 2912)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 2912) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 3024)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3024) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3136)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3136) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 3248)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3248) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3360)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3360) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3472)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3472) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 3584)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3584) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3696)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3696) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 3808)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3808) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 3920)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 3920) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 4032)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96768)];
- kernel_shared[(((int)threadIdx.x) + 4144)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4144) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4256)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4256) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 4368)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4368) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4480)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4480) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 4592)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4592) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4704)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4704) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 96) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 4816)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4816) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 48)];
- kernel_shared[(((int)threadIdx.x) + 4928)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 4928) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 128) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5040)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5040) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 144)];
- kernel_shared[(((int)threadIdx.x) + 5152)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5152) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 160) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5264)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5264) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 240)];
- kernel_shared[(((int)threadIdx.x) + 5376)] = kernel[(((((((int)blockIdx.x) * 147456) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 129024)];
- kernel_shared[(((int)threadIdx.x) + 5488)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5488) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 112) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5600)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5600) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 96)];
- kernel_shared[(((int)threadIdx.x) + 5712)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5712) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 144) % 192) * 3)) + rx_outer_outer)];
- kernel_shared[(((int)threadIdx.x) + 5824)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5824) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 192)];
- kernel_shared[(((int)threadIdx.x) + 5936)] = kernel[(((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 5936) / 192) * 4608)) + (rc_outer_outer * 576)) + (((((int)threadIdx.x) + 176) % 192) * 3)) + rx_outer_outer)];
- if (((int)threadIdx.x) < 96) {
- kernel_shared[(((int)threadIdx.x) + 6048)] = kernel[((((((((int)blockIdx.x) * 147456) + (((((int)threadIdx.x) + 6048) / 192) * 4608)) + (rc_outer_outer * 576)) + (((int)threadIdx.x) * 3)) + rx_outer_outer) + 288)];
+ 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);
}
- __syncthreads();
- for (int rc_outer_inner = 0; rc_outer_inner < 64; ++rc_outer_inner) {
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[(((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3))]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[((rc_outer_inner * 63) + (((int)threadIdx.x) % 7))] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3072)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 1)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 7)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3073)]));
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 2)]));
- conv2d_nchw[7] = (conv2d_nchw[7] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 14)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[8] = (conv2d_nchw[8] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 21)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[9] = (conv2d_nchw[9] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 28)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[10] = (conv2d_nchw[10] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 35)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[11] = (conv2d_nchw[11] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 42)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[12] = (conv2d_nchw[12] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 49)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
- conv2d_nchw[13] = (conv2d_nchw[13] + (pad_temp_shared[(((rc_outer_inner * 63) + (((int)threadIdx.x) % 7)) + 56)] * kernel_shared[((((((int)threadIdx.x) / 7) * 192) + (rc_outer_inner * 3)) + 3074)]));
+ 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) / 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);
}
}
- compute[(((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7))] = max((conv2d_nchw[0] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 7)] = max((conv2d_nchw[1] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 14)] = max((conv2d_nchw[2] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 21)] = max((conv2d_nchw[3] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 28)] = max((conv2d_nchw[4] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 35)] = max((conv2d_nchw[5] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 42)] = max((conv2d_nchw[6] + bias[((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7))]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 784)] = max((conv2d_nchw[7] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 791)] = max((conv2d_nchw[8] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 798)] = max((conv2d_nchw[9] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 805)] = max((conv2d_nchw[10] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 812)] = max((conv2d_nchw[11] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 819)] = max((conv2d_nchw[12] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
- compute[((((((int)blockIdx.x) * 1568) + ((((int)threadIdx.x) / 7) * 49)) + (((int)threadIdx.x) % 7)) + 826)] = max((conv2d_nchw[13] + bias[(((((int)blockIdx.x) * 32) + (((int)threadIdx.x) / 7)) + 16)]), 0.000000e+00f);
}
</pre></div>
</div>
@@ -1125,7 +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 14.904 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 6 minutes 8.709 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 187ff9f5fe..a957822b33 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -921,7 +921,7 @@ so we can read the log file and load the best schedules.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 7.8994 7.9014 7.9033 7.8934 0.0043
+ 7.8820 7.8852 7.8878 7.8730 0.0064
</pre></div>
</div>
</div>
@@ -943,7 +943,7 @@ to learn how to use the RPC Tracker and RPC Server.
To use the RPC Tracker in auto-scheduler, replace the runner in <code class="code docutils literal notranslate"><span class="pre">TuningOptions</span></code>
with <a class="reference internal" href="../../reference/api/python/auto_scheduler.html#tvm.auto_scheduler.RPCRunner" title="tvm.auto_scheduler.RPCRunner"><code class="xref any py py-class docutils literal notranslate"><span class="pre">auto_scheduler.RPCRunner</span></code></a>.</p></li>
</ol>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 7.384 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 7.519 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 7f023acd77..3aba194354 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)
- 758.6970 758.7265 761.0138 756.3507 1.9038
+ 762.1932 762.1844 762.2230 762.1721 0.0217
</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 41.717 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 42.014 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 3a5ea68a81..25af6ffe79 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -637,74 +637,27 @@ class Module:
@T.prim_func
def main(placeholder: T.Buffer((128, 256), "float32"), placeholder_1: T.Buffer((4916, 16, 1), "float32"), placeholder_2: T.Buffer((4916,), "int32"), placeholder_3: T.Buffer((33,), "int32"), placeholder_4: T.Buffer((128, 512), "float32"), compute: T.Buffer((128, 512), "float32")):
T.func_attr({"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True})
- for i0_outer_i1_outer_fused in T.parallel(32):
- compute_1 = T.allocate([2048], "float32", "global")
- compute_2 = T.Buffer((2048,), data=compute_1)
- for nb_j_inner in range(2):
- for i_inner_init in range(64):
- cse_var_1: T.int32 = i_inner_init * 32 + nb_j_inner * 16
- compute_2[cse_var_1] = T.float32(0)
- compute_2[cse_var_1 + 1] = T.float32(0)
- compute_2[cse_var_1 + 2] = T.float32(0)
- compute_2[cse_var_1 + 3] = T.float32(0)
- compute_2[cse_var_1 + 4] = T.float32(0)
- compute_2[cse_var_1 + 5] = T.float32(0)
- compute_2[cse_var_1 + 6] = T.float32(0)
- compute_2[cse_var_1 + 7] = T.float32(0)
- compute_2[cse_var_1 + 8] = T.float32(0)
- compute_2[cse_var_1 + 9] = T.float32(0)
- compute_2[cse_var_1 + 10] = T.float32(0)
- compute_2[cse_var_1 + 11] = T.float32(0)
- compute_2[cse_var_1 + 12] = T.float32(0)
- compute_2[cse_var_1 + 13] = T.float32(0)
- compute_2[cse_var_1 + 14] = T.float32(0)
- compute_2[cse_var_1 + 15] = T.float32(0)
- for elem_idx, i_inner in T.grid(T.Let(placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2], where={cse_var_2: i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner}), 64):
- cse_var_2 = T.int32()
+ for i0_outer_i1_outer_fused in T.parallel(128):
+ compute_1 = T.allocate([512], "float32", "global")
+ compute_2 = T.Buffer((512,), data=compute_1)
+ for i_outer_inner in range(64):
+ for i_inner_init, j_init in T.grid(2, 4):
+ compute_2[i_outer_inner * 8 + i_inner_init * 4 + 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: i0_outer_i1_outer_fused // 4}), 2, 4):
+ cse_var_1 = T.int32()
placeholder_5 = T.Buffer((33,), "int32", data=placeholder_3.data)
- cse_var_21: T.int32 = elem_idx * 16
- cse_var_20: T.int32 = i_inner * 32 + nb_j_inner * 16
- cse_var_19: T.int32 = i0_outer_i1_outer_fused % 16 * 2 + nb_j_inner
- cse_var_18: T.int32 = i0_outer_i1_outer_fused // 16 * 16384 + i_inner * 256
- cse_var_17: T.int32 = cse_var_20 + 9
- cse_var_16: T.int32 = cse_var_20 + 8
- cse_var_15: T.int32 = cse_var_20 + 7
- cse_var_14: T.int32 = cse_var_20 + 6
- cse_var_13: T.int32 = cse_var_20 + 5
- cse_var_12: T.int32 = cse_var_20 + 4
- cse_var_11: T.int32 = cse_var_20 + 3
- cse_var_10: T.int32 = cse_var_20 + 2
- cse_var_9: T.int32 = cse_var_20 + 15
- cse_var_8: T.int32 = cse_var_20 + 14
- cse_var_7: T.int32 = cse_var_20 + 13
- cse_var_6: T.int32 = cse_var_20 + 12
- cse_var_5: T.int32 = cse_var_20 + 11
- cse_var_4: T.int32 = cse_var_20 + 10
- cse_var_3: T.int32 = cse_var_20 + 1
- placeholder_6 = T.Buffer((78656,), data=placeholder_1.data)
- placeholder_7 = T.Buffer((32768,), data=placeholder.data)
- placeholder_8 = T.Buffer((4916,), "int32", data=placeholder_2.data)
- compute_2[cse_var_20] = compute_2[cse_var_20] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 1] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_10] = compute_2[cse_var_10] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 2] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_11] = compute_2[cse_var_11] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 3] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_12] = compute_2[cse_var_12] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 4] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_13] = compute_2[cse_var_13] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 5] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_14] = compute_2[cse_var_14] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 6] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_15] = compute_2[cse_var_15] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 7] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_16] = compute_2[cse_var_16] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 8] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_17] = compute_2[cse_var_17] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 9] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_4] = compute_2[cse_var_4] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 10] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_5] = compute_2[cse_var_5] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 11] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_6] = compute_2[cse_var_6] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 12] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_7] = compute_2[cse_var_7] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 13] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_8] = compute_2[cse_var_8] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 14] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- compute_2[cse_var_9] = compute_2[cse_var_9] + placeholder_6[placeholder_5[cse_var_19] * 16 + cse_var_21 + 15] * T.max(placeholder_7[cse_var_18 + placeholder_8[placeholder_5[cse_var_19] + elem_idx]], T.float32(0))
- for i0_inner, i1_inner in T.grid(64, 32):
- cse_var_22: T.int32 = i0_outer_i1_outer_fused // 16 * 32768 + i0_inner * 512 + i0_outer_i1_outer_fused % 16 * 32 + i1_inner
+ cse_var_2: T.int32 = i0_outer_i1_outer_fused // 4
+ if T.likely(elem_idx < placeholder_5[cse_var_2 + 1] - placeholder_5[cse_var_2]):
+ 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)
+ cse_var_3: T.int32 = i_outer_inner * 8 + i_inner * 4 + j
+ compute_2[cse_var_3] = compute_2[cse_var_3] + placeholder_6[placeholder_5[cse_var_2] * 16 + elem_idx * 16 + i0_outer_i1_outer_fused % 4 * 4 + j] * T.max(placeholder_7[i_outer_inner * 512 + i_inner * 256 + placeholder_8[placeholder_5[cse_var_2] + elem_idx]], T.float32(0))
+ for i0_inner in range(128):
+ cse_var_4: T.int32 = i0_inner * 512 + i0_outer_i1_outer_fused * 4
compute_3 = T.Buffer((65536,), data=compute.data)
placeholder_5 = T.Buffer((65536,), data=placeholder_4.data)
- compute_3[cse_var_22] = T.max(compute_2[i0_inner * 32 + i1_inner] + placeholder_5[cse_var_22], T.float32(0))
+ compute_3[cse_var_4:cse_var_4 + 4] = T.max(compute_2[i0_inner * 4:i0_inner * 4 + 4] + placeholder_5[cse_var_4:cse_var_4 + 4], T.Broadcast(T.float32(0), 4))
</pre></div>
</div>
</div>
@@ -738,7 +691,7 @@ 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: 1.850 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 3.125 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 2192cd79e1..438c776b6f 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>00:31.133</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:35.099</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 84%" />
@@ -354,11 +354,11 @@
</colgroup>
<tbody>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_conv2d_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-conv2d-cuda-py"><span class="std std-ref">Tuning High Performance Convolution on NVIDIA GPUs</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_conv2d_cuda.py</span></code>)</p></td>
-<td><p>00:31.100</p></td>
+<td><p>00:35.060</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="tune_relay_x86.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-x86-py"><span class="std std-ref">Auto-tuning a Convolutional Network for x86 CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_x86.py</span></code>)</p></td>
-<td><p>00:00.020</p></td>
+<td><p>00:00.025</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tune_relay_cuda.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-cuda-py"><span class="std std-ref">Auto-tuning a Convolutional Network for NVIDIA GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_cuda.py</span></code>)</p></td>
diff --git a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
index bab0759e2f..66b59a95ba 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -695,8 +695,9 @@ 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, 64, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9074474
-No: 2 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 256, 2, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5959158
+No: 2 GFLOPS: 12.04/12.04 result: MeasureResult(costs=(0.01922388833333333,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.596773386001587, timestamp=1678274198.0724711) [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3687096
+No: 3 GFLOPS: 0.00/12.04 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
@@ -818,8 +819,8 @@ 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, 16, 8, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4858385
-No: 3 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 64, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1556414
+No: 4 GFLOPS: 0.00/12.04 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
@@ -941,8 +942,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 1, 64]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 32, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4084063
-No: 4 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 32, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7305670
+No: 5 GFLOPS: 0.00/12.04 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
@@ -1064,8 +1065,8 @@ 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, 128, 1, 4]), ('tile_y', [-1, 1, 7, 1]), ('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', 1500), ('unroll_explicit', 1)],None,10386307
-No: 5 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4246113
+No: 6 GFLOPS: 0.00/12.04 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
@@ -1187,8 +1188,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 2, 32]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8031292
-No: 6 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 2, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 16, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8048670
+No: 7 GFLOPS: 0.00/12.04 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
@@ -1310,8 +1311,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 64, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,5852536
-No: 7 GFLOPS: 0.00/0.00 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 256]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1348496
+No: 8 GFLOPS: 0.00/12.04 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
@@ -1433,10 +1434,8 @@ 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, 16, 2, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6032634
-No: 8 GFLOPS: 5.46/5.46 result: MeasureResult(costs=(0.04242529675,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.1453189849853516, timestamp=1678271999.5114067) [('tile_f', [-1, 2, 1, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 64, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4702857
-No: 9 GFLOPS: 234.60/234.60 result: MeasureResult(costs=(0.0009867967423312883,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5571238994598389, timestamp=1678272001.4171991) [('tile_f', [-1, 4, 16, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4714196
-No: 10 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 8, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4195579
+No: 9 GFLOPS: 0.00/12.04 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
@@ -1558,8 +1557,8 @@ 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, 128, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,677869
-No: 11 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 128, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 128, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8579146
+No: 10 GFLOPS: 0.00/12.04 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
@@ -1681,9 +1680,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 2, 2]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9650809
-No: 12 GFLOPS: 1.28/234.60 result: MeasureResult(costs=(0.18103141775,), error_no=MeasureErrorNo.NO_ERROR, all_cost=3.474294662475586, timestamp=1678272004.214991) [('tile_f', [-1, 4, 2, 64]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 1, 2]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3329246
-No: 13 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 32, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 2]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7210152
+No: 11 GFLOPS: 0.00/12.04 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
@@ -1805,8 +1803,8 @@ 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, 1, 4]), ('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, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4983106
-No: 14 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 2, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1499693
+No: 12 GFLOPS: 0.00/12.04 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
@@ -1928,8 +1926,9 @@ 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, 16, 4, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8529993
-No: 15 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 32, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,1891602
+No: 13 GFLOPS: 7.71/12.04 result: MeasureResult(costs=(0.03004289625,), error_no=MeasureErrorNo.NO_ERROR, all_cost=2.9099907875061035, timestamp=1678274204.5159023) [('tile_f', [-1, 32, 4, 2]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3581237
+No: 14 GFLOPS: 0.00/12.04 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
@@ -2051,8 +2050,8 @@ 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, 1, 256]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 2, 32]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4212776
-No: 16 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+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, 7, 1, 1]), ('tile_rc', [-1, 8, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6154661
+No: 15 GFLOPS: 0.00/12.04 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
@@ -2174,8 +2173,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 64, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 64, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8609747
-No: 17 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 64, 4, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1175758
+No: 16 GFLOPS: 0.00/12.04 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
@@ -2297,9 +2296,8 @@ Traceback (most recent call last):
File "tvm/_ffi/_cython/./packed_func.pxi", line 56, in tvm._ffi._cy3.core.tvm_callback
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 875, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
-tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 64, 8]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,3078403
-No: 18 GFLOPS: 171.84/234.60 result: MeasureResult(costs=(0.0013472277733333335,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.2511568069458008, timestamp=1678272005.8511157) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 8]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1457959
-No: 19 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2672315
+No: 17 GFLOPS: 0.00/12.04 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
@@ -2421,8 +2419,9 @@ 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, 8]), ('tile_y', [-1, 1, 1, 7]), ('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,9656180
-No: 20 GFLOPS: 0.00/234.60 result: Traceback (most recent call last):
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 128, 1]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3882829
+No: 18 GFLOPS: 9.93/12.04 result: MeasureResult(costs=(0.023322064666666666,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.7776870727539062, timestamp=1678274206.5398247) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2281400
+No: 19 GFLOPS: 0.00/12.04 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
@@ -2544,7 +2543,130 @@ 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, 4, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 2, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10230560
+tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 16, 2]), ('tile_y', [-1, 1, 7, 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', 512), ('unroll_explicit', 1)],None,7913488
+No: 20 GFLOPS: 0.00/12.04 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, 1, 32, 2]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,8245030
</pre></div>
</div>
<p>Finally we can inspect the best config from log file, check correctness,
@@ -2583,9 +2705,9 @@ 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, 4, 16, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 4]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4714196
+[('tile_f', [-1, 2, 4, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3687096
Finish loading 20 records
-Time cost of this operator: 0.001362
+Time cost of this operator: 0.011204
</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 92be8012ba..a73a025992 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -648,10 +648,10 @@ the tuned operator.</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build without Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 324.8 98.785 (1, 2, 10, 10, 3) 2 1 [324.8]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.042 0.925 (1, 6, 10, 10) 1 1 [3.042]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.953 0.29 (1, 1, 10, 10, 3) 1 1 [0.953]
-Total_time - 328.795 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 315.0 98.742 (1, 2, 10, 10, 3) 2 1 [315.0]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.057 0.958 (1, 6, 10, 10) 1 1 [3.057]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.958 0.3 (1, 1, 10, 10, 3) 1 1 [0.958]
+Total_time - 319.014 - - - - -
</pre></div>
</div>
</div>
@@ -703,13 +703,13 @@ Total_time -
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs Measurements(us)
--------- --- -------- ------- ----- ------ ------- ----------------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 100.3 97.348 (1, 6, 10, 10, 1) 2 1 [100.3]
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.757 1.705 (1, 6, 10, 10) 1 1 [1.757]
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.976 0.947 (1, 1, 10, 10, 3) 1 1 [0.976]
-Total_time - 103.033 - - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 108.1 97.604 (1, 6, 10, 10, 1) 2 1 [108.1]
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.795 1.621 (1, 6, 10, 10) 1 1 [1.795]
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.858 0.775 (1, 3, 10, 10, 1) 1 1 [0.858]
+Total_time - 110.754 - - - - -
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 22.752 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 20.648 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 91f57e19a4..83227b3cde 100644
--- a/docs/how_to/work_with_microtvm/micro_pytorch.html
+++ b/docs/how_to/work_with_microtvm/micro_pytorch.html
@@ -459,7 +459,7 @@ download a cat image and preprocess it to use as the model input.</p>
Downloading: "https://download.pytorch.org/models/quantized/mobilenet_v2_qnnpack_37f702c5.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2_qnnpack_37f702c5.pth
0%| | 0.00/3.42M [00:00<?, ?B/s]
-100%|##########| 3.42M/3.42M [00:00<00:00, 105MB/s]
+100%|##########| 3.42M/3.42M [00:00<00:00, 124MB/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.
@@ -585,7 +585,7 @@ 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 17.022 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 19.092 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 9fe0eaeb3b..c74c45f853 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/tmpmkfhz42r/images/random'
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>'/tmp/tmp96e_gj6m/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="[0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [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/tmpmkfhz42r/images/target contains 8144 images
-/tmp/tmpmkfhz42r/images/random contains 5000 images
+<img src="../../_images/sphx_glr_micro_train_001.png" srcset="../../_images/sphx_glr_micro_train_001.png" alt="[1.0, 0.0], [0.0, 1.0], [0.0, 1.0], [0.0, 1.0], [1.0, 0.0], [1.0, 0.0], [1.0, 0.0], [0.0, 1.0], [1.0, 0.0], [0.0, 1.0]" class = "sphx-glr-single-img"/><div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>/tmp/tmp96e_gj6m/images/target contains 8144 images
+/tmp/tmp96e_gj6m/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 - 48s - loss: 0.2043 - accuracy: 0.9280 - val_loss: 0.1360 - val_accuracy: 0.9562 - 48s/epoch - 145ms/step
+328/328 - 48s - loss: 0.2257 - accuracy: 0.9211 - val_loss: 0.1038 - val_accuracy: 0.9622 - 48s/epoch - 148ms/step
Epoch 2/3
-328/328 - 44s - loss: 0.1009 - accuracy: 0.9624 - val_loss: 0.1234 - val_accuracy: 0.9585 - 44s/epoch - 134ms/step
+328/328 - 43s - loss: 0.1021 - accuracy: 0.9618 - val_loss: 0.0971 - val_accuracy: 0.9649 - 43s/epoch - 132ms/step
Epoch 3/3
-328/328 - 44s - loss: 0.0591 - accuracy: 0.9779 - val_loss: 0.1900 - val_accuracy: 0.9415 - 44s/epoch - 133ms/step
+328/328 - 43s - loss: 0.0728 - accuracy: 0.9731 - val_loss: 0.1309 - val_accuracy: 0.9558 - 43s/epoch - 132ms/step
-<keras.callbacks.History object at 0x7f7bc84ed950>
+<keras.callbacks.History object at 0x7f149e420b90>
</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 24.406 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 4 minutes 26.784 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 418dee3316..2b2b6c51bb 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>07:21.426</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>07:24.526</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 82%" />
@@ -354,30 +354,30 @@
</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>04:24.406</p></td>
+<td><p>04:26.784</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>01:22.752</p></td>
+<td><p>01:20.648</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>01:17.022</p></td>
+<td><p>01:19.092</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>00:10.087</p></td>
+<td><p>00:10.165</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">2. microTVM TFLite Tutorial</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></td>
-<td><p>00:07.159</p></td>
+<td><p>00:07.838</p></td>
<td><p>0.0 MB</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">7. Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU with CMSIS-NN</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></td>
+<tr class="row-even"><td><p><a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">1. microTVM CLI Tool</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></td>
<td><p>00:00.000</p></td>
<td><p>0.0 MB</p></td>
</tr>
-<tr class="row-odd"><td><p><a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">1. microTVM CLI Tool</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></td>
+<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>
<td><p>00:00.000</p></td>
<td><p>0.0 MB</p></td>
</tr>
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 8017f42cc9..32e6f31d1f 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -345,7 +345,7 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-work-with-relay-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:46.439</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:50.749</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>00:34.165</p></td>
+<td><p>00:36.555</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="using_external_lib.html#sphx-glr-how-to-work-with-relay-using-external-lib-py"><span class="std std-ref">Using External Libraries in Relay</span></a> (<code class="docutils literal notranslate"><span class="pre">using_external_lib.py</span></code>)</p></td>
-<td><p>00:10.412</p></td>
+<td><p>00:12.092</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>00:01.856</p></td>
+<td><p>00:02.095</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 11b070086a..7dc7cd2972 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 0x7f7a7598b200>
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span><function my_cuda_math_rule at 0x7f135015d200>
</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 1cf25eae77..f475a5e23a 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>00:06.702</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:07.645</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
<table class="docutils align-default">
<colgroup>
<col style="width: 83%" />
@@ -354,31 +354,31 @@
</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>00:04.170</p></td>
+<td><p>00:04.633</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="tensorize.html#sphx-glr-how-to-work-with-schedules-tensorize-py"><span class="std std-ref">Use Tensorize to Leverage Hardware Intrinsics</span></a> (<code class="docutils literal notranslate"><span class="pre">tensorize.py</span></code>)</p></td>
-<td><p>00:01.135</p></td>
+<td><p>00:01.587</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>00:00.589</p></td>
+<tr class="row-odd"><td><p><a class="reference internal" href="scan.html#sphx-glr-how-to-work-with-schedules-scan-py"><span class="std std-ref">Scan and Recurrent Kernel</span></a> (<code class="docutils literal notranslate"><span class="pre">scan.py</span></code>)</p></td>
+<td><p>00:00.594</p></td>
<td><p>0.0 MB</p></td>
</tr>
-<tr class="row-even"><td><p><a class="reference internal" href="scan.html#sphx-glr-how-to-work-with-schedules-scan-py"><span class="std std-ref">Scan and Recurrent Kernel</span></a> (<code class="docutils literal notranslate"><span class="pre">scan.py</span></code>)</p></td>
-<td><p>00:00.572</p></td>
+<tr class="row-even"><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>00:00.589</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="extern_op.html#sphx-glr-how-to-work-with-schedules-extern-op-py"><span class="std std-ref">External Tensor Functions</span></a> (<code class="docutils literal notranslate"><span class="pre">extern_op.py</span></code>)</p></td>
-<td><p>00:00.118</p></td>
+<td><p>00:00.117</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-even"><td><p><a class="reference internal" href="schedule_primitives.html#sphx-glr-how-to-work-with-schedules-schedule-primitives-py"><span class="std std-ref">Schedule Primitives in TVM</span></a> (<code class="docutils literal notranslate"><span class="pre">schedule_primitives.py</span></code>)</p></td>
-<td><p>00:00.053</p></td>
+<td><p>00:00.065</p></td>
<td><p>0.0 MB</p></td>
</tr>
<tr class="row-odd"><td><p><a class="reference internal" href="tedd.html#sphx-glr-how-to-work-with-schedules-tedd-py"><span class="std std-ref">Use Tensor Expression Debug Display (TEDD) for Visualization</span></a> (<code class="docutils literal notranslate"><span class="pre">tedd.py</span></code>)</p></td>
-<td><p>00:00.038</p></td>
+<td><p>00:00.033</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/reference/api/doxygen/block__scope_8h_source.html b/docs/reference/api/doxygen/block__scope_8h_source.html
index d5c6057902..0916ee953c 100644
--- a/docs/reference/api/doxygen/block__scope_8h_source.html
+++ b/docs/reference/api/doxygen/block__scope_8h_source.html
@@ -102,7 +102,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1tir_html_a82f5fe564edbe0dadbc4a56a82dfe12eaf3b1d6a59161fbe655adccdd10c6eda4"><div class="ttname"><a href="namespacetvm_1_1tir.html#a82f5fe564edbe0dadbc4a56a82dfe12eaf3b1d6a59161fbe655adccdd10c6eda4">tvm::tir::DepKind::kWAW</a></div></div>
<div class="ttc" id="classtvm_1_1tir_1_1StmtSRefNode_html_a1ce9920064b0c2ed42be0e9412fd3939"><div class="ttname"><a href="classtvm_1_1tir_1_1StmtSRefNode.html#a1ce9920064b0c2ed42be0e9412fd3939">tvm::tir::StmtSRefNode::seq_index</a></div><div class="ttdeci">int64_t seq_index</div><div class="ttdoc">If the statement the sref points to is an element of a SeqStmt in the AST, then seq_index is set to i...</div><div class="ttdef"><b>Definition:</b> block_scope.h:63</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1StmtSRefNode_html_ad4d0dab331c6e42c9197e5c0020099b5"><div class="ttname"><a href="classtvm_1_1tir_1_1StmtSRefNode.html#ad4d0dab331c6e42c9197e5c0020099b5">tvm::tir::StmtSRefNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> block_scope.h:65</div></div>
-<div class="ttc" id="namespacetvm_1_1tir_html_add7d0a6b1dd91f0c3c5dd2f4cf64358eaf324873e6195114a186db7f910559b2c"><div class="ttname"><a href="namespacetvm_1_1tir.html#add7d0a6b1dd91f0c3c5dd2f4cf64358eaf324873e6195114a186db7f910559b2c">tvm::tir::kOpaque</a></div><div class="ttdoc">IterVar is opaque,. </div><div class="ttdef"><b>Definition:</b> var.h:220</div></div>
+<div class="ttc" id="namespacetvm_1_1tir_html_add7d0a6b1dd91f0c3c5dd2f4cf64358eaf324873e6195114a186db7f910559b2c"><div class="ttname"><a href="namespacetvm_1_1tir.html#add7d0a6b1dd91f0c3c5dd2f4cf64358eaf324873e6195114a186db7f910559b2c">tvm::tir::kOpaque</a></div><div class="ttdoc">IterVar is opaque,. </div><div class="ttdef"><b>Definition:</b> var.h:227</div></div>
<div class="ttc" id="object_8h_html_a782d0de62fbf75736e29c1e79c22c7f1"><div class="ttname"><a href="object_8h.html#a782d0de62fbf75736e29c1e79c22c7f1">TVM_DEFINE_NOTNULLABLE_OBJECT_REF_METHODS</a></div><div class="ttdeci">#define TVM_DEFINE_NOTNULLABLE_OBJECT_REF_METHODS(TypeName, ParentType, ObjectName)</div><div class="ttdef"><b>Definition:</b> object.h:728</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1StmtSRefNode_html_afc61714fbac246f72d02d0729fb9ba2d"><div class="ttname"><a href="classtvm_1_1tir_1_1StmtSRefNode.html#afc61714fbac246f72d02d0729fb9ba2d">tvm::tir::StmtSRefNode::StmtAs</a></div><div class="ttdeci">const StmtType * StmtAs() const</div><div class="ttdoc">Get the referenced statement with proper type checking. It serves the same purpose as ObjectRef::as...</div><div class="ttdef"><b>Definition:</b> block_scope.h:89</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1BlockScopeNode_html_a460aa1dc8f4f608db1b7f529d51b9d29"><div class="ttname"><a href="classtvm_1_1tir_1_1BlockScopeNode.html#a460aa1dc8f4f608db1b7f529d51b9d29">tvm::tir::BlockScopeNode::VisitAttrs</a></div><div class="ttdeci">void VisitAttrs(AttrVisitor *v)</div><div class="ttdef"><b>Definition:</b> block_scope.h:230</div></div>
diff --git a/docs/reference/api/doxygen/bound_8h_source.html b/docs/reference/api/doxygen/bound_8h_source.html
index a3387a4850..28b7baefe3 100644
--- a/docs/reference/api/doxygen/bound_8h_source.html
+++ b/docs/reference/api/doxygen/bound_8h_source.html
@@ -71,7 +71,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1arith_html_af0860828b70ea8ff0d2f7c036c37afd5"><div class="ttname"><a href="namespacetvm_1_1arith.html#af0860828b70ea8ff0d2f7c036c37afd5">tvm::arith::DeduceBound</a></div><div class="ttdeci">IntSet DeduceBound(PrimExpr v, PrimExpr cond, const Map< Var, IntSet > &hint_map, const Map< Var, IntSet > &relax_map)</div><div class="ttdoc">Deduce the bound of the target variable in a expression, give the domain of each variables. Return und [...]
<div class="ttc" id="stmt_8h_html"><div class="ttname"><a href="stmt_8h.html">stmt.h</a></div><div class="ttdoc">TIR statements. </div></div>
<div class="ttc" id="tir_2expr_8h_html"><div class="ttname"><a href="tir_2expr_8h.html">expr.h</a></div><div class="ttdoc">TIR expressions. </div></div>
-<div class="ttc" id="namespacetvm_1_1tir_html_a8277e2a3d81a80a4776705673df51e0a"><div class="ttname"><a href="namespacetvm_1_1tir.html#a8277e2a3d81a80a4776705673df51e0a">tvm::tir::Region</a></div><div class="ttdeci">Array< Range > Region</div><div class="ttdef"><b>Definition:</b> var.h:169</div></div>
+<div class="ttc" id="namespacetvm_1_1tir_html_a8277e2a3d81a80a4776705673df51e0a"><div class="ttname"><a href="namespacetvm_1_1tir.html#a8277e2a3d81a80a4776705673df51e0a">tvm::tir::Region</a></div><div class="ttdeci">Array< Range > Region</div><div class="ttdef"><b>Definition:</b> var.h:176</div></div>
<div class="ttc" id="namespacetvm_1_1arith_html_a7641fa29ca6d8941f2696aacccf23998"><div class="ttname"><a href="namespacetvm_1_1arith.html#a7641fa29ca6d8941f2696aacccf23998">tvm::arith::DomainTouched</a></div><div class="ttdeci">Region DomainTouched(const Stmt &body, const tir::Buffer &buffer, bool consider_loads, bool consider_stores)</div><div class="ttdoc">Infer a regular domain that covers all the calls or provides within the given statement. </div></div>
<div class="ttc" id="int__set_8h_html"><div class="ttname"><a href="int__set_8h.html">int_set.h</a></div><div class="ttdoc">Integer set. </div></div>
</div><!-- fragment --></div><!-- contents -->
diff --git a/docs/reference/api/doxygen/classtvm_1_1PrimExpr__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1PrimExpr__inherit__graph.svg
index f233c8a82b..fd7c3016a7 100644
--- a/docs/reference/api/doxygen/classtvm_1_1PrimExpr__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1PrimExpr__inherit__graph.svg
@@ -988,22 +988,23 @@
<g id="node43" class="node">
<title>Node42</title>
<g id="a_node43"><a xlink:href="classtvm_1_1tir_1_1SizeVar.html" target="_top" xlink:title="a named variable represents a tensor index size ">
-<polygon fill="#ffffff" stroke="#000000" points="6001.5,-39 6001.5,-129 6096.5,-129 6096.5,-39 6001.5,-39"/>
-<text text-anchor="middle" x="6049" y="-117" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
-<polyline fill="none" stroke="#000000" points="6001.5,-110 6096.5,-110 "/>
-<text text-anchor="middle" x="6049" y="-98" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="6001.5,-91 6096.5,-91 "/>
-<text text-anchor="start" x="6009.5" y="-79" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
-<text text-anchor="start" x="6009.5" y="-68" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
-<text text-anchor="start" x="6009.5" y="-57" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="6009.5" y="-46" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<polygon fill="#ffffff" stroke="#000000" points="6001.5,-33.5 6001.5,-134.5 6096.5,-134.5 6096.5,-33.5 6001.5,-33.5"/>
+<text text-anchor="middle" x="6049" y="-122.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
+<polyline fill="none" stroke="#000000" points="6001.5,-115.5 6096.5,-115.5 "/>
+<text text-anchor="middle" x="6049" y="-103.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="6001.5,-96.5 6096.5,-96.5 "/>
+<text text-anchor="start" x="6009.5" y="-84.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
+<text text-anchor="start" x="6009.5" y="-73.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
+<text text-anchor="start" x="6009.5" y="-62.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
+<text text-anchor="start" x="6009.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="6009.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
</a>
</g>
</g>
<!-- Node41->Node42 -->
<g id="edge42" class="edge">
<title>Node41->Node42</title>
-<path fill="none" stroke="#191970" d="M6049,-194.1149C6049,-172.2126 6049,-148.7632 6049,-129.1568"/>
+<path fill="none" stroke="#191970" d="M6049,-194.1211C6049,-174.234 6049,-153.0708 6049,-134.6669"/>
<polygon fill="none" stroke="#191970" points="6045.5001,-194.3831 6049,-204.3831 6052.5001,-194.3831 6045.5001,-194.3831"/>
</g>
</g>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar-members.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar-members.html
index ecbed39503..8ef76394ed 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar-members.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar-members.html
@@ -93,13 +93,14 @@ $(function() {
<tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#ae31a5b9f40781d60a2901994ead700e8">same_as</a>(const ObjectRef &other) const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
<tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#ac470249315d9e395ad581d35dd5dcb05">SizeVar</a>(ObjectPtr< Object > n)</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html">tvm::tir::SizeVar</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
<tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#a0f8cb8a92feb96343939d223db90f7cd">SizeVar</a>(String name_hint="s", DataType t=DataType::Int(32), Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html">tvm::tir::SizeVar</a></td><td class="entry"><span class="mlabel">explicit</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1PrimExpr.html#a3ad47a31c4ce693077a93f154b2b1e12">TVM_DEFINE_OBJECT_REF_METHODS</a>(PrimExpr, BaseExpr, PrimExprNode)</td><td class="entry"><a class="el" href="classtvm_1_1PrimExpr.html">tvm::PrimExpr</a></td><td class="entry"></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1BaseExpr.html#aa513c6abed6e5b76c7fc9441649b3e4c">tvm::BaseExpr::TVM_DEFINE_OBJECT_REF_METHODS</a>(BaseExpr, ObjectRef, BaseExprNode)</td><td class="entry"><a class="el" href="classtvm_1_1BaseExpr.html">tvm::BaseExpr</a></td><td class="entry"></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7">unique</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#a0ae0da21d247cd87ea94fe3777c4405e">use_count</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a21ba7568a83bfc2a5896f9e0ff181129">Var</a>(ObjectPtr< Object > n)</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
- <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a8d4803f34057c3a0460cf2147239e17a">Var</a>(String name_hint="v", DataType dtype=DataType::Int(32), Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">explicit</span></td></tr>
- <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a241222d3b660a733dd45e871fe8bf2e5">Var</a>(String name_hint, Type type_annotation, Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">explicit</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#ab089bab85206d8e306cc61e879e525be">SizeVar</a>(String name_hint, Type type_annotation, Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html">tvm::tir::SizeVar</a></td><td class="entry"><span class="mlabel">explicit</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1PrimExpr.html#a3ad47a31c4ce693077a93f154b2b1e12">TVM_DEFINE_OBJECT_REF_METHODS</a>(PrimExpr, BaseExpr, PrimExprNode)</td><td class="entry"><a class="el" href="classtvm_1_1PrimExpr.html">tvm::PrimExpr</a></td><td class="entry"></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1BaseExpr.html#aa513c6abed6e5b76c7fc9441649b3e4c">tvm::BaseExpr::TVM_DEFINE_OBJECT_REF_METHODS</a>(BaseExpr, ObjectRef, BaseExprNode)</td><td class="entry"><a class="el" href="classtvm_1_1BaseExpr.html">tvm::BaseExpr</a></td><td class="entry"></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#a4e7cdb1574b93a59e784d70aa47b8da7">unique</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html#a0ae0da21d247cd87ea94fe3777c4405e">use_count</a>() const</td><td class="entry"><a class="el" href="classtvm_1_1runtime_1_1ObjectRef.html">tvm::runtime::ObjectRef</a></td><td class="entry"><span class="mlabel">inline</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a21ba7568a83bfc2a5896f9e0ff181129">Var</a>(ObjectPtr< Object > n)</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">inline</span><span class="mlabel">explicit</span></td></tr>
+ <tr class="even"><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a8d4803f34057c3a0460cf2147239e17a">Var</a>(String name_hint="v", DataType dtype=DataType::Int(32), Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">explicit</span></td></tr>
+ <tr><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html#a241222d3b660a733dd45e871fe8bf2e5">Var</a>(String name_hint, Type type_annotation, Span span=Span())</td><td class="entry"><a class="el" href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></td><td class="entry"><span class="mlabel">explicit</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_1SizeVar.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar.html
index c99f852936..731e06f248 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar.html
@@ -78,13 +78,13 @@ $(function() {
<div class="dynheader">
Inheritance diagram for tvm::tir::SizeVar:</div>
<div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1SizeVar__inherit__graph.svg" width="216" height="1030"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1SizeVar__inherit__graph.svg" width="216" height="1044"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
</div>
</div>
<div class="dynheader">
Collaboration diagram for tvm::tir::SizeVar:</div>
<div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1SizeVar__coll__graph.svg" width="216" height="1322"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1SizeVar__coll__graph.svg" width="216" height="1336"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
</div>
</div>
<table class="memberdecls">
@@ -109,6 +109,9 @@ Public Member Functions</h2></td></tr>
<tr class="memitem:a0f8cb8a92feb96343939d223db90f7cd"><td class="memItemLeft" align="right" valign="top"> </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#a0f8cb8a92feb96343939d223db90f7cd">SizeVar</a> (<a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> name_hint="s", DataType t=<a class="el" href="classtvm_1_1runtime_1_1DataType.html#ab45f13dd70d982d9f977c79b6f7fac98">DataType::Int</a>(32), <a class="el" hre [...]
<tr class="memdesc:a0f8cb8a92feb96343939d223db90f7cd"><td class="mdescLeft"> </td><td class="mdescRight">constructor <a href="#a0f8cb8a92feb96343939d223db90f7cd">More...</a><br /></td></tr>
<tr class="separator:a0f8cb8a92feb96343939d223db90f7cd"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab089bab85206d8e306cc61e879e525be"><td class="memItemLeft" align="right" valign="top"> </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#ab089bab85206d8e306cc61e879e525be">SizeVar</a> (<a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> name_hint, <a class="el" href="classtvm_1_1Type.html">Type</a> type_annotation, <a class="el" href="classtvm_1_1Span.html">Span</a> span=<a class="el" href="classtvm_1_ [...]
+<tr class="memdesc:ab089bab85206d8e306cc61e879e525be"><td class="mdescLeft"> </td><td class="mdescRight"><a class="el" href="classtvm_1_1Constructor.html" title="Managed reference to ConstructorNode. ">Constructor</a> which provides a more detailed type annotation. <a href="#ab089bab85206d8e306cc61e879e525be">More...</a><br /></td></tr>
+<tr class="separator:ab089bab85206d8e306cc61e879e525be"><td class="memSeparator" colspan="2"> </td></tr>
<tr class="memitem:ada44b624c1397c943ad9a0427378b048"><td class="memItemLeft" align="right" valign="top">const <a class="el" href="classtvm_1_1tir_1_1SizeVarNode.html">SizeVarNode</a> * </td><td class="memItemRight" valign="bottom"><a class="el" href="classtvm_1_1tir_1_1SizeVar.html#ada44b624c1397c943ad9a0427378b048">operator-></a> () const</td></tr>
<tr class="memdesc:ada44b624c1397c943ad9a0427378b048"><td class="mdescLeft"> </td><td class="mdescRight">Get pointer to the internal value. <a href="#ada44b624c1397c943ad9a0427378b048">More...</a><br /></td></tr>
<tr class="separator:ada44b624c1397c943ad9a0427378b048"><td class="memSeparator" colspan="2"> </td></tr>
@@ -230,7 +233,7 @@ Additional Inherited Members</h2></td></tr>
</div>
<h2 class="groupheader">Constructor & Destructor Documentation</h2>
<a id="ac470249315d9e395ad581d35dd5dcb05"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ac470249315d9e395ad581d35dd5dcb05">◆ </a></span>SizeVar() <span class="overload">[1/2]</span></h2>
+<h2 class="memtitle"><span class="permalink"><a href="#ac470249315d9e395ad581d35dd5dcb05">◆ </a></span>SizeVar() <span class="overload">[1/3]</span></h2>
<div class="memitem">
<div class="memproto">
@@ -256,7 +259,7 @@ Additional Inherited Members</h2></td></tr>
</div>
</div>
<a id="a0f8cb8a92feb96343939d223db90f7cd"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a0f8cb8a92feb96343939d223db90f7cd">◆ </a></span>SizeVar() <span class="overload">[2/2]</span></h2>
+<h2 class="memtitle"><span class="permalink"><a href="#a0f8cb8a92feb96343939d223db90f7cd">◆ </a></span>SizeVar() <span class="overload">[2/3]</span></h2>
<div class="memitem">
<div class="memproto">
@@ -305,6 +308,58 @@ Additional Inherited Members</h2></td></tr>
</dd>
</dl>
+</div>
+</div>
+<a id="ab089bab85206d8e306cc61e879e525be"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab089bab85206d8e306cc61e879e525be">◆ </a></span>SizeVar() <span class="overload">[3/3]</span></h2>
+
+<div class="memitem">
+<div class="memproto">
+<table class="mlabels">
+ <tr>
+ <td class="mlabels-left">
+ <table class="memname">
+ <tr>
+ <td class="memname">tvm::tir::SizeVar::SizeVar </td>
+ <td>(</td>
+ <td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> </td>
+ <td class="paramname"><em>name_hint</em>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype"><a class="el" href="classtvm_1_1Type.html">Type</a> </td>
+ <td class="paramname"><em>type_annotation</em>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype"><a class="el" href="classtvm_1_1Span.html">Span</a> </td>
+ <td class="paramname"><em>span</em> = <code><a class="el" href="classtvm_1_1Span.html">Span</a>()</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
+ </tr>
+ </table>
+ </td>
+ <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">explicit</span></span> </td>
+ </tr>
+</table>
+</div><div class="memdoc">
+
+<p><a class="el" href="classtvm_1_1Constructor.html" title="Managed reference to ConstructorNode. ">Constructor</a> which provides a more detailed type annotation. </p>
+<dl class="params"><dt>Parameters</dt><dd>
+ <table class="params">
+ <tr><td class="paramname">name_hint</td><td>variable name. </td></tr>
+ <tr><td class="paramname">type_annotation</td><td>The type annotation. </td></tr>
+ <tr><td class="paramname">span</td><td>The location of this object in the source code. </td></tr>
+ </table>
+ </dd>
+</dl>
+
</div>
</div>
<h2 class="groupheader">Member Function Documentation</h2>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__coll__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__coll__graph.svg
index 8d8c025866..2543471ebe 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__coll__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__coll__graph.svg
@@ -4,19 +4,20 @@
<!-- Generated by graphviz version 2.40.1 (20161225.0304)
-->
<!-- Title: tvm::tir::SizeVar Pages: 1 -->
-<svg width="162pt" height="991pt"
- viewBox="0.00 0.00 162.00 991.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 987)">
+<svg width="162pt" height="1002pt"
+ viewBox="0.00 0.00 162.00 1002.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 998)">
<title>tvm::tir::SizeVar</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-987 158,-987 158,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-998 158,-998 158,4 -4,4"/>
<!-- Node5 -->
<g id="node1" class="node">
<title>Node5</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="29.5,-.5 29.5,-90.5 124.5,-90.5 124.5,-.5 29.5,-.5"/>
-<text text-anchor="middle" x="77" y="-78.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
-<polyline fill="none" stroke="#000000" points="29.5,-71.5 124.5,-71.5 "/>
-<text text-anchor="middle" x="77" y="-59.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="29.5,-52.5 124.5,-52.5 "/>
+<polygon fill="#bfbfbf" stroke="#000000" points="29.5,-.5 29.5,-101.5 124.5,-101.5 124.5,-.5 29.5,-.5"/>
+<text text-anchor="middle" x="77" y="-89.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
+<polyline fill="none" stroke="#000000" points="29.5,-82.5 124.5,-82.5 "/>
+<text text-anchor="middle" x="77" y="-70.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="29.5,-63.5 124.5,-63.5 "/>
+<text text-anchor="start" x="37.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
@@ -26,134 +27,134 @@
<g id="node2" class="node">
<title>Node6</title>
<g id="a_node2"><a xlink:href="classtvm_1_1tir_1_1Var.html" target="_top" xlink:title="a named variable in TIR ">
-<polygon fill="#ffffff" stroke="#000000" points="18,-128.5 18,-251.5 136,-251.5 136,-128.5 18,-128.5"/>
-<text text-anchor="middle" x="77" y="-239.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
-<polyline fill="none" stroke="#000000" points="18,-232.5 136,-232.5 "/>
-<text text-anchor="middle" x="77" y="-220.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="18,-213.5 136,-213.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="18,-139.5 18,-262.5 136,-262.5 136,-139.5 18,-139.5"/>
+<text text-anchor="middle" x="77" y="-250.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
+<polyline fill="none" stroke="#000000" points="18,-243.5 136,-243.5 "/>
+<text text-anchor="middle" x="77" y="-231.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="18,-224.5 136,-224.5 "/>
+<text text-anchor="start" x="26" y="-212.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-201.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-190.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-179.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-168.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
-<text text-anchor="start" x="26" y="-157.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
-<text text-anchor="start" x="26" y="-146.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="26" y="-135.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="26" y="-179.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
+<text text-anchor="start" x="26" y="-168.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
+<text text-anchor="start" x="26" y="-157.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="26" y="-146.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
</a>
</g>
</g>
<!-- Node6->Node5 -->
<g id="edge1" class="edge">
<title>Node6->Node5</title>
-<path fill="none" stroke="#191970" d="M77,-118.2121C77,-108.8848 77,-99.5472 77,-90.8211"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-118.2998 77,-128.2999 80.5001,-118.2999 73.5001,-118.2998"/>
+<path fill="none" stroke="#191970" d="M77,-128.8462C77,-119.6965 77,-110.4904 77,-101.7734"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-129.1232 77,-139.1232 80.5001,-129.1233 73.5001,-129.1232"/>
</g>
<!-- Node7 -->
<g id="node3" class="node">
<title>Node7</title>
<g id="a_node3"><a xlink:href="classtvm_1_1PrimExpr.html" target="_top" xlink:title="Reference to PrimExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-289.5 0,-390.5 154,-390.5 154,-289.5 0,-289.5"/>
-<text text-anchor="middle" x="77" y="-378.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-371.5 154,-371.5 "/>
-<text text-anchor="middle" x="77" y="-359.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-352.5 154,-352.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="0,-300.5 0,-401.5 154,-401.5 154,-300.5 0,-300.5"/>
+<text text-anchor="middle" x="77" y="-389.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-382.5 154,-382.5 "/>
+<text text-anchor="middle" x="77" y="-370.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-363.5 154,-363.5 "/>
+<text text-anchor="start" x="8" y="-351.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
<text text-anchor="start" x="8" y="-340.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-329.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-318.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
-<text text-anchor="start" x="8" y="-307.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-296.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<text text-anchor="start" x="8" y="-329.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
+<text text-anchor="start" x="8" y="-318.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-307.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node7->Node6 -->
<g id="edge2" class="edge">
<title>Node7->Node6</title>
-<path fill="none" stroke="#191970" d="M77,-279.1359C77,-270.0734 77,-260.7251 77,-251.6466"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-279.2168 77,-289.2168 80.5001,-279.2169 73.5001,-279.2168"/>
+<path fill="none" stroke="#191970" d="M77,-290.1359C77,-281.0734 77,-271.7251 77,-262.6466"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-290.2168 77,-300.2168 80.5001,-290.2169 73.5001,-290.2168"/>
</g>
<!-- Node8 -->
<g id="node4" class="node">
<title>Node8</title>
<g id="a_node4"><a xlink:href="classtvm_1_1BaseExpr.html" target="_top" xlink:title="Managed reference to BaseExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-428.5 0,-496.5 154,-496.5 154,-428.5 0,-428.5"/>
-<text text-anchor="middle" x="77" y="-484.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-477.5 154,-477.5 "/>
-<text text-anchor="middle" x="77" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-458.5 154,-458.5 "/>
-<text text-anchor="start" x="8" y="-446.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-435.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<polygon fill="#ffffff" stroke="#000000" points="0,-439.5 0,-507.5 154,-507.5 154,-439.5 0,-439.5"/>
+<text text-anchor="middle" x="77" y="-495.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-488.5 154,-488.5 "/>
+<text text-anchor="middle" x="77" y="-476.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-469.5 154,-469.5 "/>
+<text text-anchor="start" x="8" y="-457.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-446.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node8->Node7 -->
<g id="edge3" class="edge">
<title>Node8->Node7</title>
-<path fill="none" stroke="#191970" d="M77,-418.2437C77,-409.3671 77,-399.9733 77,-390.8923"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-418.3769 77,-428.3769 80.5001,-418.377 73.5001,-418.3769"/>
+<path fill="none" stroke="#191970" d="M77,-429.2437C77,-420.3671 77,-410.9733 77,-401.8923"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-429.3769 77,-439.3769 80.5001,-429.377 73.5001,-429.3769"/>
</g>
<!-- Node9 -->
<g id="node5" class="node">
<title>Node9</title>
<g id="a_node5"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="10,-534.5 10,-756.5 144,-756.5 144,-534.5 10,-534.5"/>
-<text text-anchor="middle" x="77" y="-744.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="10,-737.5 144,-737.5 "/>
-<text text-anchor="start" x="18" y="-725.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<polyline fill="none" stroke="#000000" points="10,-718.5 144,-718.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="10,-545.5 10,-767.5 144,-767.5 144,-545.5 10,-545.5"/>
+<text text-anchor="middle" x="77" y="-755.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="10,-748.5 144,-748.5 "/>
+<text text-anchor="start" x="18" y="-736.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<polyline fill="none" stroke="#000000" points="10,-729.5 144,-729.5 "/>
+<text text-anchor="start" x="18" y="-717.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
<text text-anchor="start" x="18" y="-706.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-695.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-684.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="18" y="-673.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="18" y="-662.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="18" y="-651.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
-<text text-anchor="start" x="18" y="-640.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="18" y="-629.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="18" y="-618.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="18" y="-607.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="18" y="-596.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="18" y="-585.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="18" y="-574.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="18" y="-563.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="18" y="-552.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="18" y="-541.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<text text-anchor="start" x="18" y="-695.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="18" y="-684.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="18" y="-673.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="18" y="-662.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
+<text text-anchor="start" x="18" y="-651.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="18" y="-640.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="18" y="-629.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="18" y="-618.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="18" y="-607.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="18" y="-596.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="18" y="-585.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="18" y="-574.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="18" y="-563.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="18" y="-552.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
</a>
</g>
</g>
<!-- Node9->Node8 -->
<g id="edge4" class="edge">
<title>Node9->Node8</title>
-<path fill="none" stroke="#191970" d="M77,-524.2198C77,-514.2931 77,-504.8914 77,-496.5572"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-524.3012 77,-534.3012 80.5001,-524.3012 73.5001,-524.3012"/>
+<path fill="none" stroke="#191970" d="M77,-535.2198C77,-525.2931 77,-515.8914 77,-507.5572"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-535.3012 77,-545.3012 80.5001,-535.3012 73.5001,-535.3012"/>
</g>
<!-- Node10 -->
<g id="node6" class="node">
<title>Node10</title>
<g id="a_node6"><a xlink:href="classtvm_1_1runtime_1_1ObjectPtr.html" target="_top" xlink:title="{tvm::runtime::ObjectPtr\l\< tvm::runtime::Object \>\n||+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ObjectPtr()\l+ ~ObjectPtr()\l+ swap()\l+ get()\l+ operator-\>()\land 11 more...\l}">
-<polygon fill="#ffffff" stroke="#000000" points="7,-804.5 7,-982.5 147,-982.5 147,-804.5 7,-804.5"/>
-<text text-anchor="start" x="15" y="-970.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
-<text text-anchor="middle" x="77" y="-959.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">< tvm::runtime::Object ></text>
-<polyline fill="none" stroke="#000000" points="7,-952.5 147,-952.5 "/>
-<text text-anchor="middle" x="77" y="-940.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="7,-933.5 147,-933.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="7,-815.5 7,-993.5 147,-993.5 147,-815.5 7,-815.5"/>
+<text text-anchor="start" x="15" y="-981.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectPtr</text>
+<text text-anchor="middle" x="77" y="-970.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">< tvm::runtime::Object ></text>
+<polyline fill="none" stroke="#000000" points="7,-963.5 147,-963.5 "/>
+<text text-anchor="middle" x="77" y="-951.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="7,-944.5 147,-944.5 "/>
+<text text-anchor="start" x="15" y="-932.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
<text text-anchor="start" x="15" y="-921.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
<text text-anchor="start" x="15" y="-910.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
<text text-anchor="start" x="15" y="-899.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
<text text-anchor="start" x="15" y="-888.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
<text text-anchor="start" x="15" y="-877.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="15" y="-866.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectPtr()</text>
-<text text-anchor="start" x="15" y="-855.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
-<text text-anchor="start" x="15" y="-844.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
-<text text-anchor="start" x="15" y="-833.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="15" y="-822.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="15" y="-811.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
+<text text-anchor="start" x="15" y="-866.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ~ObjectPtr()</text>
+<text text-anchor="start" x="15" y="-855.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ swap()</text>
+<text text-anchor="start" x="15" y="-844.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="15" y="-833.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="15" y="-822.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">and 11 more...</text>
</a>
</g>
</g>
<!-- Node10->Node9 -->
<g id="edge5" class="edge">
<title>Node10->Node9</title>
-<path fill="none" stroke="#404040" d="M77,-804.3167C77,-792.8765 77,-781.0062 77,-769.1402"/>
-<polygon fill="none" stroke="#404040" points="77.0001,-768.7944 73,-762.7944 77,-756.7944 81,-762.7943 77.0001,-768.7944"/>
-<text text-anchor="middle" x="96.5" y="-778" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
+<path fill="none" stroke="#404040" d="M77,-815.3167C77,-803.8765 77,-792.0062 77,-780.1402"/>
+<polygon fill="none" stroke="#404040" points="77.0001,-779.7944 73,-773.7944 77,-767.7944 81,-773.7943 77.0001,-779.7944"/>
+<text text-anchor="middle" x="96.5" y="-789" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> #data_</text>
</g>
</g>
</svg>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__inherit__graph.svg
index 5c18da1695..1417eae11e 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1SizeVar__inherit__graph.svg
@@ -4,19 +4,20 @@
<!-- Generated by graphviz version 2.40.1 (20161225.0304)
-->
<!-- Title: tvm::tir::SizeVar Pages: 1 -->
-<svg width="162pt" height="772pt"
- viewBox="0.00 0.00 162.00 772.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 768)">
+<svg width="162pt" height="783pt"
+ viewBox="0.00 0.00 162.00 783.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 779)">
<title>tvm::tir::SizeVar</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-768 158,-768 158,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-779 158,-779 158,4 -4,4"/>
<!-- Node0 -->
<g id="node1" class="node">
<title>Node0</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="29.5,-.5 29.5,-90.5 124.5,-90.5 124.5,-.5 29.5,-.5"/>
-<text text-anchor="middle" x="77" y="-78.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
-<polyline fill="none" stroke="#000000" points="29.5,-71.5 124.5,-71.5 "/>
-<text text-anchor="middle" x="77" y="-59.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="29.5,-52.5 124.5,-52.5 "/>
+<polygon fill="#bfbfbf" stroke="#000000" points="29.5,-.5 29.5,-101.5 124.5,-101.5 124.5,-.5 29.5,-.5"/>
+<text text-anchor="middle" x="77" y="-89.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
+<polyline fill="none" stroke="#000000" points="29.5,-82.5 124.5,-82.5 "/>
+<text text-anchor="middle" x="77" y="-70.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="29.5,-63.5 124.5,-63.5 "/>
+<text text-anchor="start" x="37.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
@@ -26,104 +27,104 @@
<g id="node2" class="node">
<title>Node1</title>
<g id="a_node2"><a xlink:href="classtvm_1_1tir_1_1Var.html" target="_top" xlink:title="a named variable in TIR ">
-<polygon fill="#ffffff" stroke="#000000" points="18,-127.5 18,-250.5 136,-250.5 136,-127.5 18,-127.5"/>
-<text text-anchor="middle" x="77" y="-238.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
-<polyline fill="none" stroke="#000000" points="18,-231.5 136,-231.5 "/>
-<text text-anchor="middle" x="77" y="-219.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="18,-212.5 136,-212.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="18,-138.5 18,-261.5 136,-261.5 136,-138.5 18,-138.5"/>
+<text text-anchor="middle" x="77" y="-249.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
+<polyline fill="none" stroke="#000000" points="18,-242.5 136,-242.5 "/>
+<text text-anchor="middle" x="77" y="-230.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="18,-223.5 136,-223.5 "/>
+<text text-anchor="start" x="26" y="-211.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-200.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-189.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-178.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-167.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
-<text text-anchor="start" x="26" y="-156.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
-<text text-anchor="start" x="26" y="-145.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="26" y="-134.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="26" y="-178.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
+<text text-anchor="start" x="26" y="-167.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
+<text text-anchor="start" x="26" y="-156.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="26" y="-145.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
</a>
</g>
</g>
<!-- Node1->Node0 -->
<g id="edge1" class="edge">
<title>Node1->Node0</title>
-<path fill="none" stroke="#191970" d="M77,-117.1555C77,-108.135 77,-99.1166 77,-90.6704"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-117.3102 77,-127.3103 80.5001,-117.3103 73.5001,-117.3102"/>
+<path fill="none" stroke="#191970" d="M77,-128.0596C77,-119.1193 77,-110.1299 77,-101.607"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-128.105 77,-138.105 80.5001,-128.1051 73.5001,-128.105"/>
</g>
<!-- Node2 -->
<g id="node3" class="node">
<title>Node2</title>
<g id="a_node3"><a xlink:href="classtvm_1_1PrimExpr.html" target="_top" xlink:title="Reference to PrimExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-287.5 0,-388.5 154,-388.5 154,-287.5 0,-287.5"/>
-<text text-anchor="middle" x="77" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-369.5 154,-369.5 "/>
-<text text-anchor="middle" x="77" y="-357.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-350.5 154,-350.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="0,-298.5 0,-399.5 154,-399.5 154,-298.5 0,-298.5"/>
+<text text-anchor="middle" x="77" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-380.5 154,-380.5 "/>
+<text text-anchor="middle" x="77" y="-368.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-361.5 154,-361.5 "/>
+<text text-anchor="start" x="8" y="-349.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
<text text-anchor="start" x="8" y="-338.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-327.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-316.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
-<text text-anchor="start" x="8" y="-305.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-294.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<text text-anchor="start" x="8" y="-327.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
+<text text-anchor="start" x="8" y="-316.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-305.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node2->Node1 -->
<g id="edge2" class="edge">
<title>Node2->Node1</title>
-<path fill="none" stroke="#191970" d="M77,-277.0751C77,-268.3766 77,-259.4225 77,-250.709"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-277.1394 77,-287.1394 80.5001,-277.1394 73.5001,-277.1394"/>
+<path fill="none" stroke="#191970" d="M77,-288.0751C77,-279.3766 77,-270.4225 77,-261.709"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-288.1394 77,-298.1394 80.5001,-288.1394 73.5001,-288.1394"/>
</g>
<!-- Node3 -->
<g id="node4" class="node">
<title>Node3</title>
<g id="a_node4"><a xlink:href="classtvm_1_1BaseExpr.html" target="_top" xlink:title="Managed reference to BaseExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-425.5 0,-493.5 154,-493.5 154,-425.5 0,-425.5"/>
-<text text-anchor="middle" x="77" y="-481.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-474.5 154,-474.5 "/>
-<text text-anchor="middle" x="77" y="-462.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-455.5 154,-455.5 "/>
-<text text-anchor="start" x="8" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-432.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<polygon fill="#ffffff" stroke="#000000" points="0,-436.5 0,-504.5 154,-504.5 154,-436.5 0,-436.5"/>
+<text text-anchor="middle" x="77" y="-492.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-485.5 154,-485.5 "/>
+<text text-anchor="middle" x="77" y="-473.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-466.5 154,-466.5 "/>
+<text text-anchor="start" x="8" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node3->Node2 -->
<g id="edge3" class="edge">
<title>Node3->Node2</title>
-<path fill="none" stroke="#191970" d="M77,-415.1349C77,-406.5408 77,-397.474 77,-388.6952"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-415.3321 77,-425.3321 80.5001,-415.3322 73.5001,-415.3321"/>
+<path fill="none" stroke="#191970" d="M77,-426.1349C77,-417.5408 77,-408.474 77,-399.6952"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-426.3321 77,-436.3321 80.5001,-426.3322 73.5001,-426.3321"/>
</g>
<!-- Node4 -->
<g id="node5" class="node">
<title>Node4</title>
<g id="a_node5"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="10,-530.5 10,-763.5 144,-763.5 144,-530.5 10,-530.5"/>
-<text text-anchor="middle" x="77" y="-751.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="10,-744.5 144,-744.5 "/>
-<text text-anchor="start" x="18" y="-732.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<text text-anchor="start" x="18" y="-721.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
-<polyline fill="none" stroke="#000000" points="10,-714.5 144,-714.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="10,-541.5 10,-774.5 144,-774.5 144,-541.5 10,-541.5"/>
+<text text-anchor="middle" x="77" y="-762.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="10,-755.5 144,-755.5 "/>
+<text text-anchor="start" x="18" y="-743.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<text text-anchor="start" x="18" y="-732.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
+<polyline fill="none" stroke="#000000" points="10,-725.5 144,-725.5 "/>
+<text text-anchor="start" x="18" y="-713.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
<text text-anchor="start" x="18" y="-702.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="18" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="18" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="18" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
-<text text-anchor="start" x="18" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="18" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="18" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="18" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="18" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="18" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="18" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="18" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="18" y="-548.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="18" y="-537.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<text text-anchor="start" x="18" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="18" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="18" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="18" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
+<text text-anchor="start" x="18" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="18" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="18" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="18" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="18" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="18" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="18" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="18" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="18" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="18" y="-548.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
</a>
</g>
</g>
<!-- Node4->Node3 -->
<g id="edge4" class="edge">
<title>Node4->Node3</title>
-<path fill="none" stroke="#191970" d="M77,-520.2822C77,-510.749 77,-501.737 77,-493.7166"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-520.3857 77,-530.3857 80.5001,-520.3857 73.5001,-520.3857"/>
+<path fill="none" stroke="#191970" d="M77,-531.2822C77,-521.749 77,-512.737 77,-504.7166"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-531.3857 77,-541.3857 80.5001,-531.3857 73.5001,-531.3857"/>
</g>
</g>
</svg>
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var.html b/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var.html
index af4fa814e7..57a615ece3 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var.html
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var.html
@@ -78,7 +78,7 @@ $(function() {
<div class="dynheader">
Inheritance diagram for tvm::tir::Var:</div>
<div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1Var__inherit__graph.svg" width="216" height="1030"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
+<div class="center"><iframe scrolling="no" frameborder="0" src="classtvm_1_1tir_1_1Var__inherit__graph.svg" width="216" height="1044"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
</div>
</div>
<div class="dynheader">
diff --git a/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var__inherit__graph.svg b/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var__inherit__graph.svg
index 9372e09590..db61e0336d 100644
--- a/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var__inherit__graph.svg
+++ b/docs/reference/api/doxygen/classtvm_1_1tir_1_1Var__inherit__graph.svg
@@ -4,36 +4,37 @@
<!-- Generated by graphviz version 2.40.1 (20161225.0304)
-->
<!-- Title: tvm::tir::Var Pages: 1 -->
-<svg width="162pt" height="772pt"
- viewBox="0.00 0.00 162.00 772.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
-<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 768)">
+<svg width="162pt" height="783pt"
+ viewBox="0.00 0.00 162.00 783.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<g id="graph0" class="graph" transform="scale(1 1) rotate(0) translate(4 779)">
<title>tvm::tir::Var</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-768 158,-768 158,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-779 158,-779 158,4 -4,4"/>
<!-- Node0 -->
<g id="node1" class="node">
<title>Node0</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="18,-127.5 18,-250.5 136,-250.5 136,-127.5 18,-127.5"/>
-<text text-anchor="middle" x="77" y="-238.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
-<polyline fill="none" stroke="#000000" points="18,-231.5 136,-231.5 "/>
-<text text-anchor="middle" x="77" y="-219.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="18,-212.5 136,-212.5 "/>
+<polygon fill="#bfbfbf" stroke="#000000" points="18,-138.5 18,-261.5 136,-261.5 136,-138.5 18,-138.5"/>
+<text text-anchor="middle" x="77" y="-249.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::Var</text>
+<polyline fill="none" stroke="#000000" points="18,-242.5 136,-242.5 "/>
+<text text-anchor="middle" x="77" y="-230.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="18,-223.5 136,-223.5 "/>
+<text text-anchor="start" x="26" y="-211.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-200.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
<text text-anchor="start" x="26" y="-189.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-178.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ Var()</text>
-<text text-anchor="start" x="26" y="-167.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
-<text text-anchor="start" x="26" y="-156.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
-<text text-anchor="start" x="26" y="-145.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="26" y="-134.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="26" y="-178.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_suffix()</text>
+<text text-anchor="start" x="26" y="-167.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ copy_with_dtype()</text>
+<text text-anchor="start" x="26" y="-156.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="26" y="-145.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
</g>
<!-- Node4 -->
<g id="node5" class="node">
<title>Node4</title>
<g id="a_node5"><a xlink:href="classtvm_1_1tir_1_1SizeVar.html" target="_top" xlink:title="a named variable represents a tensor index size ">
-<polygon fill="#ffffff" stroke="#000000" points="29.5,-.5 29.5,-90.5 124.5,-90.5 124.5,-.5 29.5,-.5"/>
-<text text-anchor="middle" x="77" y="-78.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
-<polyline fill="none" stroke="#000000" points="29.5,-71.5 124.5,-71.5 "/>
-<text text-anchor="middle" x="77" y="-59.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="29.5,-52.5 124.5,-52.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="29.5,-.5 29.5,-101.5 124.5,-101.5 124.5,-.5 29.5,-.5"/>
+<text text-anchor="middle" x="77" y="-89.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::tir::SizeVar</text>
+<polyline fill="none" stroke="#000000" points="29.5,-82.5 124.5,-82.5 "/>
+<text text-anchor="middle" x="77" y="-70.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="29.5,-63.5 124.5,-63.5 "/>
+<text text-anchor="start" x="37.5" y="-51.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-40.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-29.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ SizeVar()</text>
<text text-anchor="start" x="37.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
@@ -44,86 +45,86 @@
<!-- Node0->Node4 -->
<g id="edge4" class="edge">
<title>Node0->Node4</title>
-<path fill="none" stroke="#191970" d="M77,-117.1555C77,-108.135 77,-99.1166 77,-90.6704"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-117.3102 77,-127.3103 80.5001,-117.3103 73.5001,-117.3102"/>
+<path fill="none" stroke="#191970" d="M77,-128.0596C77,-119.1193 77,-110.1299 77,-101.607"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-128.105 77,-138.105 80.5001,-128.1051 73.5001,-128.105"/>
</g>
<!-- Node1 -->
<g id="node2" class="node">
<title>Node1</title>
<g id="a_node2"><a xlink:href="classtvm_1_1PrimExpr.html" target="_top" xlink:title="Reference to PrimExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-287.5 0,-388.5 154,-388.5 154,-287.5 0,-287.5"/>
-<text text-anchor="middle" x="77" y="-376.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-369.5 154,-369.5 "/>
-<text text-anchor="middle" x="77" y="-357.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-350.5 154,-350.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="0,-298.5 0,-399.5 154,-399.5 154,-298.5 0,-298.5"/>
+<text text-anchor="middle" x="77" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::PrimExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-380.5 154,-380.5 "/>
+<text text-anchor="middle" x="77" y="-368.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-361.5 154,-361.5 "/>
+<text text-anchor="start" x="8" y="-349.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
<text text-anchor="start" x="8" y="-338.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-327.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ PrimExpr()</text>
-<text text-anchor="start" x="8" y="-316.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
-<text text-anchor="start" x="8" y="-305.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-294.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<text text-anchor="start" x="8" y="-327.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ dtype()</text>
+<text text-anchor="start" x="8" y="-316.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-305.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node1->Node0 -->
<g id="edge1" class="edge">
<title>Node1->Node0</title>
-<path fill="none" stroke="#191970" d="M77,-277.0751C77,-268.3766 77,-259.4225 77,-250.709"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-277.1394 77,-287.1394 80.5001,-277.1394 73.5001,-277.1394"/>
+<path fill="none" stroke="#191970" d="M77,-288.0751C77,-279.3766 77,-270.4225 77,-261.709"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-288.1394 77,-298.1394 80.5001,-288.1394 73.5001,-288.1394"/>
</g>
<!-- Node2 -->
<g id="node3" class="node">
<title>Node2</title>
<g id="a_node3"><a xlink:href="classtvm_1_1BaseExpr.html" target="_top" xlink:title="Managed reference to BaseExprNode. ">
-<polygon fill="#ffffff" stroke="#000000" points="0,-425.5 0,-493.5 154,-493.5 154,-425.5 0,-425.5"/>
-<text text-anchor="middle" x="77" y="-481.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
-<polyline fill="none" stroke="#000000" points="0,-474.5 154,-474.5 "/>
-<text text-anchor="middle" x="77" y="-462.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
-<polyline fill="none" stroke="#000000" points="0,-455.5 154,-455.5 "/>
-<text text-anchor="start" x="8" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
-<text text-anchor="start" x="8" y="-432.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
+<polygon fill="#ffffff" stroke="#000000" points="0,-436.5 0,-504.5 154,-504.5 154,-436.5 0,-436.5"/>
+<text text-anchor="middle" x="77" y="-492.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::BaseExpr</text>
+<polyline fill="none" stroke="#000000" points="0,-485.5 154,-485.5 "/>
+<text text-anchor="middle" x="77" y="-473.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"> </text>
+<polyline fill="none" stroke="#000000" points="0,-466.5 154,-466.5 "/>
+<text text-anchor="start" x="8" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ TVM_DEFINE_OBJECT_REF</text>
+<text text-anchor="start" x="8" y="-443.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_METHODS()</text>
</a>
</g>
</g>
<!-- Node2->Node1 -->
<g id="edge2" class="edge">
<title>Node2->Node1</title>
-<path fill="none" stroke="#191970" d="M77,-415.1349C77,-406.5408 77,-397.474 77,-388.6952"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-415.3321 77,-425.3321 80.5001,-415.3322 73.5001,-415.3321"/>
+<path fill="none" stroke="#191970" d="M77,-426.1349C77,-417.5408 77,-408.474 77,-399.6952"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-426.3321 77,-436.3321 80.5001,-426.3322 73.5001,-426.3321"/>
</g>
<!-- Node3 -->
<g id="node4" class="node">
<title>Node3</title>
<g id="a_node4"><a xlink:href="classtvm_1_1runtime_1_1ObjectRef.html" target="_top" xlink:title="Base class of all object reference. ">
-<polygon fill="#ffffff" stroke="#000000" points="10,-530.5 10,-763.5 144,-763.5 144,-530.5 10,-530.5"/>
-<text text-anchor="middle" x="77" y="-751.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
-<polyline fill="none" stroke="#000000" points="10,-744.5 144,-744.5 "/>
-<text text-anchor="start" x="18" y="-732.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
-<text text-anchor="start" x="18" y="-721.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
-<polyline fill="none" stroke="#000000" points="10,-714.5 144,-714.5 "/>
+<polygon fill="#ffffff" stroke="#000000" points="10,-541.5 10,-774.5 144,-774.5 144,-541.5 10,-541.5"/>
+<text text-anchor="middle" x="77" y="-762.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm::runtime::ObjectRef</text>
+<polyline fill="none" stroke="#000000" points="10,-755.5 144,-755.5 "/>
+<text text-anchor="start" x="18" y="-743.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ _type_is_nullable</text>
+<text text-anchor="start" x="18" y="-732.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># data_</text>
+<polyline fill="none" stroke="#000000" points="10,-725.5 144,-725.5 "/>
+<text text-anchor="start" x="18" y="-713.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
<text text-anchor="start" x="18" y="-702.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ ObjectRef()</text>
-<text text-anchor="start" x="18" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
-<text text-anchor="start" x="18" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
-<text text-anchor="start" x="18" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
-<text text-anchor="start" x="18" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
-<text text-anchor="start" x="18" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
-<text text-anchor="start" x="18" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
-<text text-anchor="start" x="18" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
-<text text-anchor="start" x="18" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
-<text text-anchor="start" x="18" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
-<text text-anchor="start" x="18" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
-<text text-anchor="start" x="18" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
-<text text-anchor="start" x="18" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
-<text text-anchor="start" x="18" y="-548.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
-<text text-anchor="start" x="18" y="-537.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
+<text text-anchor="start" x="18" y="-691.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ same_as()</text>
+<text text-anchor="start" x="18" y="-680.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator==()</text>
+<text text-anchor="start" x="18" y="-669.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator!=()</text>
+<text text-anchor="start" x="18" y="-658.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator<()</text>
+<text text-anchor="start" x="18" y="-647.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ defined()</text>
+<text text-anchor="start" x="18" y="-636.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ get()</text>
+<text text-anchor="start" x="18" y="-625.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ operator->()</text>
+<text text-anchor="start" x="18" y="-614.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ unique()</text>
+<text text-anchor="start" x="18" y="-603.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ use_count()</text>
+<text text-anchor="start" x="18" y="-592.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">+ as()</text>
+<text text-anchor="start" x="18" y="-581.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># get_mutable()</text>
+<text text-anchor="start" x="18" y="-570.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># DowncastNoCheck()</text>
+<text text-anchor="start" x="18" y="-559.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># FFIClearAfterMove()</text>
+<text text-anchor="start" x="18" y="-548.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000"># GetDataPtr()</text>
</a>
</g>
</g>
<!-- Node3->Node2 -->
<g id="edge3" class="edge">
<title>Node3->Node2</title>
-<path fill="none" stroke="#191970" d="M77,-520.2822C77,-510.749 77,-501.737 77,-493.7166"/>
-<polygon fill="none" stroke="#191970" points="73.5001,-520.3857 77,-530.3857 80.5001,-520.3857 73.5001,-520.3857"/>
+<path fill="none" stroke="#191970" d="M77,-531.2822C77,-521.749 77,-512.737 77,-504.7166"/>
+<polygon fill="none" stroke="#191970" points="73.5001,-531.3857 77,-541.3857 80.5001,-531.3857 73.5001,-531.3857"/>
</g>
</g>
</svg>
diff --git a/docs/reference/api/doxygen/cuda_2dense_8h_source.html b/docs/reference/api/doxygen/cuda_2dense_8h_source.html
index b02991e11b..e425078e3e 100644
--- a/docs/reference/api/doxygen/cuda_2dense_8h_source.html
+++ b/docs/reference/api/doxygen/cuda_2dense_8h_source.html
@@ -76,7 +76,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1te_html"><div class="ttname"><a href="namespacetvm_1_1te.html">tvm::te</a></div><div class="ttdoc">Tensor expression language DSL. </div><div class="ttdef"><b>Definition:</b> extracted_task.h:33</div></div>
<div class="ttc" id="classtvm_1_1te_1_1Operation_html"><div class="ttname"><a href="classtvm_1_1te_1_1Operation.html">tvm::te::Operation</a></div><div class="ttdoc">Operation that produces tensors. </div><div class="ttdef"><b>Definition:</b> tensor.h:47</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1Var_html"><div class="ttname"><a href="classtvm_1_1tir_1_1Var.html">tvm::tir::Var</a></div><div class="ttdoc">a named variable in TIR </div><div class="ttdef"><b>Definition:</b> var.h:88</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Array_html_aa026b914ee05f81b6c20130b8905f257"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#aa026b914ee05f81b6c20130b8905f257">tvm::runtime::Array::push_back</a></div><div class="ttdeci">void push_back(const T &item)</div><div class="ttdoc">push a new item to the back of the list </div><div class="ttdef"><b>Definition:</b> array.h:457</div></div>
<div class="ttc" id="schedule__pass_8h_html"><div class="ttname"><a href="schedule__pass_8h.html">schedule_pass.h</a></div><div class="ttdoc">Collection of Schedule pass functions. </div></div>
<div class="ttc" id="namespacetvm_1_1topi_1_1contrib_html_a68c8492ea536608724ef6267b0785054"><div class="ttname"><a href="namespacetvm_1_1topi_1_1contrib.html#a68c8492ea536608724ef6267b0785054">tvm::topi::contrib::cublas_matmul</a></div><div class="ttdeci">Tensor cublas_matmul(const Tensor &lhs, const Tensor &rhs, bool transa, bool transb)</div><div class="ttdoc">Create an op that multiplies lhs and rhs with cuBLAS. </div><div class="ttdef"><b>Definition:</b> cublas.h:46</div></div>
diff --git a/docs/reference/api/doxygen/cuda_2injective_8h_source.html b/docs/reference/api/doxygen/cuda_2injective_8h_source.html
index 56300cf5d3..da673528d6 100644
--- a/docs/reference/api/doxygen/cuda_2injective_8h_source.html
+++ b/docs/reference/api/doxygen/cuda_2injective_8h_source.html
@@ -72,7 +72,7 @@ $(function() {
<div class="ttc" id="namespacetvm_html"><div class="ttname"><a href="namespacetvm.html">tvm</a></div><div class="ttdoc">runtime implementation for LibTorch/TorchScript. </div><div class="ttdef"><b>Definition:</b> analyzer.h:36</div></div>
<div class="ttc" id="namespacetvm_1_1te_html"><div class="ttname"><a href="namespacetvm_1_1te.html">tvm::te</a></div><div class="ttdoc">Tensor expression language DSL. </div><div class="ttdef"><b>Definition:</b> extracted_task.h:33</div></div>
<div class="ttc" id="classtvm_1_1Target_html_abed5e5cfb5d36e70ea5eaadef9fb63b2"><div class="ttname"><a href="classtvm_1_1Target.html#abed5e5cfb5d36e70ea5eaadef9fb63b2">tvm::Target::Current</a></div><div class="ttdeci">static tvm::Target Current(bool allow_not_defined=true)</div><div class="ttdoc">Get the current target context from thread local storage. </div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Array_html_aa026b914ee05f81b6c20130b8905f257"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#aa026b914ee05f81b6c20130b8905f257">tvm::runtime::Array::push_back</a></div><div class="ttdeci">void push_back(const T &item)</div><div class="ttdoc">push a new item to the back of the list </div><div class="ttdef"><b>Definition:</b> array.h:457</div></div>
<div class="ttc" id="schedule__pass_8h_html"><div class="ttname"><a href="schedule__pass_8h.html">schedule_pass.h</a></div><div class="ttdoc">Collection of Schedule pass functions. </div></div>
<div class="ttc" id="classtvm_1_1Range_html"><div class="ttname"><a href="classtvm_1_1Range.html">tvm::Range</a></div><div class="ttdoc">Range constainer. </div><div class="ttdef"><b>Definition:</b> expr.h:715</div></div>
diff --git a/docs/reference/api/doxygen/cuda_2pooling_8h_source.html b/docs/reference/api/doxygen/cuda_2pooling_8h_source.html
index 8cfdfd0746..892ef1a10a 100644
--- a/docs/reference/api/doxygen/cuda_2pooling_8h_source.html
+++ b/docs/reference/api/doxygen/cuda_2pooling_8h_source.html
@@ -73,7 +73,7 @@ $(function() {
<div class="ttc" id="namespacetvm_html"><div class="ttname"><a href="namespacetvm.html">tvm</a></div><div class="ttdoc">runtime implementation for LibTorch/TorchScript. </div><div class="ttdef"><b>Definition:</b> analyzer.h:36</div></div>
<div class="ttc" id="namespacetvm_1_1te_html"><div class="ttname"><a href="namespacetvm_1_1te.html">tvm::te</a></div><div class="ttdoc">Tensor expression language DSL. </div><div class="ttdef"><b>Definition:</b> extracted_task.h:33</div></div>
<div class="ttc" id="classtvm_1_1te_1_1Operation_html"><div class="ttname"><a href="classtvm_1_1te_1_1Operation.html">tvm::te::Operation</a></div><div class="ttdoc">Operation that produces tensors. </div><div class="ttdef"><b>Definition:</b> tensor.h:47</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="namespacetvm_1_1topi_1_1rocm_html_a7d6d48e5cbcec1b7cc0a90c7399e2a8a"><div class="ttname"><a href="namespacetvm_1_1topi_1_1rocm.html#a7d6d48e5cbcec1b7cc0a90c7399e2a8a">tvm::topi::rocm::schedule_pool</a></div><div class="ttdeci">Schedule schedule_pool(const Target &target, const Array< Tensor > &outs)</div><div class="ttdoc">Create a rocm schedule for pool. </div><div class="ttdef"><b>Definition:</b> pooling.h:49</div></div>
<div class="ttc" id="namespacetvm_1_1topi_1_1rocm_html_a45aee34b0000f98aafd958ffe9baebc0"><div class="ttname"><a href="namespacetvm_1_1topi_1_1rocm.html#a45aee34b0000f98aafd958ffe9baebc0">tvm::topi::rocm::schedule_global_pool</a></div><div class="ttdeci">Schedule schedule_global_pool(const Target &target, const Array< Tensor > &outs)</div><div class="ttdoc">Create a rocm schedule for global_pool. </div><div class="ttdef"><b>Definition:</b> pooling.h:61</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Array_html_aa026b914ee05f81b6c20130b8905f257"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#aa026b914ee05f81b6c20130b8905f257">tvm::runtime::Array::push_back</a></div><div class="ttdeci">void push_back(const T &item)</div><div class="ttdoc">push a new item to the back of the list </div><div class="ttdef"><b>Definition:</b> array.h:457</div></div>
diff --git a/docs/reference/api/doxygen/cuda_2reduction_8h_source.html b/docs/reference/api/doxygen/cuda_2reduction_8h_source.html
index 84fc12538b..54aa57eeef 100644
--- a/docs/reference/api/doxygen/cuda_2reduction_8h_source.html
+++ b/docs/reference/api/doxygen/cuda_2reduction_8h_source.html
@@ -78,7 +78,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1te_html"><div class="ttname"><a href="namespacetvm_1_1te.html">tvm::te</a></div><div class="ttdoc">Tensor expression language DSL. </div><div class="ttdef"><b>Definition:</b> extracted_task.h:33</div></div>
<div class="ttc" id="classtvm_1_1te_1_1Operation_html"><div class="ttname"><a href="classtvm_1_1te_1_1Operation.html">tvm::te::Operation</a></div><div class="ttdoc">Operation that produces tensors. </div><div class="ttdef"><b>Definition:</b> tensor.h:47</div></div>
<div class="ttc" id="namespacetvm_1_1topi_html_a5975401def8848aaa49ae83729263919"><div class="ttname"><a href="namespacetvm_1_1topi.html#a5975401def8848aaa49ae83729263919">tvm::topi::is_injective</a></div><div class="ttdeci">bool is_injective(std::string tag)</div><div class="ttdef"><b>Definition:</b> tags.h:51</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="classtvm_1_1te_1_1PlaceholderOpNode_html"><div class="ttname"><a href="classtvm_1_1te_1_1PlaceholderOpNode.html">tvm::te::PlaceholderOpNode</a></div><div class="ttdoc">A placeholder op represents an input placeholder. </div><div class="ttdef"><b>Definition:</b> operation.h:152</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Array_html_aa026b914ee05f81b6c20130b8905f257"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#aa026b914ee05f81b6c20130b8905f257">tvm::runtime::Array::push_back</a></div><div class="ttdeci">void push_back(const T &item)</div><div class="ttdoc">push a new item to the back of the list </div><div class="ttdef"><b>Definition:</b> array.h:457</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Object_html_a90e90b3f4ba8a590baff78c75807bbc7"><div class="ttname"><a href="classtvm_1_1runtime_1_1Object.html#a90e90b3f4ba8a590baff78c75807bbc7">tvm::runtime::Object::IsInstance</a></div><div class="ttdeci">bool IsInstance() const</div><div class="ttdef"><b>Definition:</b> object.h:829</div></div>
diff --git a/docs/reference/api/doxygen/cuda_2softmax_8h_source.html b/docs/reference/api/doxygen/cuda_2softmax_8h_source.html
index a75fcc5d7e..f4693efdb7 100644
--- a/docs/reference/api/doxygen/cuda_2softmax_8h_source.html
+++ b/docs/reference/api/doxygen/cuda_2softmax_8h_source.html
@@ -71,7 +71,7 @@ $(function() {
<div class="ttc" id="namespacetvm_1_1te_html_a485034766309df280239e0994913b34b"><div class="ttname"><a href="namespacetvm_1_1te.html#a485034766309df280239e0994913b34b">tvm::te::create_schedule</a></div><div class="ttdeci">Schedule create_schedule(Array< Operation > ops)</div><div class="ttdoc">Create a schedule for array of ops(and their dependencies). </div><div class="ttdef"><b>Definition:</b> schedule.h:654</div></div>
<div class="ttc" id="namespacetvm_html"><div class="ttname"><a href="namespacetvm.html">tvm</a></div><div class="ttdoc">runtime implementation for LibTorch/TorchScript. </div><div class="ttdef"><b>Definition:</b> analyzer.h:36</div></div>
<div class="ttc" id="namespacetvm_1_1te_html"><div class="ttname"><a href="namespacetvm_1_1te.html">tvm::te</a></div><div class="ttdoc">Tensor expression language DSL. </div><div class="ttdef"><b>Definition:</b> extracted_task.h:33</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="namespacetvm_1_1topi_1_1nn_html_aa2cb22c64412c3eacb351c12b883333b"><div class="ttname"><a href="namespacetvm_1_1topi_1_1nn.html#aa2cb22c64412c3eacb351c12b883333b">tvm::topi::nn::softmax</a></div><div class="ttdeci">Tensor softmax(const Tensor &x, int axis=-1, std::string name="tensor", std::string tag="softmax_output")</div><div class="ttdoc">Softmax activation. </div><div class="ttdef"><b>Definition:</b> softmax.h:50</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Array_html_aa026b914ee05f81b6c20130b8905f257"><div class="ttname"><a href="classtvm_1_1runtime_1_1Array.html#aa026b914ee05f81b6c20130b8905f257">tvm::runtime::Array::push_back</a></div><div class="ttdeci">void push_back(const T &item)</div><div class="ttdoc">push a new item to the back of the list </div><div class="ttdef"><b>Definition:</b> array.h:457</div></div>
<div class="ttc" id="schedule__pass_8h_html"><div class="ttname"><a href="schedule__pass_8h.html">schedule_pass.h</a></div><div class="ttdoc">Collection of Schedule pass functions. </div></div>
diff --git a/docs/reference/api/doxygen/data__layout_8h_source.html b/docs/reference/api/doxygen/data__layout_8h_source.html
index a6855aafed..d86a9b9fc0 100644
--- a/docs/reference/api/doxygen/data__layout_8h_source.html
+++ b/docs/reference/api/doxygen/data__layout_8h_source.html
@@ -77,7 +77,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1tir_1_1LayoutAxis_html_a5f510601a3a63b5b192278a18347b6a7"><div class="ttname"><a href="classtvm_1_1tir_1_1LayoutAxis.html#a5f510601a3a63b5b192278a18347b6a7">tvm::tir::LayoutAxis::operator<<</a></div><div class="ttdeci">friend std::ostream & operator<<(std::ostream &os, const LayoutAxis &l)</div><div class="ttdef"><b>Definition:</b> data_layout.h:73</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1LayoutNode_html_a4287f43c0ae133205d9803f4fadc3d88"><div class="ttname"><a href="classtvm_1_1tir_1_1LayoutNode.html#a4287f43c0ae133205d9803f4fadc3d88">tvm::tir::LayoutNode::axes</a></div><div class="ttdeci">Array< tir::IterVar > axes</div><div class="ttdoc">specify each axis of the layout, in which the variable name is the name of the axis. The IterVar&#39;s extent indicates the size of the axis, it is a variable for a primal axis, but a c [...]
<div class="ttc" id="classtvm_1_1tir_1_1VarNode_html_aefe69c9361d4ab7282a9c07efebe4414"><div class="ttname"><a href="classtvm_1_1tir_1_1VarNode.html#aefe69c9361d4ab7282a9c07efebe4414">tvm::tir::VarNode::name_hint</a></div><div class="ttdeci">String name_hint</div><div class="ttdoc">The hint to the variable name. </div><div class="ttdef"><b>Definition:</b> var.h:53</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="namespacetvm_1_1tir_1_1builtin_html_a77a72a2e21fe0ea2118479924b4fb877"><div class="ttname"><a href="namespacetvm_1_1tir_1_1builtin.html#a77a72a2e21fe0ea2118479924b4fb877">tvm::tir::builtin::undef</a></div><div class="ttdeci">const Op & undef()</div><div class="ttdoc">Returns an initialized but arbitrary value. </div></div>
<div class="ttc" id="namespacestd_html"><div class="ttname"><a href="namespacestd.html">std</a></div><div class="ttdef"><b>Definition:</b> loop_state.h:456</div></div>
<div class="ttc" id="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>
diff --git a/docs/reference/api/doxygen/data__type__rewriter_8h_source.html b/docs/reference/api/doxygen/data__type__rewriter_8h_source.html
index be231f7e0d..9ab9d53c90 100644
--- a/docs/reference/api/doxygen/data__type__rewriter_8h_source.html
+++ b/docs/reference/api/doxygen/data__type__rewriter_8h_source.html
@@ -73,7 +73,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1tir_1_1IndexDataTypeNormalizer_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IndexDataTypeNormalizer.html">tvm::tir::IndexDataTypeNormalizer</a></div><div class="ttdoc">Normalize the data types of buffer shapes and indices to the same data type. </div><div class="ttdef"><b>Definition:</b> data_type_rewriter.h:140</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1IndexDataTypeRewriter_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IndexDataTypeRewriter.html">tvm::tir::IndexDataTypeRewriter</a></div><div class="ttdoc">Data type rewriter for buffer indices. </div><div class="ttdef"><b>Definition:</b> data_type_rewriter.h:97</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1IfThenElseNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IfThenElseNode.html">tvm::tir::IfThenElseNode</a></div><div class="ttdoc">IfThenElse statment. </div><div class="ttdef"><b>Definition:</b> stmt.h:830</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1VarNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1VarNode.html">tvm::tir::VarNode</a></div><div class="ttdoc">A variable node in the IR. </div><div class="ttdef"><b>Definition:</b> var.h:47</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1MulNode_html"><div class="ttname"><a href="classtvm_1_1tir_1_1MulNode.html">tvm::tir::MulNode</a></div><div class="ttdoc">a * b </div><div class="ttdef"><b>Definition:</b> expr.h:191</div></div>
<div class="ttc" id="classtvm_1_1tir_1_1BufferRegion_html"><div class="ttname"><a href="classtvm_1_1tir_1_1BufferRegion.html">tvm::tir::BufferRegion</a></div><div class="ttdoc">Managed reference to BufferRegionNode. </div><div class="ttdef"><b>Definition:</b> stmt.h:1151</div></div>
diff --git a/docs/reference/api/doxygen/functions_func_s.html b/docs/reference/api/doxygen/functions_func_s.html
index bdf3482a02..5d244e21d7 100644
--- a/docs/reference/api/doxygen/functions_func_s.html
+++ b/docs/reference/api/doxygen/functions_func_s.html
@@ -643,7 +643,7 @@ $(function() {
, <a class="el" href="classtvm_1_1tir_1_1SeqStmtNode.html#a0b17eae7cf993b7e3c3e5b50b4fa5982">tvm::tir::SeqStmtNode</a>
</li>
<li>SizeVar()
-: <a class="el" href="classtvm_1_1tir_1_1SizeVar.html#a0f8cb8a92feb96343939d223db90f7cd">tvm::tir::SizeVar</a>
+: <a class="el" href="classtvm_1_1tir_1_1SizeVar.html#ac470249315d9e395ad581d35dd5dcb05">tvm::tir::SizeVar</a>
</li>
<li>Slice()
: <a class="el" href="classtvm_1_1te_1_1Tensor_1_1Slice.html#ab314819e8bcca6421e9a4f33e48578c3">tvm::te::Tensor::Slice</a>
@@ -655,7 +655,7 @@ $(function() {
: <a class="el" href="classtvm_1_1Source.html#a794f5989e4e89cb39b638a5cd837eaa9">tvm::Source</a>
</li>
<li>SourceMap()
-: <a class="el" href="classtvm_1_1SourceMap.html#a9f10049893326844c3f01daad7c121e9">tvm::SourceMap</a>
+: <a class="el" href="classtvm_1_1SourceMap.html#ad4517cedaea581d34c28cb9903205eeb">tvm::SourceMap</a>
</li>
<li>SpaceGenerator()
: <a class="el" href="classtvm_1_1meta__schedule_1_1SpaceGenerator.html#a4f11a841acdf7a66893eeefc1feb04ae">tvm::meta_schedule::SpaceGenerator</a>
@@ -665,7 +665,7 @@ $(function() {
</li>
<li>Span()
: <a class="el" href="classtvm_1_1Span.html#a5216631b639e8c802263d87d3fe9e5f6">tvm::Span</a>
-, <a class="el" href="classtvm_1_1support_1_1Span.html#a3c22dd06856e7029e7107adf38eb72f5">tvm::support::Span< T, W ></a>
+, <a class="el" href="classtvm_1_1support_1_1Span.html#a77653730a2542edf93b7c4413a72f3ec">tvm::support::Span< T, W ></a>
</li>
<li>SpecializedCondition()
: <a class="el" href="classtvm_1_1te_1_1SpecializedCondition.html#a48d119ee1c6033929a5592cfc2592e60">tvm::te::SpecializedCondition</a>
@@ -693,8 +693,8 @@ $(function() {
: <a class="el" href="classtvm_1_1runtime_1_1String.html#a88aa19338db83ae838183e7478f09e60">tvm::runtime::String</a>
</li>
<li>Stage()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#a39ffbb1b4e189180bc4067e74965f42b">tvm::auto_scheduler::Stage</a>
-, <a class="el" href="classtvm_1_1te_1_1Stage.html#a1ecdc9a000be62c9cc26a96d4c33e36e">tvm::te::Stage</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#af0643fe8c1298451c9a322f915c48843">tvm::auto_scheduler::Stage</a>
+, <a class="el" href="classtvm_1_1te_1_1Stage.html#aa6ace38b6312e42aaf9389c8749ae0a4">tvm::te::Stage</a>
</li>
<li>Start()
: <a class="el" href="classtvm_1_1runtime_1_1profiling_1_1MetricCollectorNode.html#a44fadfb7b0f961a7fb2275e3b5dbcd88">tvm::runtime::profiling::MetricCollectorNode</a>
@@ -775,7 +775,7 @@ $(function() {
: <a class="el" href="classtvm_1_1runtime_1_1DeviceAPI.html#ac29b9295c432a87658392872c644864f">tvm::runtime::DeviceAPI</a>
</li>
<li>String()
-: <a class="el" href="classtvm_1_1runtime_1_1String.html#a02fca36e3ff55cc1e83635b02a11fca3">tvm::runtime::String</a>
+: <a class="el" href="classtvm_1_1runtime_1_1String.html#a68df7bab89fca339e3918438dd80300d">tvm::runtime::String</a>
</li>
<li>StringImm()
: <a class="el" href="classtvm_1_1tir_1_1StringImm.html#a0f2830290e055f677c5d5dea98aab726">tvm::tir::StringImm</a>
diff --git a/docs/reference/api/doxygen/functions_s.html b/docs/reference/api/doxygen/functions_s.html
index 018d2c76e2..1649079561 100644
--- a/docs/reference/api/doxygen/functions_s.html
+++ b/docs/reference/api/doxygen/functions_s.html
@@ -915,7 +915,7 @@ $(function() {
: <a class="el" href="classtvm_1_1te_1_1Stage.html#a51432f38d9ec4792a2525023179ae604">tvm::te::Stage</a>
</li>
<li>SplitStep()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1SplitStep.html#a184575a8029d77f7a3bee23d81141df5">tvm::auto_scheduler::SplitStep</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1SplitStep.html#a64ed86582a56a2645b3e4eb44ecb31af">tvm::auto_scheduler::SplitStep</a>
</li>
<li>src
: <a class="el" href="classtvm_1_1arith_1_1IntConstraintsTransformNode.html#a8ce159fc6db748e5092fa937de3fde53">tvm::arith::IntConstraintsTransformNode</a>
@@ -949,8 +949,8 @@ $(function() {
: <a class="el" href="classtvm_1_1runtime_1_1String.html#a88aa19338db83ae838183e7478f09e60">tvm::runtime::String</a>
</li>
<li>Stage()
-: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#a39ffbb1b4e189180bc4067e74965f42b">tvm::auto_scheduler::Stage</a>
-, <a class="el" href="classtvm_1_1te_1_1Stage.html#a1ecdc9a000be62c9cc26a96d4c33e36e">tvm::te::Stage</a>
+: <a class="el" href="classtvm_1_1auto__scheduler_1_1Stage.html#af0643fe8c1298451c9a322f915c48843">tvm::auto_scheduler::Stage</a>
+, <a class="el" href="classtvm_1_1te_1_1Stage.html#aa6ace38b6312e42aaf9389c8749ae0a4">tvm::te::Stage</a>
</li>
<li>stage_id
: <a class="el" href="classtvm_1_1auto__scheduler_1_1StepNode.html#afcc7aaf263348f66139307affbfcee09">tvm::auto_scheduler::StepNode</a>
@@ -1033,7 +1033,7 @@ $(function() {
: <a class="el" href="classtvm_1_1script_1_1printer_1_1StmtDoc.html#adec8d59e41d8a4093fb310089bf2c3ba">tvm::script::printer::StmtDoc</a>
</li>
<li>StmtNode()
-: <a class="el" href="classtvm_1_1tir_1_1StmtNode.html#a67693c4e97ae49890ea74605fe1b1f74">tvm::tir::StmtNode</a>
+: <a class="el" href="classtvm_1_1tir_1_1StmtNode.html#a79e21b14d3ab57209577bf4a8f694a87">tvm::tir::StmtNode</a>
</li>
<li>stmts
: <a class="el" href="classtvm_1_1script_1_1ir__builder_1_1tir_1_1TIRFrameNode.html#a13776bb5c2e5403138fbee06d4fdad40">tvm::script::ir_builder::tir::TIRFrameNode</a>
diff --git a/docs/reference/api/doxygen/ir_2type_8h.html b/docs/reference/api/doxygen/ir_2type_8h.html
index f9ba355a3c..316e4e9df3 100644
--- a/docs/reference/api/doxygen/ir_2type_8h.html
+++ b/docs/reference/api/doxygen/ir_2type_8h.html
@@ -194,7 +194,7 @@ Functions</h2></td></tr>
<p>This file contains types that are common across IR variants.</p>
<h2>Relation between Type and runtime::DataType</h2>
<p>Besides Type, we also store a dtype field in the low-level PrimExpr. runtime::DataType(dtype) provides coarse grained type information during compile time and runtime. It is eagerly built in low-level expression construction and can be used for quick type checking in the low-level IR. For example, when an Expr's dtype is int32, we know for sure that its type is also int32.</p>
-<p>On the other hand, Type provides more fine grained information. For example, a low level expression can have <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a939b10df4c1a607526694b98ddbf8fd7" title="Create a TIR var that represents a pointer. ">DataType::Handle()</a> as its dtype and MemRef[float32] as its type. Types are usually lazily constructed via type checking, so they may not readily be available during IR construction.</p>
+<p>On the other hand, Type provides more fine grained information. For example, a low level expression can have <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaa2a6b8cd99890d0966ce5c83eb0ef77" title="Create a TIR var that represents a pointer. ">DataType::Handle()</a> as its dtype and MemRef[float32] as its type. Types are usually lazily constructed via type checking, so they may not readily be available during IR construction.</p>
<p>The unified Type serves as a common bridge across IR dialects. For example, we require all the functions to have a type signature, which allow us to build cross dialect function calls. </p>
</div></div><!-- contents -->
<!-- start footer part -->
diff --git a/docs/reference/api/doxygen/map_8h_source.html b/docs/reference/api/doxygen/map_8h_source.html
index 40f9ea3a1e..f2f973e5ba 100644
--- a/docs/reference/api/doxygen/map_8h_source.html
+++ b/docs/reference/api/doxygen/map_8h_source.html
@@ -107,7 +107,7 @@ $(function() {
<div class="ttc" id="classtvm_1_1runtime_1_1MapNode_html_a5c0c770f7667f911aa8bec879e3ac214"><div class="ttname"><a href="classtvm_1_1runtime_1_1MapNode.html#a5c0c770f7667f911aa8bec879e3ac214">tvm::runtime::MapNode::size</a></div><div class="ttdeci">size_t size() const</div><div class="ttdoc">Number of elements in the SmallMapNode. </div><div class="ttdef"><b>Definition:</b> map.h:196</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1DenseMapNode_html_a6071908cdeb00617d3b28a70d05ac649"><div class="ttname"><a href="classtvm_1_1runtime_1_1DenseMapNode.html#a6071908cdeb00617d3b28a70d05ac649">tvm::runtime::DenseMapNode::at</a></div><div class="ttdeci">const mapped_type & at(const key_type &key) const</div><div class="ttdoc">Index value associated with a key, throw exception if the key does not exist. </div><div class="ttdef"><b>Definition:</b> map.h:606</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1MapNode_html_abc18d8e58770915331c3257ebc80eadc"><div class="ttname"><a href="classtvm_1_1runtime_1_1MapNode.html#abc18d8e58770915331c3257ebc80eadc">tvm::runtime::MapNode::TVM_DECLARE_FINAL_OBJECT_INFO</a></div><div class="ttdeci">TVM_DECLARE_FINAL_OBJECT_INFO(MapNode, Object)</div></div>
-<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:301</div></div>
+<div class="ttc" id="classtvm_1_1tir_1_1IterVar_html"><div class="ttname"><a href="classtvm_1_1tir_1_1IterVar.html">tvm::tir::IterVar</a></div><div class="ttdoc">Iteration Variable, represents an iteration over an integer interval. </div><div class="ttdef"><b>Definition:</b> var.h:308</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Map_html_ac40191ef3c2de0c546b48102fa43cd88"><div class="ttname"><a href="classtvm_1_1runtime_1_1Map.html#ac40191ef3c2de0c546b48102fa43cd88">tvm::runtime::Map::Map</a></div><div class="ttdeci">Map(const std::unordered_map< K, V, Hash, Equal > &init)</div><div class="ttdoc">constructor from unordered_map </div><div class="ttdef"><b>Definition:</b> map.h:1335</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1Map_1_1iterator_html_a8052ae36e24a4973c1a123c99cf5152c"><div class="ttname"><a href="classtvm_1_1runtime_1_1Map_1_1iterator.html#a8052ae36e24a4973c1a123c99cf5152c">tvm::runtime::Map::iterator::operator++</a></div><div class="ttdeci">iterator & operator++()</div><div class="ttdoc">Prefix self increment, e.g. ++iter. </div><div class="ttdef"><b>Definition:</b> map.h:1436</div></div>
<div class="ttc" id="classtvm_1_1runtime_1_1MapNode_html_ad486baa9df3b8061218bbad6cea53df9"><div class="ttname"><a href="classtvm_1_1runtime_1_1MapNode.html#ad486baa9df3b8061218bbad6cea53df9">tvm::runtime::MapNode::Empty</a></div><div class="ttdeci">static ObjectPtr< MapNode > Empty()</div><div class="ttdoc">Create an empty container. </div><div class="ttdef"><b>Definition:</b> map.h:1202</div></div>
diff --git a/docs/reference/api/doxygen/namespacemembers_b.html b/docs/reference/api/doxygen/namespacemembers_b.html
index 883be4ccba..22cb0f1655 100644
--- a/docs/reference/api/doxygen/namespacemembers_b.html
+++ b/docs/reference/api/doxygen/namespacemembers_b.html
@@ -136,7 +136,7 @@ $(function() {
: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac547caf48dbfd8f4b8072dcb2c3af683">tvm::script::ir_builder::tir</a>
</li>
<li>Boolean()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a95b135dca5d84261342b70d3d8a34acc">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a25dce58350b5f2cb65d2367f16fec4b6">tvm::script::ir_builder::tir</a>
</li>
<li>BoundTypeVars()
: <a class="el" href="namespacetvm_1_1relay.html#a1e5aa65b13f8ca172009aa2ff3ba59d6">tvm::relay</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_f.html b/docs/reference/api/doxygen/namespacemembers_f.html
index d56748bf1c..afc9383687 100644
--- a/docs/reference/api/doxygen/namespacemembers_f.html
+++ b/docs/reference/api/doxygen/namespacemembers_f.html
@@ -185,76 +185,76 @@ $(function() {
: <a class="el" href="namespacetvm_1_1relay.html#a926750aeda977340aa82bceac863fc9eaff508e399703958af46f99ee8eb7d1a5">tvm::relay</a>
</li>
<li>Float16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaf6b6dcbbd344633ac7549017151b1f4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab878e27ae0a3913e1d58948d47cc3818">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af233b18ee0522a900568b14b6f9ef8d6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a84ffb192123d18c7aeb22f0072847d43">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad5311856f8b8b4df884170d8c4c513c8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6970dcedb53382eb2de268867c9a1b69">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6843af0b415dd3c3555b176cd3233b6b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5cb22dc9a696dba7cb66574ef0e253ce">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac42c4020fca22350e8159b86cefb4031">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a741dd7d66effeba15af9b539561b2ded">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1f72082ac88788080f123bf8c93500e6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a037c836af7ac994e2ffe76300ed6e1b7">tvm::script::ir_builder::tir</a>
</li>
<li>Float32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8878f384c4a2861add3f43da4aa7e080">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad0b06e1dabf3f744a991164800d50a0a">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb206b62b60473442d995a4f868f57f1">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abfc76052bf4e687fb91c903cb97dcd83">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad859865f93db0e8b124464203253fdab">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a23c31f9b9908cc3c016592111f5d5860">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afc23e87468bd379cef2e436dec7380ec">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab61768a121397418c43f7d4170c0c5da">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad363b06dcc7a557663219774aae3ae59">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af101d8c641d0f61ce35974d5fe3ef00a">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae6706e8560ac719f2957e0186de014b3">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab88e3bd2be02f2be5805c4cd5e0bf620">tvm::script::ir_builder::tir</a>
</li>
<li>Float64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6ff858a2cfb49c10fc00ae68a1728086">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f5da5cec8dae24b9c7f94cbfd9ff9ca">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acccc0b5d430d1205b3d1a782cab67aae">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71b08acd9bf3b1eb73f83b4edd5a5182">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae07732bbb9fea30ff6b9c03a7150beeb">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab0f70432f1d4056ab3a434e3259b3e9e">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4051e0d30c5fe712f65fffb8eee90d29">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae86a4246835fe2a1356a0ce18fc7a9e0">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab2aebc810c46e788e51c8e01a9eb7536">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aad6a7dc76ebbdcdb5eed91efc07fbff8">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4bb3d4dabf70259c2820a0f36c9e5661">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4022cf159b063cb6671f0331ae33792a">tvm::script::ir_builder::tir</a>
</li>
<li>Float8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8f02a339b55424163e97c46f1c9f7976">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0b31ddab3910ef43268f150db4e2266c">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac5b707e9c9bd8ff20640e9aa5609f7ad">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6735673e92f3697be33f5cf311c06f9d">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0c1842297c526585f5e7ecd5cf9fa835">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a98a7d7a7f408fec2091af9d2195c655d">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9d6a011a6d9a84e818bbed17821fca43">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2e3565ed4e1713896da4ee27b3317757">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9493b11608cc2420b3e3780726556f0c">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8cff6e0d907d3c4c19fea0a88e9bedf0">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8d984cfc2c4f432a5e2bb16ed9723c9c">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aedf5c6024355c112045906632353bcfa">tvm::script::ir_builder::tir</a>
</li>
<li>FloatImmNode
: <a class="el" href="namespacetvm_1_1tir.html#adb123be81447d2871f4c716ce2d1dc00">tvm::tir</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_b.html b/docs/reference/api/doxygen/namespacemembers_func_b.html
index 303d960b29..df0589ff90 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_b.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_b.html
@@ -127,7 +127,7 @@ $(function() {
: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac547caf48dbfd8f4b8072dcb2c3af683">tvm::script::ir_builder::tir</a>
</li>
<li>Boolean()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a95b135dca5d84261342b70d3d8a34acc">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a25dce58350b5f2cb65d2367f16fec4b6">tvm::script::ir_builder::tir</a>
</li>
<li>BoundTypeVars()
: <a class="el" href="namespacetvm_1_1relay.html#a1e5aa65b13f8ca172009aa2ff3ba59d6">tvm::relay</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_f.html b/docs/reference/api/doxygen/namespacemembers_func_f.html
index f971d14b86..3ef958882a 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_f.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_f.html
@@ -119,76 +119,76 @@ $(function() {
: <a class="el" href="namespacetvm_1_1relay.html#a91168c24ecbf1cbf719d2ba28841da5c">tvm::relay</a>
</li>
<li>Float16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaf6b6dcbbd344633ac7549017151b1f4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab878e27ae0a3913e1d58948d47cc3818">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af233b18ee0522a900568b14b6f9ef8d6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a84ffb192123d18c7aeb22f0072847d43">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad5311856f8b8b4df884170d8c4c513c8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6970dcedb53382eb2de268867c9a1b69">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6843af0b415dd3c3555b176cd3233b6b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5cb22dc9a696dba7cb66574ef0e253ce">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac42c4020fca22350e8159b86cefb4031">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a741dd7d66effeba15af9b539561b2ded">tvm::script::ir_builder::tir</a>
</li>
<li>Float16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1f72082ac88788080f123bf8c93500e6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a037c836af7ac994e2ffe76300ed6e1b7">tvm::script::ir_builder::tir</a>
</li>
<li>Float32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8878f384c4a2861add3f43da4aa7e080">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad0b06e1dabf3f744a991164800d50a0a">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb206b62b60473442d995a4f868f57f1">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abfc76052bf4e687fb91c903cb97dcd83">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad859865f93db0e8b124464203253fdab">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a23c31f9b9908cc3c016592111f5d5860">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afc23e87468bd379cef2e436dec7380ec">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab61768a121397418c43f7d4170c0c5da">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad363b06dcc7a557663219774aae3ae59">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af101d8c641d0f61ce35974d5fe3ef00a">tvm::script::ir_builder::tir</a>
</li>
<li>Float32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae6706e8560ac719f2957e0186de014b3">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab88e3bd2be02f2be5805c4cd5e0bf620">tvm::script::ir_builder::tir</a>
</li>
<li>Float64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6ff858a2cfb49c10fc00ae68a1728086">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f5da5cec8dae24b9c7f94cbfd9ff9ca">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acccc0b5d430d1205b3d1a782cab67aae">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71b08acd9bf3b1eb73f83b4edd5a5182">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae07732bbb9fea30ff6b9c03a7150beeb">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab0f70432f1d4056ab3a434e3259b3e9e">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4051e0d30c5fe712f65fffb8eee90d29">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae86a4246835fe2a1356a0ce18fc7a9e0">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab2aebc810c46e788e51c8e01a9eb7536">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aad6a7dc76ebbdcdb5eed91efc07fbff8">tvm::script::ir_builder::tir</a>
</li>
<li>Float64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4bb3d4dabf70259c2820a0f36c9e5661">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4022cf159b063cb6671f0331ae33792a">tvm::script::ir_builder::tir</a>
</li>
<li>Float8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8f02a339b55424163e97c46f1c9f7976">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0b31ddab3910ef43268f150db4e2266c">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac5b707e9c9bd8ff20640e9aa5609f7ad">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6735673e92f3697be33f5cf311c06f9d">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0c1842297c526585f5e7ecd5cf9fa835">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a98a7d7a7f408fec2091af9d2195c655d">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9d6a011a6d9a84e818bbed17821fca43">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2e3565ed4e1713896da4ee27b3317757">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9493b11608cc2420b3e3780726556f0c">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8cff6e0d907d3c4c19fea0a88e9bedf0">tvm::script::ir_builder::tir</a>
</li>
<li>Float8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8d984cfc2c4f432a5e2bb16ed9723c9c">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aedf5c6024355c112045906632353bcfa">tvm::script::ir_builder::tir</a>
</li>
<li>floor()
: <a class="el" href="namespacetvm.html#aaff65dde3044433b2220677aedf4855f">tvm</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_h.html b/docs/reference/api/doxygen/namespacemembers_func_h.html
index 87b4cdaa8e..e02ec22300 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_h.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_h.html
@@ -62,7 +62,7 @@ $(function() {
<h3><a id="index_h"></a>- h -</h3><ul>
<li>Handle()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a939b10df4c1a607526694b98ddbf8fd7">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaa2a6b8cd99890d0966ce5c83eb0ef77">tvm::script::ir_builder::tir</a>
</li>
<li>HillClimb()
: <a class="el" href="namespacetvm_1_1tir_1_1usmp_1_1algo.html#ad7f4265f15be71c0db15d9fff4272148">tvm::tir::usmp::algo</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_i.html b/docs/reference/api/doxygen/namespacemembers_func_i.html
index 4262103338..b5829e1c5d 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_i.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_i.html
@@ -136,76 +136,76 @@ $(function() {
: <a class="el" href="namespacetvm_1_1tir_1_1transform.html#a2ac039d4db9a6821d8a3e0b0191e7e20">tvm::tir::transform</a>
</li>
<li>Int16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71a8950ee36a0cc96b08f6a38a67beb8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae8f12ddc9fce3fea6fe5dfe870a8dc6b">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a384760861c5679c8f3228fb069c4525b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4ab76571223be0c914419c09d0e5fbf5">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a535df06d8e6eec52c711be076ca919da">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aab48e91f64d8c71e8687517920295583">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aee6b73b52838f0921c9dae3a6820b9a6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae49920dbdafc3d66234dd5b9815efa38">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af817e48b37ce691b1538708ee4a96ae4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2ea4609ed809b61fc078d3cbfbb588bf">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a42335d7e171d3a46711e43ffd241e68d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa85096ef46c3453be8c0e446a5e6fcd3">tvm::script::ir_builder::tir</a>
</li>
<li>Int32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af90ae386a2b910cc52c17ebb6390404e">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a88c9e3f761e2247a0518161db6a8358c">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4945336e84e5ce9b60b583ac0adb28d6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e64669b1260b85c4e1722d1b98afc17">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f621b090a72a31cf4b80978d88ccf76">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae800a458fed6eb05504f889a7f94b6b5">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1d7bdb8458778293ced5af6ad9378283">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a99e797d54d8ffcbf02d3a220cf86489a">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a57b983ed9d4d5000d0372e64e526f9e0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2f7c6be2d1881aa9cbb6e5d5579a8fcc">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae89c85eabf166d4efddff53988ff57b5">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5e81ae25cecfd56e2f3b474063324c3a">tvm::script::ir_builder::tir</a>
</li>
<li>Int64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e3533536d109d1a5dd037c356ac2662">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5be7437e66fa41a4c249f81a7efe7470">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3c2d6fdb8690f5593c1d43b0cdf43d4e">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a682890d013c50542ded6cfdad936f88f">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e047daab47ebe417cdda5e3209607af">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad828ae492de313cf743fbaa293189cba">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa84d61b0037299683391cfb1792064f9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aae04dd74f177452906e9832e3e41da1c">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a61b20548797f6cbb80de229afe1cce53">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a439334cb39d3e5877553053b51973a10">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a972550fc782bca80bfc88281e88d6022">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a85202b4d4ad9b820b7ac1361905e233c">tvm::script::ir_builder::tir</a>
</li>
<li>Int8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8e3f67af7d6676f0980fa5fa6406ee7">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7811d90b33a2e5d399003f735d11ba3">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4738b9544893c5f7992efefd8959af1d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8b37883752e594abbfc23bec33a5c44">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8092c0c7d3e9c111c2c247cae4eda1e9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab5ab0626b32c85ed15634859354bae60">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8324f437b88e52d038fb35b35ee094a0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa5dd0f6b10d8a8512193987ba9c1564d">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6750fab4bf57aeb380dabe6fce99c5a4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac76b90af462de39cffd71f7a08340ec5">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac575b8dc0c148ea90512e7443fde3073">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9a74627806410f80db60b17bee94cbc4">tvm::script::ir_builder::tir</a>
</li>
<li>Intersect()
: <a class="el" href="namespacetvm_1_1arith.html#ad66a987ddb3d8aee69f9175bf412eb8a">tvm::arith</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_u.html b/docs/reference/api/doxygen/namespacemembers_func_u.html
index e19ae131b2..507483552f 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_u.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_u.html
@@ -62,76 +62,76 @@ $(function() {
<h3><a id="index_u"></a>- u -</h3><ul>
<li>UInt16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac18cfb33b4b074bebb62a0022e3fddd9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7a3cd3dbea53feb2e7043e6452260d8a">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5505117dc914ce1bfe08fe140134dbc9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4e2e296491ba53b33d0555f9edd7b9d2">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a22c54a2af8641b558a1ad2895acb3005">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0dbc9e2fb673e412abfb1c77d49a053d">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a31de2e593c18ff9cff149cc1c5f7ebb8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae3b27f43f0de67f3e5ba799888bbefa7">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afcd528fd1230b26fce48701ce44619e3">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2a0336d85cc5c8061fe29886dc871b2b">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0110fa08ba2f1b0d29292e4dc6d6f41d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb9032673c555649c524dcb423bd3ec9">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7dab1502059beec51115d2fc80d7467">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad46043363fa7a3d8cc4b53bd69c5c57a">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab1a1aa826e5d680ad549247c55437325">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada1efb68ea879efc0ecc27196a3068a4">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5a86c4391013e0b0791fa8d74e925f8f">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a17c245f82bc5002bbe5eb9c225325a85">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79e5972bf85473593165a557c7ac1ded">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9492cfc58991919e32a9a9653a0544a6">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1895112a229f0e1129328368f8124361">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a096f4ffaf512c1fefa0ceaefad55e593">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3e9853322f7173158dba8bea26cfeefc">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac81a6cbb9e308a006df32b073e05098f">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#addcd399b813b5989ff7c45a50b9a0925">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a11d2e4a9eba311bd85a70cc8793aaf75">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a998c1aa94e4462047190261120acdae9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a33feb22ecf4d6708312e93a950e43637">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6b7b72765526b8f96cf3291f5e1c083b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a625cdc68ea89d04b9bf314c28f397234">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaef8e2bd7738e488d4fac10c757a3acf">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afa4e70a7e0ad235f3d2a1160c0093625">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4b6ead30ca701c211d3eac598d86b25a">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abdc950ea7ac026f00b8858e86b0fd81b">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2fcd410ad9a6e3132821db76bf7eb537">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a920b270cb2b11f7ce4fee459f7f1a619">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79bce011a3c96be3481e02fa595026e0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae492cc7293526e8e42ad4bf2b4a64cc0">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4c909574d4809116f228e6212b13ec56">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#affd5053dacbb25688ef6f067bbb78761">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa9ab96e8f14b6ee6b140bd3b7261fdd9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abbe8c3cb652b5fe7948080f8caf1570c">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab796b555e2ec6855db23676cdde3bb2d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab84a0811f857d82faecf6c9cbbad5780">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a297d6deaea7f8520c75b70828bb017cb">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a97cd0230ef0458bc166a387d49fa0250">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada38005ba6aef4e9dbae3028c1e40dd4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a227658e263028f9233da9f55c248a43f">tvm::script::ir_builder::tir</a>
</li>
<li>UnCPS()
: <a class="el" href="namespacetvm_1_1relay.html#a1677d450d58078811f52fd70d67db515">tvm::relay</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_func_v.html b/docs/reference/api/doxygen/namespacemembers_func_v.html
index 0192b3590e..124835a354 100644
--- a/docs/reference/api/doxygen/namespacemembers_func_v.html
+++ b/docs/reference/api/doxygen/namespacemembers_func_v.html
@@ -108,7 +108,7 @@ $(function() {
: <a class="el" href="namespacetvm_1_1tir.html#a1f69889b13455cd290e6b1a6453e3c1d">tvm::tir</a>
</li>
<li>Void()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afbc8756e7d3f4322e8e4506f6a0698a5">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2b6c9317de9b507692d67a8e77dafc94">tvm::script::ir_builder::tir</a>
</li>
<li>VoidType()
: <a class="el" href="namespacetvm.html#a27df956aaa2e5e25005e0f9dc1638a08">tvm</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_h.html b/docs/reference/api/doxygen/namespacemembers_h.html
index 75eef5ce89..a54d1d2463 100644
--- a/docs/reference/api/doxygen/namespacemembers_h.html
+++ b/docs/reference/api/doxygen/namespacemembers_h.html
@@ -65,7 +65,7 @@ $(function() {
: <a class="el" href="namespacetvm_1_1tir_1_1attr.html#a0d026645d3f86d9cc2e693fa232fddec">tvm::tir::attr</a>
</li>
<li>Handle()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a939b10df4c1a607526694b98ddbf8fd7">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaa2a6b8cd99890d0966ce5c83eb0ef77">tvm::script::ir_builder::tir</a>
</li>
<li>HillClimb()
: <a class="el" href="namespacetvm_1_1tir_1_1usmp_1_1algo.html#ad7f4265f15be71c0db15d9fff4272148">tvm::tir::usmp::algo</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_i.html b/docs/reference/api/doxygen/namespacemembers_i.html
index 153044f908..009f99dd54 100644
--- a/docs/reference/api/doxygen/namespacemembers_i.html
+++ b/docs/reference/api/doxygen/namespacemembers_i.html
@@ -148,76 +148,76 @@ $(function() {
: <a class="el" href="namespacetvm_1_1tir_1_1transform.html#a2ac039d4db9a6821d8a3e0b0191e7e20">tvm::tir::transform</a>
</li>
<li>Int16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71a8950ee36a0cc96b08f6a38a67beb8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae8f12ddc9fce3fea6fe5dfe870a8dc6b">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a384760861c5679c8f3228fb069c4525b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4ab76571223be0c914419c09d0e5fbf5">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a535df06d8e6eec52c711be076ca919da">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aab48e91f64d8c71e8687517920295583">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aee6b73b52838f0921c9dae3a6820b9a6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae49920dbdafc3d66234dd5b9815efa38">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af817e48b37ce691b1538708ee4a96ae4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2ea4609ed809b61fc078d3cbfbb588bf">tvm::script::ir_builder::tir</a>
</li>
<li>Int16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a42335d7e171d3a46711e43ffd241e68d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa85096ef46c3453be8c0e446a5e6fcd3">tvm::script::ir_builder::tir</a>
</li>
<li>Int32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af90ae386a2b910cc52c17ebb6390404e">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a88c9e3f761e2247a0518161db6a8358c">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4945336e84e5ce9b60b583ac0adb28d6">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e64669b1260b85c4e1722d1b98afc17">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f621b090a72a31cf4b80978d88ccf76">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae800a458fed6eb05504f889a7f94b6b5">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1d7bdb8458778293ced5af6ad9378283">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a99e797d54d8ffcbf02d3a220cf86489a">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a57b983ed9d4d5000d0372e64e526f9e0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2f7c6be2d1881aa9cbb6e5d5579a8fcc">tvm::script::ir_builder::tir</a>
</li>
<li>Int32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae89c85eabf166d4efddff53988ff57b5">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5e81ae25cecfd56e2f3b474063324c3a">tvm::script::ir_builder::tir</a>
</li>
<li>Int64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e3533536d109d1a5dd037c356ac2662">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5be7437e66fa41a4c249f81a7efe7470">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3c2d6fdb8690f5593c1d43b0cdf43d4e">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a682890d013c50542ded6cfdad936f88f">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e047daab47ebe417cdda5e3209607af">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad828ae492de313cf743fbaa293189cba">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa84d61b0037299683391cfb1792064f9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aae04dd74f177452906e9832e3e41da1c">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a61b20548797f6cbb80de229afe1cce53">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a439334cb39d3e5877553053b51973a10">tvm::script::ir_builder::tir</a>
</li>
<li>Int64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a972550fc782bca80bfc88281e88d6022">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a85202b4d4ad9b820b7ac1361905e233c">tvm::script::ir_builder::tir</a>
</li>
<li>Int8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8e3f67af7d6676f0980fa5fa6406ee7">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7811d90b33a2e5d399003f735d11ba3">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4738b9544893c5f7992efefd8959af1d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8b37883752e594abbfc23bec33a5c44">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8092c0c7d3e9c111c2c247cae4eda1e9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab5ab0626b32c85ed15634859354bae60">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8324f437b88e52d038fb35b35ee094a0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa5dd0f6b10d8a8512193987ba9c1564d">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6750fab4bf57aeb380dabe6fce99c5a4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac76b90af462de39cffd71f7a08340ec5">tvm::script::ir_builder::tir</a>
</li>
<li>Int8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac575b8dc0c148ea90512e7443fde3073">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9a74627806410f80db60b17bee94cbc4">tvm::script::ir_builder::tir</a>
</li>
<li>Intersect()
: <a class="el" href="namespacetvm_1_1arith.html#ad66a987ddb3d8aee69f9175bf412eb8a">tvm::arith</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_u.html b/docs/reference/api/doxygen/namespacemembers_u.html
index a6ae30b37e..0c084b56dd 100644
--- a/docs/reference/api/doxygen/namespacemembers_u.html
+++ b/docs/reference/api/doxygen/namespacemembers_u.html
@@ -62,76 +62,76 @@ $(function() {
<h3><a id="index_u"></a>- u -</h3><ul>
<li>UInt16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac18cfb33b4b074bebb62a0022e3fddd9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7a3cd3dbea53feb2e7043e6452260d8a">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5505117dc914ce1bfe08fe140134dbc9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4e2e296491ba53b33d0555f9edd7b9d2">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a22c54a2af8641b558a1ad2895acb3005">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0dbc9e2fb673e412abfb1c77d49a053d">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a31de2e593c18ff9cff149cc1c5f7ebb8">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae3b27f43f0de67f3e5ba799888bbefa7">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afcd528fd1230b26fce48701ce44619e3">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2a0336d85cc5c8061fe29886dc871b2b">tvm::script::ir_builder::tir</a>
</li>
<li>UInt16x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0110fa08ba2f1b0d29292e4dc6d6f41d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb9032673c555649c524dcb423bd3ec9">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7dab1502059beec51115d2fc80d7467">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad46043363fa7a3d8cc4b53bd69c5c57a">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab1a1aa826e5d680ad549247c55437325">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada1efb68ea879efc0ecc27196a3068a4">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5a86c4391013e0b0791fa8d74e925f8f">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a17c245f82bc5002bbe5eb9c225325a85">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79e5972bf85473593165a557c7ac1ded">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9492cfc58991919e32a9a9653a0544a6">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1895112a229f0e1129328368f8124361">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a096f4ffaf512c1fefa0ceaefad55e593">tvm::script::ir_builder::tir</a>
</li>
<li>UInt32x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3e9853322f7173158dba8bea26cfeefc">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac81a6cbb9e308a006df32b073e05098f">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#addcd399b813b5989ff7c45a50b9a0925">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a11d2e4a9eba311bd85a70cc8793aaf75">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a998c1aa94e4462047190261120acdae9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a33feb22ecf4d6708312e93a950e43637">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6b7b72765526b8f96cf3291f5e1c083b">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a625cdc68ea89d04b9bf314c28f397234">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaef8e2bd7738e488d4fac10c757a3acf">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afa4e70a7e0ad235f3d2a1160c0093625">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4b6ead30ca701c211d3eac598d86b25a">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abdc950ea7ac026f00b8858e86b0fd81b">tvm::script::ir_builder::tir</a>
</li>
<li>UInt64x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2fcd410ad9a6e3132821db76bf7eb537">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a920b270cb2b11f7ce4fee459f7f1a619">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79bce011a3c96be3481e02fa595026e0">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae492cc7293526e8e42ad4bf2b4a64cc0">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x16()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4c909574d4809116f228e6212b13ec56">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#affd5053dacbb25688ef6f067bbb78761">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x32()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa9ab96e8f14b6ee6b140bd3b7261fdd9">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abbe8c3cb652b5fe7948080f8caf1570c">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x4()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab796b555e2ec6855db23676cdde3bb2d">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab84a0811f857d82faecf6c9cbbad5780">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x64()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a297d6deaea7f8520c75b70828bb017cb">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a97cd0230ef0458bc166a387d49fa0250">tvm::script::ir_builder::tir</a>
</li>
<li>UInt8x8()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada38005ba6aef4e9dbae3028c1e40dd4">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a227658e263028f9233da9f55c248a43f">tvm::script::ir_builder::tir</a>
</li>
<li>UnCPS()
: <a class="el" href="namespacetvm_1_1relay.html#a1677d450d58078811f52fd70d67db515">tvm::relay</a>
diff --git a/docs/reference/api/doxygen/namespacemembers_v.html b/docs/reference/api/doxygen/namespacemembers_v.html
index 959383200b..fa29c2dfc5 100644
--- a/docs/reference/api/doxygen/namespacemembers_v.html
+++ b/docs/reference/api/doxygen/namespacemembers_v.html
@@ -111,7 +111,7 @@ $(function() {
: <a class="el" href="namespacetvm_1_1tir.html#a1f69889b13455cd290e6b1a6453e3c1d">tvm::tir</a>
</li>
<li>Void()
-: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afbc8756e7d3f4322e8e4506f6a0698a5">tvm::script::ir_builder::tir</a>
+: <a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2b6c9317de9b507692d67a8e77dafc94">tvm::script::ir_builder::tir</a>
</li>
<li>VoidType()
: <a class="el" href="namespacetvm.html#a27df956aaa2e5e25005e0f9dc1638a08">tvm</a>
diff --git a/docs/reference/api/doxygen/namespacetvm_1_1script_1_1ir__builder_1_1tir.html b/docs/reference/api/doxygen/namespacetvm_1_1script_1_1ir__builder_1_1tir.html
index 4b90201c91..5b947185d3 100644
--- a/docs/reference/api/doxygen/namespacetvm_1_1script_1_1ir__builder_1_1tir.html
+++ b/docs/reference/api/doxygen/namespacetvm_1_1script_1_1ir__builder_1_1tir.html
@@ -295,157 +295,157 @@ Functions</h2></td></tr>
<tr class="memitem:a101850df27120c29355e6215b9b2b525"><td class="memItemLeft" align="right" valign="top">void </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a101850df27120c29355e6215b9b2b525">Evaluate</a> (<a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> value)</td></tr>
<tr class="memdesc:a101850df27120c29355e6215b9b2b525"><td class="mdescLeft"> </td><td class="mdescRight">Evaluate the input expression. <a href="#a101850df27120c29355e6215b9b2b525">More...</a><br /></td></tr>
<tr class="separator:a101850df27120c29355e6215b9b2b525"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a939b10df4c1a607526694b98ddbf8fd7"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1tir_1_1Var.html">Var</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a939b10df4c1a607526694b98ddbf8fd7">Handle</a> (<a class="el" href="classtvm_1_1runtime_1_1DataType.html">runtime::DataType</a> dtype=<a class="el" href="classtvm_1_1runtime_1_1DataType.html#ab8dc0832aff8fd7421 [...]
-<tr class="memdesc:a939b10df4c1a607526694b98ddbf8fd7"><td class="mdescLeft"> </td><td class="mdescRight">Create a TIR var that represents a pointer. <a href="#a939b10df4c1a607526694b98ddbf8fd7">More...</a><br /></td></tr>
-<tr class="separator:a939b10df4c1a607526694b98ddbf8fd7"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a8f02a339b55424163e97c46f1c9f7976"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8f02a339b55424163e97c46f1c9f7976">Float8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a8f02a339b55424163e97c46f1c9f7976"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aaf6b6dcbbd344633ac7549017151b1f4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaf6b6dcbbd344633ac7549017151b1f4">Float16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:aaf6b6dcbbd344633ac7549017151b1f4"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a8878f384c4a2861add3f43da4aa7e080"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8878f384c4a2861add3f43da4aa7e080">Float32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a8878f384c4a2861add3f43da4aa7e080"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a6ff858a2cfb49c10fc00ae68a1728086"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6ff858a2cfb49c10fc00ae68a1728086">Float64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a6ff858a2cfb49c10fc00ae68a1728086"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a79bce011a3c96be3481e02fa595026e0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79bce011a3c96be3481e02fa595026e0">UInt8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
-<tr class="separator:a79bce011a3c96be3481e02fa595026e0"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ac18cfb33b4b074bebb62a0022e3fddd9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac18cfb33b4b074bebb62a0022e3fddd9">UInt16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:ac18cfb33b4b074bebb62a0022e3fddd9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aa7dab1502059beec51115d2fc80d7467"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7dab1502059beec51115d2fc80d7467">UInt32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:aa7dab1502059beec51115d2fc80d7467"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:addcd399b813b5989ff7c45a50b9a0925"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#addcd399b813b5989ff7c45a50b9a0925">UInt64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:addcd399b813b5989ff7c45a50b9a0925"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aa8e3f67af7d6676f0980fa5fa6406ee7"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8e3f67af7d6676f0980fa5fa6406ee7">Int8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" hr [...]
-<tr class="separator:aa8e3f67af7d6676f0980fa5fa6406ee7"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a71a8950ee36a0cc96b08f6a38a67beb8"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71a8950ee36a0cc96b08f6a38a67beb8">Int16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
-<tr class="separator:a71a8950ee36a0cc96b08f6a38a67beb8"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:af90ae386a2b910cc52c17ebb6390404e"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af90ae386a2b910cc52c17ebb6390404e">Int32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
-<tr class="separator:af90ae386a2b910cc52c17ebb6390404e"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a7e3533536d109d1a5dd037c356ac2662"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e3533536d109d1a5dd037c356ac2662">Int64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
-<tr class="separator:a7e3533536d109d1a5dd037c356ac2662"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a9d6a011a6d9a84e818bbed17821fca43"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9d6a011a6d9a84e818bbed17821fca43">Float8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a9d6a011a6d9a84e818bbed17821fca43"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a8d984cfc2c4f432a5e2bb16ed9723c9c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8d984cfc2c4f432a5e2bb16ed9723c9c">Float8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a8d984cfc2c4f432a5e2bb16ed9723c9c"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ac5b707e9c9bd8ff20640e9aa5609f7ad"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac5b707e9c9bd8ff20640e9aa5609f7ad">Float8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:ac5b707e9c9bd8ff20640e9aa5609f7ad"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a0c1842297c526585f5e7ecd5cf9fa835"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0c1842297c526585f5e7ecd5cf9fa835">Float8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a0c1842297c526585f5e7ecd5cf9fa835"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a9493b11608cc2420b3e3780726556f0c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9493b11608cc2420b3e3780726556f0c">Float8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a9493b11608cc2420b3e3780726556f0c"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a6843af0b415dd3c3555b176cd3233b6b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6843af0b415dd3c3555b176cd3233b6b">Float16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a6843af0b415dd3c3555b176cd3233b6b"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a1f72082ac88788080f123bf8c93500e6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1f72082ac88788080f123bf8c93500e6">Float16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a1f72082ac88788080f123bf8c93500e6"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:af233b18ee0522a900568b14b6f9ef8d6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af233b18ee0522a900568b14b6f9ef8d6">Float16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:af233b18ee0522a900568b14b6f9ef8d6"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ad5311856f8b8b4df884170d8c4c513c8"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad5311856f8b8b4df884170d8c4c513c8">Float16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ad5311856f8b8b4df884170d8c4c513c8"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ac42c4020fca22350e8159b86cefb4031"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac42c4020fca22350e8159b86cefb4031">Float16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ac42c4020fca22350e8159b86cefb4031"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:afc23e87468bd379cef2e436dec7380ec"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afc23e87468bd379cef2e436dec7380ec">Float32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:afc23e87468bd379cef2e436dec7380ec"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ae6706e8560ac719f2957e0186de014b3"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae6706e8560ac719f2957e0186de014b3">Float32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:ae6706e8560ac719f2957e0186de014b3"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:acb206b62b60473442d995a4f868f57f1"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb206b62b60473442d995a4f868f57f1">Float32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:acb206b62b60473442d995a4f868f57f1"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ad859865f93db0e8b124464203253fdab"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad859865f93db0e8b124464203253fdab">Float32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ad859865f93db0e8b124464203253fdab"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ad363b06dcc7a557663219774aae3ae59"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad363b06dcc7a557663219774aae3ae59">Float32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ad363b06dcc7a557663219774aae3ae59"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4051e0d30c5fe712f65fffb8eee90d29"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4051e0d30c5fe712f65fffb8eee90d29">Float64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a4051e0d30c5fe712f65fffb8eee90d29"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4bb3d4dabf70259c2820a0f36c9e5661"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4bb3d4dabf70259c2820a0f36c9e5661">Float64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a4bb3d4dabf70259c2820a0f36c9e5661"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:acccc0b5d430d1205b3d1a782cab67aae"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acccc0b5d430d1205b3d1a782cab67aae">Float64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:acccc0b5d430d1205b3d1a782cab67aae"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ae07732bbb9fea30ff6b9c03a7150beeb"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae07732bbb9fea30ff6b9c03a7150beeb">Float64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ae07732bbb9fea30ff6b9c03a7150beeb"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ab2aebc810c46e788e51c8e01a9eb7536"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab2aebc810c46e788e51c8e01a9eb7536">Float64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
-<tr class="separator:ab2aebc810c46e788e51c8e01a9eb7536"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ab796b555e2ec6855db23676cdde3bb2d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab796b555e2ec6855db23676cdde3bb2d">UInt8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:ab796b555e2ec6855db23676cdde3bb2d"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ada38005ba6aef4e9dbae3028c1e40dd4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada38005ba6aef4e9dbae3028c1e40dd4">UInt8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:ada38005ba6aef4e9dbae3028c1e40dd4"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4c909574d4809116f228e6212b13ec56"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4c909574d4809116f228e6212b13ec56">UInt8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a4c909574d4809116f228e6212b13ec56"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aa9ab96e8f14b6ee6b140bd3b7261fdd9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa9ab96e8f14b6ee6b140bd3b7261fdd9">UInt8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:aa9ab96e8f14b6ee6b140bd3b7261fdd9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a297d6deaea7f8520c75b70828bb017cb"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a297d6deaea7f8520c75b70828bb017cb">UInt8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a297d6deaea7f8520c75b70828bb017cb"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a31de2e593c18ff9cff149cc1c5f7ebb8"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a31de2e593c18ff9cff149cc1c5f7ebb8">UInt16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a31de2e593c18ff9cff149cc1c5f7ebb8"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a0110fa08ba2f1b0d29292e4dc6d6f41d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0110fa08ba2f1b0d29292e4dc6d6f41d">UInt16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a0110fa08ba2f1b0d29292e4dc6d6f41d"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a5505117dc914ce1bfe08fe140134dbc9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5505117dc914ce1bfe08fe140134dbc9">UInt16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a5505117dc914ce1bfe08fe140134dbc9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a22c54a2af8641b558a1ad2895acb3005"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a22c54a2af8641b558a1ad2895acb3005">UInt16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a22c54a2af8641b558a1ad2895acb3005"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:afcd528fd1230b26fce48701ce44619e3"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afcd528fd1230b26fce48701ce44619e3">UInt16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:afcd528fd1230b26fce48701ce44619e3"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a79e5972bf85473593165a557c7ac1ded"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a79e5972bf85473593165a557c7ac1ded">UInt32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a79e5972bf85473593165a557c7ac1ded"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a3e9853322f7173158dba8bea26cfeefc"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3e9853322f7173158dba8bea26cfeefc">UInt32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a3e9853322f7173158dba8bea26cfeefc"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ab1a1aa826e5d680ad549247c55437325"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab1a1aa826e5d680ad549247c55437325">UInt32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:ab1a1aa826e5d680ad549247c55437325"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a5a86c4391013e0b0791fa8d74e925f8f"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5a86c4391013e0b0791fa8d74e925f8f">UInt32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a5a86c4391013e0b0791fa8d74e925f8f"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a1895112a229f0e1129328368f8124361"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1895112a229f0e1129328368f8124361">UInt32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a1895112a229f0e1129328368f8124361"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aaef8e2bd7738e488d4fac10c757a3acf"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaef8e2bd7738e488d4fac10c757a3acf">UInt64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:aaef8e2bd7738e488d4fac10c757a3acf"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a2fcd410ad9a6e3132821db76bf7eb537"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2fcd410ad9a6e3132821db76bf7eb537">UInt64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a2fcd410ad9a6e3132821db76bf7eb537"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a998c1aa94e4462047190261120acdae9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a998c1aa94e4462047190261120acdae9">UInt64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a998c1aa94e4462047190261120acdae9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a6b7b72765526b8f96cf3291f5e1c083b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6b7b72765526b8f96cf3291f5e1c083b">UInt64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a6b7b72765526b8f96cf3291f5e1c083b"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4b6ead30ca701c211d3eac598d86b25a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4b6ead30ca701c211d3eac598d86b25a">UInt64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
-<tr class="separator:a4b6ead30ca701c211d3eac598d86b25a"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a8324f437b88e52d038fb35b35ee094a0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8324f437b88e52d038fb35b35ee094a0">Int8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a8324f437b88e52d038fb35b35ee094a0"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ac575b8dc0c148ea90512e7443fde3073"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac575b8dc0c148ea90512e7443fde3073">Int8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:ac575b8dc0c148ea90512e7443fde3073"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4738b9544893c5f7992efefd8959af1d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4738b9544893c5f7992efefd8959af1d">Int8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a4738b9544893c5f7992efefd8959af1d"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a8092c0c7d3e9c111c2c247cae4eda1e9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8092c0c7d3e9c111c2c247cae4eda1e9">Int8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a8092c0c7d3e9c111c2c247cae4eda1e9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a6750fab4bf57aeb380dabe6fce99c5a4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6750fab4bf57aeb380dabe6fce99c5a4">Int8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a6750fab4bf57aeb380dabe6fce99c5a4"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aee6b73b52838f0921c9dae3a6820b9a6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aee6b73b52838f0921c9dae3a6820b9a6">Int16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:aee6b73b52838f0921c9dae3a6820b9a6"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a42335d7e171d3a46711e43ffd241e68d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a42335d7e171d3a46711e43ffd241e68d">Int16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a42335d7e171d3a46711e43ffd241e68d"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a384760861c5679c8f3228fb069c4525b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a384760861c5679c8f3228fb069c4525b">Int16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a384760861c5679c8f3228fb069c4525b"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a535df06d8e6eec52c711be076ca919da"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a535df06d8e6eec52c711be076ca919da">Int16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a535df06d8e6eec52c711be076ca919da"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:af817e48b37ce691b1538708ee4a96ae4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af817e48b37ce691b1538708ee4a96ae4">Int16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:af817e48b37ce691b1538708ee4a96ae4"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a1d7bdb8458778293ced5af6ad9378283"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a1d7bdb8458778293ced5af6ad9378283">Int32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a1d7bdb8458778293ced5af6ad9378283"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:ae89c85eabf166d4efddff53988ff57b5"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae89c85eabf166d4efddff53988ff57b5">Int32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:ae89c85eabf166d4efddff53988ff57b5"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a4945336e84e5ce9b60b583ac0adb28d6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4945336e84e5ce9b60b583ac0adb28d6">Int32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a4945336e84e5ce9b60b583ac0adb28d6"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a9f621b090a72a31cf4b80978d88ccf76"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f621b090a72a31cf4b80978d88ccf76">Int32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a9f621b090a72a31cf4b80978d88ccf76"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a57b983ed9d4d5000d0372e64e526f9e0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a57b983ed9d4d5000d0372e64e526f9e0">Int32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a57b983ed9d4d5000d0372e64e526f9e0"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:aa84d61b0037299683391cfb1792064f9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa84d61b0037299683391cfb1792064f9">Int64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:aa84d61b0037299683391cfb1792064f9"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a972550fc782bca80bfc88281e88d6022"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a972550fc782bca80bfc88281e88d6022">Int64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a972550fc782bca80bfc88281e88d6022"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a3c2d6fdb8690f5593c1d43b0cdf43d4e"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a3c2d6fdb8690f5593c1d43b0cdf43d4e">Int64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a3c2d6fdb8690f5593c1d43b0cdf43d4e"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a7e047daab47ebe417cdda5e3209607af"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e047daab47ebe417cdda5e3209607af">Int64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a7e047daab47ebe417cdda5e3209607af"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a61b20548797f6cbb80de229afe1cce53"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a61b20548797f6cbb80de229afe1cce53">Int64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
-<tr class="separator:a61b20548797f6cbb80de229afe1cce53"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:a95b135dca5d84261342b70d3d8a34acc"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a95b135dca5d84261342b70d3d8a34acc">Boolean</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
-<tr class="separator:a95b135dca5d84261342b70d3d8a34acc"><td class="memSeparator" colspan="2"> </td></tr>
-<tr class="memitem:afbc8756e7d3f4322e8e4506f6a0698a5"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afbc8756e7d3f4322e8e4506f6a0698a5">Void</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" hr [...]
-<tr class="separator:afbc8756e7d3f4322e8e4506f6a0698a5"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aaa2a6b8cd99890d0966ce5c83eb0ef77"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1tir_1_1Var.html">Var</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aaa2a6b8cd99890d0966ce5c83eb0ef77">Handle</a> (<a class="el" href="classtvm_1_1runtime_1_1DataType.html">runtime::DataType</a> dtype=<a class="el" href="classtvm_1_1runtime_1_1DataType.html#ab8dc0832aff8fd7421 [...]
+<tr class="memdesc:aaa2a6b8cd99890d0966ce5c83eb0ef77"><td class="mdescLeft"> </td><td class="mdescRight">Create a TIR var that represents a pointer. <a href="#aaa2a6b8cd99890d0966ce5c83eb0ef77">More...</a><br /></td></tr>
+<tr class="separator:aaa2a6b8cd99890d0966ce5c83eb0ef77"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a0b31ddab3910ef43268f150db4e2266c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0b31ddab3910ef43268f150db4e2266c">Float8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a0b31ddab3910ef43268f150db4e2266c"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab878e27ae0a3913e1d58948d47cc3818"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab878e27ae0a3913e1d58948d47cc3818">Float16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ab878e27ae0a3913e1d58948d47cc3818"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ad0b06e1dabf3f744a991164800d50a0a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad0b06e1dabf3f744a991164800d50a0a">Float32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ad0b06e1dabf3f744a991164800d50a0a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a9f5da5cec8dae24b9c7f94cbfd9ff9ca"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9f5da5cec8dae24b9c7f94cbfd9ff9ca">Float64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a9f5da5cec8dae24b9c7f94cbfd9ff9ca"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae492cc7293526e8e42ad4bf2b4a64cc0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae492cc7293526e8e42ad4bf2b4a64cc0">UInt8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
+<tr class="separator:ae492cc7293526e8e42ad4bf2b4a64cc0"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a7a3cd3dbea53feb2e7043e6452260d8a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7a3cd3dbea53feb2e7043e6452260d8a">UInt16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a7a3cd3dbea53feb2e7043e6452260d8a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ad46043363fa7a3d8cc4b53bd69c5c57a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad46043363fa7a3d8cc4b53bd69c5c57a">UInt32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ad46043363fa7a3d8cc4b53bd69c5c57a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a11d2e4a9eba311bd85a70cc8793aaf75"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a11d2e4a9eba311bd85a70cc8793aaf75">UInt64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a11d2e4a9eba311bd85a70cc8793aaf75"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aa7811d90b33a2e5d399003f735d11ba3"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa7811d90b33a2e5d399003f735d11ba3">Int8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" hr [...]
+<tr class="separator:aa7811d90b33a2e5d399003f735d11ba3"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae8f12ddc9fce3fea6fe5dfe870a8dc6b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae8f12ddc9fce3fea6fe5dfe870a8dc6b">Int16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
+<tr class="separator:ae8f12ddc9fce3fea6fe5dfe870a8dc6b"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a88c9e3f761e2247a0518161db6a8358c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a88c9e3f761e2247a0518161db6a8358c">Int32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
+<tr class="separator:a88c9e3f761e2247a0518161db6a8358c"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a5be7437e66fa41a4c249f81a7efe7470"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5be7437e66fa41a4c249f81a7efe7470">Int64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" h [...]
+<tr class="separator:a5be7437e66fa41a4c249f81a7efe7470"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a2e3565ed4e1713896da4ee27b3317757"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2e3565ed4e1713896da4ee27b3317757">Float8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a2e3565ed4e1713896da4ee27b3317757"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aedf5c6024355c112045906632353bcfa"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aedf5c6024355c112045906632353bcfa">Float8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:aedf5c6024355c112045906632353bcfa"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a6735673e92f3697be33f5cf311c06f9d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6735673e92f3697be33f5cf311c06f9d">Float8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a6735673e92f3697be33f5cf311c06f9d"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a98a7d7a7f408fec2091af9d2195c655d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a98a7d7a7f408fec2091af9d2195c655d">Float8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a98a7d7a7f408fec2091af9d2195c655d"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a8cff6e0d907d3c4c19fea0a88e9bedf0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a8cff6e0d907d3c4c19fea0a88e9bedf0">Float8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a8cff6e0d907d3c4c19fea0a88e9bedf0"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a5cb22dc9a696dba7cb66574ef0e253ce"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5cb22dc9a696dba7cb66574ef0e253ce">Float16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a5cb22dc9a696dba7cb66574ef0e253ce"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a037c836af7ac994e2ffe76300ed6e1b7"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a037c836af7ac994e2ffe76300ed6e1b7">Float16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a037c836af7ac994e2ffe76300ed6e1b7"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a84ffb192123d18c7aeb22f0072847d43"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a84ffb192123d18c7aeb22f0072847d43">Float16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:a84ffb192123d18c7aeb22f0072847d43"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a6970dcedb53382eb2de268867c9a1b69"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a6970dcedb53382eb2de268867c9a1b69">Float16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:a6970dcedb53382eb2de268867c9a1b69"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a741dd7d66effeba15af9b539561b2ded"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a741dd7d66effeba15af9b539561b2ded">Float16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:a741dd7d66effeba15af9b539561b2ded"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab61768a121397418c43f7d4170c0c5da"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab61768a121397418c43f7d4170c0c5da">Float32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:ab61768a121397418c43f7d4170c0c5da"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab88e3bd2be02f2be5805c4cd5e0bf620"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab88e3bd2be02f2be5805c4cd5e0bf620">Float32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:ab88e3bd2be02f2be5805c4cd5e0bf620"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:abfc76052bf4e687fb91c903cb97dcd83"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abfc76052bf4e687fb91c903cb97dcd83">Float32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:abfc76052bf4e687fb91c903cb97dcd83"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a23c31f9b9908cc3c016592111f5d5860"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a23c31f9b9908cc3c016592111f5d5860">Float32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:a23c31f9b9908cc3c016592111f5d5860"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:af101d8c641d0f61ce35974d5fe3ef00a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#af101d8c641d0f61ce35974d5fe3ef00a">Float32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:af101d8c641d0f61ce35974d5fe3ef00a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae86a4246835fe2a1356a0ce18fc7a9e0"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae86a4246835fe2a1356a0ce18fc7a9e0">Float64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:ae86a4246835fe2a1356a0ce18fc7a9e0"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a4022cf159b063cb6671f0331ae33792a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4022cf159b063cb6671f0331ae33792a">Float64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a4022cf159b063cb6671f0331ae33792a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a71b08acd9bf3b1eb73f83b4edd5a5182"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a71b08acd9bf3b1eb73f83b4edd5a5182">Float64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:a71b08acd9bf3b1eb73f83b4edd5a5182"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab0f70432f1d4056ab3a434e3259b3e9e"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab0f70432f1d4056ab3a434e3259b3e9e">Float64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:ab0f70432f1d4056ab3a434e3259b3e9e"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aad6a7dc76ebbdcdb5eed91efc07fbff8"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aad6a7dc76ebbdcdb5eed91efc07fbff8">Float64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class=" [...]
+<tr class="separator:aad6a7dc76ebbdcdb5eed91efc07fbff8"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab84a0811f857d82faecf6c9cbbad5780"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab84a0811f857d82faecf6c9cbbad5780">UInt8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ab84a0811f857d82faecf6c9cbbad5780"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a227658e263028f9233da9f55c248a43f"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a227658e263028f9233da9f55c248a43f">UInt8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a227658e263028f9233da9f55c248a43f"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:affd5053dacbb25688ef6f067bbb78761"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#affd5053dacbb25688ef6f067bbb78761">UInt8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:affd5053dacbb25688ef6f067bbb78761"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:abbe8c3cb652b5fe7948080f8caf1570c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abbe8c3cb652b5fe7948080f8caf1570c">UInt8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:abbe8c3cb652b5fe7948080f8caf1570c"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a97cd0230ef0458bc166a387d49fa0250"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a97cd0230ef0458bc166a387d49fa0250">UInt8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a97cd0230ef0458bc166a387d49fa0250"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae3b27f43f0de67f3e5ba799888bbefa7"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae3b27f43f0de67f3e5ba799888bbefa7">UInt16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:ae3b27f43f0de67f3e5ba799888bbefa7"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:acb9032673c555649c524dcb423bd3ec9"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#acb9032673c555649c524dcb423bd3ec9">UInt16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:acb9032673c555649c524dcb423bd3ec9"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a4e2e296491ba53b33d0555f9edd7b9d2"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4e2e296491ba53b33d0555f9edd7b9d2">UInt16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a4e2e296491ba53b33d0555f9edd7b9d2"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a0dbc9e2fb673e412abfb1c77d49a053d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a0dbc9e2fb673e412abfb1c77d49a053d">UInt16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a0dbc9e2fb673e412abfb1c77d49a053d"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a2a0336d85cc5c8061fe29886dc871b2b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2a0336d85cc5c8061fe29886dc871b2b">UInt16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a2a0336d85cc5c8061fe29886dc871b2b"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a9492cfc58991919e32a9a9653a0544a6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9492cfc58991919e32a9a9653a0544a6">UInt32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a9492cfc58991919e32a9a9653a0544a6"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ac81a6cbb9e308a006df32b073e05098f"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac81a6cbb9e308a006df32b073e05098f">UInt32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:ac81a6cbb9e308a006df32b073e05098f"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ada1efb68ea879efc0ecc27196a3068a4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ada1efb68ea879efc0ecc27196a3068a4">UInt32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:ada1efb68ea879efc0ecc27196a3068a4"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a17c245f82bc5002bbe5eb9c225325a85"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a17c245f82bc5002bbe5eb9c225325a85">UInt32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a17c245f82bc5002bbe5eb9c225325a85"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a096f4ffaf512c1fefa0ceaefad55e593"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a096f4ffaf512c1fefa0ceaefad55e593">UInt32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a096f4ffaf512c1fefa0ceaefad55e593"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:afa4e70a7e0ad235f3d2a1160c0093625"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#afa4e70a7e0ad235f3d2a1160c0093625">UInt64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:afa4e70a7e0ad235f3d2a1160c0093625"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a920b270cb2b11f7ce4fee459f7f1a619"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a920b270cb2b11f7ce4fee459f7f1a619">UInt64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a920b270cb2b11f7ce4fee459f7f1a619"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a33feb22ecf4d6708312e93a950e43637"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a33feb22ecf4d6708312e93a950e43637">UInt64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a33feb22ecf4d6708312e93a950e43637"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a625cdc68ea89d04b9bf314c28f397234"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a625cdc68ea89d04b9bf314c28f397234">UInt64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:a625cdc68ea89d04b9bf314c28f397234"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:abdc950ea7ac026f00b8858e86b0fd81b"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#abdc950ea7ac026f00b8858e86b0fd81b">UInt64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="e [...]
+<tr class="separator:abdc950ea7ac026f00b8858e86b0fd81b"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aa5dd0f6b10d8a8512193987ba9c1564d"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa5dd0f6b10d8a8512193987ba9c1564d">Int8x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:aa5dd0f6b10d8a8512193987ba9c1564d"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a9a74627806410f80db60b17bee94cbc4"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a9a74627806410f80db60b17bee94cbc4">Int8x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a9a74627806410f80db60b17bee94cbc4"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aa8b37883752e594abbfc23bec33a5c44"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa8b37883752e594abbfc23bec33a5c44">Int8x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:aa8b37883752e594abbfc23bec33a5c44"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ab5ab0626b32c85ed15634859354bae60"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ab5ab0626b32c85ed15634859354bae60">Int8x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ab5ab0626b32c85ed15634859354bae60"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ac76b90af462de39cffd71f7a08340ec5"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ac76b90af462de39cffd71f7a08340ec5">Int8x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ac76b90af462de39cffd71f7a08340ec5"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae49920dbdafc3d66234dd5b9815efa38"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae49920dbdafc3d66234dd5b9815efa38">Int16x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:ae49920dbdafc3d66234dd5b9815efa38"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aa85096ef46c3453be8c0e446a5e6fcd3"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aa85096ef46c3453be8c0e446a5e6fcd3">Int16x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:aa85096ef46c3453be8c0e446a5e6fcd3"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a4ab76571223be0c914419c09d0e5fbf5"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a4ab76571223be0c914419c09d0e5fbf5">Int16x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a4ab76571223be0c914419c09d0e5fbf5"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aab48e91f64d8c71e8687517920295583"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aab48e91f64d8c71e8687517920295583">Int16x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:aab48e91f64d8c71e8687517920295583"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a2ea4609ed809b61fc078d3cbfbb588bf"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2ea4609ed809b61fc078d3cbfbb588bf">Int16x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a2ea4609ed809b61fc078d3cbfbb588bf"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a99e797d54d8ffcbf02d3a220cf86489a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a99e797d54d8ffcbf02d3a220cf86489a">Int32x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a99e797d54d8ffcbf02d3a220cf86489a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a5e81ae25cecfd56e2f3b474063324c3a"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a5e81ae25cecfd56e2f3b474063324c3a">Int32x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a5e81ae25cecfd56e2f3b474063324c3a"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a7e64669b1260b85c4e1722d1b98afc17"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a7e64669b1260b85c4e1722d1b98afc17">Int32x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a7e64669b1260b85c4e1722d1b98afc17"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ae800a458fed6eb05504f889a7f94b6b5"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ae800a458fed6eb05504f889a7f94b6b5">Int32x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:ae800a458fed6eb05504f889a7f94b6b5"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a2f7c6be2d1881aa9cbb6e5d5579a8fcc"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2f7c6be2d1881aa9cbb6e5d5579a8fcc">Int32x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a2f7c6be2d1881aa9cbb6e5d5579a8fcc"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:aae04dd74f177452906e9832e3e41da1c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#aae04dd74f177452906e9832e3e41da1c">Int64x4</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:aae04dd74f177452906e9832e3e41da1c"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a85202b4d4ad9b820b7ac1361905e233c"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a85202b4d4ad9b820b7ac1361905e233c">Int64x8</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a85202b4d4ad9b820b7ac1361905e233c"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a682890d013c50542ded6cfdad936f88f"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a682890d013c50542ded6cfdad936f88f">Int64x16</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a682890d013c50542ded6cfdad936f88f"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:ad828ae492de313cf743fbaa293189cba"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#ad828ae492de313cf743fbaa293189cba">Int64x32</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:ad828ae492de313cf743fbaa293189cba"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a439334cb39d3e5877553053b51973a10"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a439334cb39d3e5877553053b51973a10">Int64x64</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el [...]
+<tr class="separator:a439334cb39d3e5877553053b51973a10"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a25dce58350b5f2cb65d2367f16fec4b6"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a25dce58350b5f2cb65d2367f16fec4b6">Boolean</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" [...]
+<tr class="separator:a25dce58350b5f2cb65d2367f16fec4b6"><td class="memSeparator" colspan="2"> </td></tr>
+<tr class="memitem:a2b6c9317de9b507692d67a8e77dafc94"><td class="memItemLeft" align="right" valign="top"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> </td><td class="memItemRight" valign="bottom"><a class="el" href="namespacetvm_1_1script_1_1ir__builder_1_1tir.html#a2b6c9317de9b507692d67a8e77dafc94">Void</a> (<a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > expr=<a class="el" hr [...]
+<tr class="separator:a2b6c9317de9b507692d67a8e77dafc94"><td class="memSeparator" colspan="2"> </td></tr>
</table>
<h2 class="groupheader">Function Documentation</h2>
<a id="aa9787864f45dc62b7c3b6833003f96ea"></a>
@@ -876,8 +876,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a95b135dca5d84261342b70d3d8a34acc"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a95b135dca5d84261342b70d3d8a34acc">◆ </a></span>Boolean()</h2>
+<a id="a25dce58350b5f2cb65d2367f16fec4b6"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a25dce58350b5f2cb65d2367f16fec4b6">◆ </a></span>Boolean()</h2>
<div class="memitem">
<div class="memproto">
@@ -889,8 +889,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Boolean </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1221,8 +1231,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="aaf6b6dcbbd344633ac7549017151b1f4"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#aaf6b6dcbbd344633ac7549017151b1f4">◆ </a></span>Float16()</h2>
+<a id="ab878e27ae0a3913e1d58948d47cc3818"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab878e27ae0a3913e1d58948d47cc3818">◆ </a></span>Float16()</h2>
<div class="memitem">
<div class="memproto">
@@ -1234,8 +1244,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1247,8 +1267,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="af233b18ee0522a900568b14b6f9ef8d6"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#af233b18ee0522a900568b14b6f9ef8d6">◆ </a></span>Float16x16()</h2>
+<a id="a84ffb192123d18c7aeb22f0072847d43"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a84ffb192123d18c7aeb22f0072847d43">◆ </a></span>Float16x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -1260,8 +1280,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1273,8 +1303,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ad5311856f8b8b4df884170d8c4c513c8"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ad5311856f8b8b4df884170d8c4c513c8">◆ </a></span>Float16x32()</h2>
+<a id="a6970dcedb53382eb2de268867c9a1b69"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a6970dcedb53382eb2de268867c9a1b69">◆ </a></span>Float16x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -1286,8 +1316,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1299,8 +1339,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a6843af0b415dd3c3555b176cd3233b6b"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a6843af0b415dd3c3555b176cd3233b6b">◆ </a></span>Float16x4()</h2>
+<a id="a5cb22dc9a696dba7cb66574ef0e253ce"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a5cb22dc9a696dba7cb66574ef0e253ce">◆ </a></span>Float16x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -1312,8 +1352,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1325,8 +1375,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ac42c4020fca22350e8159b86cefb4031"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ac42c4020fca22350e8159b86cefb4031">◆ </a></span>Float16x64()</h2>
+<a id="a741dd7d66effeba15af9b539561b2ded"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a741dd7d66effeba15af9b539561b2ded">◆ </a></span>Float16x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -1338,8 +1388,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1351,8 +1411,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a1f72082ac88788080f123bf8c93500e6"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a1f72082ac88788080f123bf8c93500e6">◆ </a></span>Float16x8()</h2>
+<a id="a037c836af7ac994e2ffe76300ed6e1b7"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a037c836af7ac994e2ffe76300ed6e1b7">◆ </a></span>Float16x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -1364,8 +1424,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float16x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1377,8 +1447,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a8878f384c4a2861add3f43da4aa7e080"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a8878f384c4a2861add3f43da4aa7e080">◆ </a></span>Float32()</h2>
+<a id="ad0b06e1dabf3f744a991164800d50a0a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ad0b06e1dabf3f744a991164800d50a0a">◆ </a></span>Float32()</h2>
<div class="memitem">
<div class="memproto">
@@ -1390,8 +1460,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1403,8 +1483,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="acb206b62b60473442d995a4f868f57f1"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#acb206b62b60473442d995a4f868f57f1">◆ </a></span>Float32x16()</h2>
+<a id="abfc76052bf4e687fb91c903cb97dcd83"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#abfc76052bf4e687fb91c903cb97dcd83">◆ </a></span>Float32x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -1416,8 +1496,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1429,8 +1519,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ad859865f93db0e8b124464203253fdab"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ad859865f93db0e8b124464203253fdab">◆ </a></span>Float32x32()</h2>
+<a id="a23c31f9b9908cc3c016592111f5d5860"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a23c31f9b9908cc3c016592111f5d5860">◆ </a></span>Float32x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -1442,8 +1532,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1455,8 +1555,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="afc23e87468bd379cef2e436dec7380ec"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#afc23e87468bd379cef2e436dec7380ec">◆ </a></span>Float32x4()</h2>
+<a id="ab61768a121397418c43f7d4170c0c5da"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab61768a121397418c43f7d4170c0c5da">◆ </a></span>Float32x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -1468,8 +1568,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1481,8 +1591,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ad363b06dcc7a557663219774aae3ae59"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ad363b06dcc7a557663219774aae3ae59">◆ </a></span>Float32x64()</h2>
+<a id="af101d8c641d0f61ce35974d5fe3ef00a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#af101d8c641d0f61ce35974d5fe3ef00a">◆ </a></span>Float32x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -1494,8 +1604,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1507,8 +1627,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ae6706e8560ac719f2957e0186de014b3"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ae6706e8560ac719f2957e0186de014b3">◆ </a></span>Float32x8()</h2>
+<a id="ab88e3bd2be02f2be5805c4cd5e0bf620"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab88e3bd2be02f2be5805c4cd5e0bf620">◆ </a></span>Float32x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -1520,8 +1640,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float32x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1533,8 +1663,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a6ff858a2cfb49c10fc00ae68a1728086"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a6ff858a2cfb49c10fc00ae68a1728086">◆ </a></span>Float64()</h2>
+<a id="a9f5da5cec8dae24b9c7f94cbfd9ff9ca"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a9f5da5cec8dae24b9c7f94cbfd9ff9ca">◆ </a></span>Float64()</h2>
<div class="memitem">
<div class="memproto">
@@ -1546,8 +1676,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1559,8 +1699,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="acccc0b5d430d1205b3d1a782cab67aae"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#acccc0b5d430d1205b3d1a782cab67aae">◆ </a></span>Float64x16()</h2>
+<a id="a71b08acd9bf3b1eb73f83b4edd5a5182"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a71b08acd9bf3b1eb73f83b4edd5a5182">◆ </a></span>Float64x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -1572,8 +1712,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1585,8 +1735,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ae07732bbb9fea30ff6b9c03a7150beeb"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ae07732bbb9fea30ff6b9c03a7150beeb">◆ </a></span>Float64x32()</h2>
+<a id="ab0f70432f1d4056ab3a434e3259b3e9e"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ab0f70432f1d4056ab3a434e3259b3e9e">◆ </a></span>Float64x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -1598,8 +1748,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1611,8 +1771,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a4051e0d30c5fe712f65fffb8eee90d29"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a4051e0d30c5fe712f65fffb8eee90d29">◆ </a></span>Float64x4()</h2>
+<a id="ae86a4246835fe2a1356a0ce18fc7a9e0"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ae86a4246835fe2a1356a0ce18fc7a9e0">◆ </a></span>Float64x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -1624,8 +1784,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1637,8 +1807,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ab2aebc810c46e788e51c8e01a9eb7536"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ab2aebc810c46e788e51c8e01a9eb7536">◆ </a></span>Float64x64()</h2>
+<a id="aad6a7dc76ebbdcdb5eed91efc07fbff8"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aad6a7dc76ebbdcdb5eed91efc07fbff8">◆ </a></span>Float64x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -1650,8 +1820,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1663,8 +1843,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a4bb3d4dabf70259c2820a0f36c9e5661"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a4bb3d4dabf70259c2820a0f36c9e5661">◆ </a></span>Float64x8()</h2>
+<a id="a4022cf159b063cb6671f0331ae33792a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a4022cf159b063cb6671f0331ae33792a">◆ </a></span>Float64x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -1676,8 +1856,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float64x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1689,8 +1879,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a8f02a339b55424163e97c46f1c9f7976"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a8f02a339b55424163e97c46f1c9f7976">◆ </a></span>Float8()</h2>
+<a id="a0b31ddab3910ef43268f150db4e2266c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a0b31ddab3910ef43268f150db4e2266c">◆ </a></span>Float8()</h2>
<div class="memitem">
<div class="memproto">
@@ -1702,8 +1892,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1715,8 +1915,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ac5b707e9c9bd8ff20640e9aa5609f7ad"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ac5b707e9c9bd8ff20640e9aa5609f7ad">◆ </a></span>Float8x16()</h2>
+<a id="a6735673e92f3697be33f5cf311c06f9d"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a6735673e92f3697be33f5cf311c06f9d">◆ </a></span>Float8x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -1728,8 +1928,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1741,8 +1951,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a0c1842297c526585f5e7ecd5cf9fa835"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a0c1842297c526585f5e7ecd5cf9fa835">◆ </a></span>Float8x32()</h2>
+<a id="a98a7d7a7f408fec2091af9d2195c655d"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a98a7d7a7f408fec2091af9d2195c655d">◆ </a></span>Float8x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -1754,8 +1964,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1767,8 +1987,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a9d6a011a6d9a84e818bbed17821fca43"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a9d6a011a6d9a84e818bbed17821fca43">◆ </a></span>Float8x4()</h2>
+<a id="a2e3565ed4e1713896da4ee27b3317757"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a2e3565ed4e1713896da4ee27b3317757">◆ </a></span>Float8x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -1780,8 +2000,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1793,8 +2023,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a9493b11608cc2420b3e3780726556f0c"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a9493b11608cc2420b3e3780726556f0c">◆ </a></span>Float8x64()</h2>
+<a id="a8cff6e0d907d3c4c19fea0a88e9bedf0"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a8cff6e0d907d3c4c19fea0a88e9bedf0">◆ </a></span>Float8x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -1806,8 +2036,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1819,8 +2059,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a8d984cfc2c4f432a5e2bb16ed9723c9c"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a8d984cfc2c4f432a5e2bb16ed9723c9c">◆ </a></span>Float8x8()</h2>
+<a id="aedf5c6024355c112045906632353bcfa"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aedf5c6024355c112045906632353bcfa">◆ </a></span>Float8x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -1832,8 +2072,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Float8x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -1951,11 +2201,14 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a939b10df4c1a607526694b98ddbf8fd7"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a939b10df4c1a607526694b98ddbf8fd7">◆ </a></span>Handle()</h2>
+<a id="aaa2a6b8cd99890d0966ce5c83eb0ef77"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aaa2a6b8cd99890d0966ce5c83eb0ef77">◆ </a></span>Handle()</h2>
<div class="memitem">
<div class="memproto">
+<table class="mlabels">
+ <tr>
+ <td class="mlabels-left">
<table class="memname">
<tr>
<td class="memname"><a class="el" href="classtvm_1_1tir_1_1Var.html">Var</a> tvm::script::ir_builder::tir::Handle </td>
@@ -1967,7 +2220,13 @@ Functions</h2></td></tr>
<td class="paramkey"></td>
<td></td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1String.html">String</a> </td>
- <td class="paramname"><em>storage_scope</em> = <code>"global"</code> </td>
+ <td class="paramname"><em>storage_scope</em> = <code>"global"</code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
</tr>
<tr>
<td></td>
@@ -1975,6 +2234,11 @@ Functions</h2></td></tr>
<td></td><td></td>
</tr>
</table>
+ </td>
+ <td class="mlabels-right">
+<span class="mlabels"><span class="mlabel">inline</span></span> </td>
+ </tr>
+</table>
</div><div class="memdoc">
<p>Create a TIR var that represents a pointer. </p>
@@ -1982,6 +2246,7 @@ Functions</h2></td></tr>
<table class="params">
<tr><td class="paramname">dtype</td><td>The data type of the pointer. </td></tr>
<tr><td class="paramname">storage_scope</td><td>The storage scope of the pointer. </td></tr>
+ <tr><td class="paramname">is_size_var</td><td>Whether the pointer is a size var. </td></tr>
</table>
</dd>
</dl>
@@ -2036,8 +2301,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a71a8950ee36a0cc96b08f6a38a67beb8"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a71a8950ee36a0cc96b08f6a38a67beb8">◆ </a></span>Int16()</h2>
+<a id="ae8f12ddc9fce3fea6fe5dfe870a8dc6b"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ae8f12ddc9fce3fea6fe5dfe870a8dc6b">◆ </a></span>Int16()</h2>
<div class="memitem">
<div class="memproto">
@@ -2049,8 +2314,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2062,8 +2337,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a384760861c5679c8f3228fb069c4525b"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a384760861c5679c8f3228fb069c4525b">◆ </a></span>Int16x16()</h2>
+<a id="a4ab76571223be0c914419c09d0e5fbf5"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a4ab76571223be0c914419c09d0e5fbf5">◆ </a></span>Int16x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -2075,8 +2350,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2088,8 +2373,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a535df06d8e6eec52c711be076ca919da"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a535df06d8e6eec52c711be076ca919da">◆ </a></span>Int16x32()</h2>
+<a id="aab48e91f64d8c71e8687517920295583"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aab48e91f64d8c71e8687517920295583">◆ </a></span>Int16x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -2101,8 +2386,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2114,8 +2409,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="aee6b73b52838f0921c9dae3a6820b9a6"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#aee6b73b52838f0921c9dae3a6820b9a6">◆ </a></span>Int16x4()</h2>
+<a id="ae49920dbdafc3d66234dd5b9815efa38"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ae49920dbdafc3d66234dd5b9815efa38">◆ </a></span>Int16x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -2127,8 +2422,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2140,8 +2445,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="af817e48b37ce691b1538708ee4a96ae4"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#af817e48b37ce691b1538708ee4a96ae4">◆ </a></span>Int16x64()</h2>
+<a id="a2ea4609ed809b61fc078d3cbfbb588bf"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a2ea4609ed809b61fc078d3cbfbb588bf">◆ </a></span>Int16x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -2153,8 +2458,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2166,8 +2481,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a42335d7e171d3a46711e43ffd241e68d"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a42335d7e171d3a46711e43ffd241e68d">◆ </a></span>Int16x8()</h2>
+<a id="aa85096ef46c3453be8c0e446a5e6fcd3"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aa85096ef46c3453be8c0e446a5e6fcd3">◆ </a></span>Int16x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -2179,8 +2494,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int16x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2192,8 +2517,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="af90ae386a2b910cc52c17ebb6390404e"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#af90ae386a2b910cc52c17ebb6390404e">◆ </a></span>Int32()</h2>
+<a id="a88c9e3f761e2247a0518161db6a8358c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a88c9e3f761e2247a0518161db6a8358c">◆ </a></span>Int32()</h2>
<div class="memitem">
<div class="memproto">
@@ -2205,8 +2530,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2218,8 +2553,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a4945336e84e5ce9b60b583ac0adb28d6"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a4945336e84e5ce9b60b583ac0adb28d6">◆ </a></span>Int32x16()</h2>
+<a id="a7e64669b1260b85c4e1722d1b98afc17"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a7e64669b1260b85c4e1722d1b98afc17">◆ </a></span>Int32x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -2231,8 +2566,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2244,8 +2589,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a9f621b090a72a31cf4b80978d88ccf76"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a9f621b090a72a31cf4b80978d88ccf76">◆ </a></span>Int32x32()</h2>
+<a id="ae800a458fed6eb05504f889a7f94b6b5"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ae800a458fed6eb05504f889a7f94b6b5">◆ </a></span>Int32x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -2257,8 +2602,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2270,8 +2625,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a1d7bdb8458778293ced5af6ad9378283"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a1d7bdb8458778293ced5af6ad9378283">◆ </a></span>Int32x4()</h2>
+<a id="a99e797d54d8ffcbf02d3a220cf86489a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a99e797d54d8ffcbf02d3a220cf86489a">◆ </a></span>Int32x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -2283,8 +2638,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2296,8 +2661,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a57b983ed9d4d5000d0372e64e526f9e0"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a57b983ed9d4d5000d0372e64e526f9e0">◆ </a></span>Int32x64()</h2>
+<a id="a2f7c6be2d1881aa9cbb6e5d5579a8fcc"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a2f7c6be2d1881aa9cbb6e5d5579a8fcc">◆ </a></span>Int32x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -2309,8 +2674,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2322,8 +2697,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="ae89c85eabf166d4efddff53988ff57b5"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#ae89c85eabf166d4efddff53988ff57b5">◆ </a></span>Int32x8()</h2>
+<a id="a5e81ae25cecfd56e2f3b474063324c3a"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a5e81ae25cecfd56e2f3b474063324c3a">◆ </a></span>Int32x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -2335,8 +2710,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int32x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2348,8 +2733,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a7e3533536d109d1a5dd037c356ac2662"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a7e3533536d109d1a5dd037c356ac2662">◆ </a></span>Int64()</h2>
+<a id="a5be7437e66fa41a4c249f81a7efe7470"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a5be7437e66fa41a4c249f81a7efe7470">◆ </a></span>Int64()</h2>
<div class="memitem">
<div class="memproto">
@@ -2361,8 +2746,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2374,8 +2769,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a3c2d6fdb8690f5593c1d43b0cdf43d4e"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a3c2d6fdb8690f5593c1d43b0cdf43d4e">◆ </a></span>Int64x16()</h2>
+<a id="a682890d013c50542ded6cfdad936f88f"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a682890d013c50542ded6cfdad936f88f">◆ </a></span>Int64x16()</h2>
<div class="memitem">
<div class="memproto">
@@ -2387,8 +2782,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64x16 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2400,8 +2805,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a7e047daab47ebe417cdda5e3209607af"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a7e047daab47ebe417cdda5e3209607af">◆ </a></span>Int64x32()</h2>
+<a id="ad828ae492de313cf743fbaa293189cba"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#ad828ae492de313cf743fbaa293189cba">◆ </a></span>Int64x32()</h2>
<div class="memitem">
<div class="memproto">
@@ -2413,8 +2818,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64x32 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2426,8 +2841,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="aa84d61b0037299683391cfb1792064f9"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#aa84d61b0037299683391cfb1792064f9">◆ </a></span>Int64x4()</h2>
+<a id="aae04dd74f177452906e9832e3e41da1c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#aae04dd74f177452906e9832e3e41da1c">◆ </a></span>Int64x4()</h2>
<div class="memitem">
<div class="memproto">
@@ -2439,8 +2854,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64x4 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
+ <td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
<td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2452,8 +2877,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a61b20548797f6cbb80de229afe1cce53"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a61b20548797f6cbb80de229afe1cce53">◆ </a></span>Int64x64()</h2>
+<a id="a439334cb39d3e5877553053b51973a10"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a439334cb39d3e5877553053b51973a10">◆ </a></span>Int64x64()</h2>
<div class="memitem">
<div class="memproto">
@@ -2465,8 +2890,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64x64 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
</table>
</td>
@@ -2478,8 +2913,8 @@ Functions</h2></td></tr>
</div>
</div>
-<a id="a972550fc782bca80bfc88281e88d6022"></a>
-<h2 class="memtitle"><span class="permalink"><a href="#a972550fc782bca80bfc88281e88d6022">◆ </a></span>Int64x8()</h2>
+<a id="a85202b4d4ad9b820b7ac1361905e233c"></a>
+<h2 class="memtitle"><span class="permalink"><a href="#a85202b4d4ad9b820b7ac1361905e233c">◆ </a></span>Int64x8()</h2>
<div class="memitem">
<div class="memproto">
@@ -2491,8 +2926,18 @@ Functions</h2></td></tr>
<td class="memname"><a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> tvm::script::ir_builder::tir::Int64x8 </td>
<td>(</td>
<td class="paramtype"><a class="el" href="classtvm_1_1runtime_1_1Optional.html">Optional</a>< <a class="el" href="classtvm_1_1PrimExpr.html">PrimExpr</a> > </td>
- <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code></td><td>)</td>
+ <td class="paramname"><em>expr</em> = <code><a class="el" href="namespacetvm.html#aae7034e3e41c18e7fb78ff32bfc6a318">NullOpt</a></code>, </td>
+ </tr>
+ <tr>
+ <td class="paramkey"></td>
<td></td>
+ <td class="paramtype">bool </td>
+ <td class="paramname"><em>is_size_var</em> = <code>false</code> </td>
+ </tr>
+ <tr>
+ <td></td>
+ <td>)</td>
+ <td></td><td></td>
</tr>
... 5307 lines suppressed ...