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 2022/04/21 19:37:15 UTC
[tvm-site] branch asf-site updated: deploying docs (apache/tvm@876e2532278052d27c55adf982c9848ef3a76d4b)
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 2a3a07734 deploying docs (apache/tvm@876e2532278052d27c55adf982c9848ef3a76d4b)
2a3a07734 is described below
commit 2a3a07734ef858d9fb7057abc8b8216775dc85ea
Author: tvm-bot <95...@users.noreply.github.com>
AuthorDate: Thu Apr 21 19:37:09 2022 +0000
deploying docs (apache/tvm@876e2532278052d27c55adf982c9848ef3a76d4b)
---
.../how_to/compile_models/from_mxnet.rst.txt | 2 +-
.../how_to/compile_models/from_paddle.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 | 20 +-
.../deploy_models/deploy_model_on_android.rst.txt | 2 +-
.../deploy_object_detection_pytorch.rst.txt | 4 +-
.../deploy_models/deploy_prequantized.rst.txt | 6 +-
.../deploy_prequantized_tflite.rst.txt | 4 +-
.../how_to/deploy_models/deploy_quantized.rst.txt | 2 +-
.../deploy_models/deploy_ssd_gluoncv.rst.txt | 4 +-
.../deploy_models/sg_execution_times.rst.txt | 18 +-
.../extend_tvm/bring_your_own_datatypes.rst.txt | 2 +-
.../how_to/extend_tvm/sg_execution_times.rst.txt | 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 | 16 +-
.../tune_conv2d_layer_cuda.rst.txt | 1060 ++------------
.../tune_network_cuda.rst.txt | 2 +-
.../tune_network_x86.rst.txt | 4 +-
.../tune_sparse_x86.rst.txt | 138 +-
.../tune_with_autotvm/sg_execution_times.rst.txt | 12 +-
.../tune_with_autotvm/tune_conv2d_cuda.rst.txt | 34 +-
.../work_with_microtvm/micro_autotune.rst.txt | 16 +-
.../work_with_microtvm/sg_execution_times.rst.txt | 12 +-
.../work_with_relay/sg_execution_times.rst.txt | 8 +-
.../work_with_schedules/sg_execution_times.rst.txt | 18 +-
.../how_to/work_with_schedules/tensorize.rst.txt | 2 +-
.../tutorials/autotvm/sg_execution_times.rst.txt | 6 +-
.../frontend/deploy_classification.rst.txt | 2 +-
.../tutorials/frontend/deploy_detection.rst.txt | 2 +-
.../tutorials/frontend/sg_execution_times.rst.txt | 6 +-
.../tutorials/optimize/sg_execution_times.rst.txt | 6 +-
.../topic/vta/tutorials/sg_execution_times.rst.txt | 6 +-
.../tutorial/auto_scheduler_matmul_x86.rst.txt | 2 +-
docs/_sources/tutorial/autotvm_relay_x86.rst.txt | 66 +-
.../tutorial/cross_compilation_and_rpc.rst.txt | 2 +-
docs/_sources/tutorial/intro_topi.rst.txt | 2 +-
docs/_sources/tutorial/sg_execution_times.rst.txt | 26 +-
.../tutorial/tensor_expr_get_started.rst.txt | 42 +-
docs/commit_hash | 2 +-
docs/genindex.html | 2 +
docs/how_to/compile_models/from_mxnet.html | 2 +-
docs/how_to/compile_models/from_paddle.html | 2 +-
docs/how_to/compile_models/from_pytorch.html | 6 +-
docs/how_to/compile_models/from_tensorflow.html | 2 +-
docs/how_to/compile_models/sg_execution_times.html | 20 +-
.../deploy_models/deploy_model_on_android.html | 2 +-
.../deploy_object_detection_pytorch.html | 19 +-
docs/how_to/deploy_models/deploy_prequantized.html | 7 +-
.../deploy_models/deploy_prequantized_tflite.html | 4 +-
docs/how_to/deploy_models/deploy_quantized.html | 2 +-
docs/how_to/deploy_models/deploy_ssd_gluoncv.html | 35 +-
docs/how_to/deploy_models/sg_execution_times.html | 18 +-
.../extend_tvm/bring_your_own_datatypes.html | 2 +-
docs/how_to/extend_tvm/sg_execution_times.html | 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 | 1060 ++------------
.../tune_with_autoscheduler/tune_network_cuda.html | 2 +-
.../tune_with_autoscheduler/tune_network_x86.html | 4 +-
.../tune_with_autoscheduler/tune_sparse_x86.html | 138 +-
.../tune_with_autotvm/sg_execution_times.html | 12 +-
.../how_to/tune_with_autotvm/tune_conv2d_cuda.html | 34 +-
docs/how_to/work_with_microtvm/micro_autotune.html | 16 +-
.../work_with_microtvm/sg_execution_times.html | 12 +-
.../how_to/work_with_relay/sg_execution_times.html | 8 +-
.../work_with_schedules/sg_execution_times.html | 18 +-
docs/how_to/work_with_schedules/tensorize.html | 2 +-
docs/objects.inv | Bin 22071 -> 22083 bytes
docs/reference/api/doxygen/array_8h__dep__incl.svg | 232 +--
.../api/doxygen/c__runtime__api_8h__dep__incl.svg | 40 +-
.../api/doxygen/data__type_8h__dep__incl.svg | 84 +-
docs/reference/api/doxygen/functor_8h.html | 2 +-
.../api/doxygen/functor_8h__dep__incl.svg | 1086 +++++++-------
docs/reference/api/doxygen/greedy_8h.html | 2 +-
docs/reference/api/doxygen/greedy_8h__incl.svg | 998 ++++++-------
.../api/doxygen/ir_2attrs_8h__dep__incl.svg | 88 +-
.../api/doxygen/ir_2expr_8h__dep__incl.svg | 148 +-
docs/reference/api/doxygen/ir_2function_8h.html | 2 +-
.../api/doxygen/ir_2function_8h__dep__incl.svg | 1050 ++++++-------
.../api/doxygen/ir_2span_8h__dep__incl.svg | 164 +--
.../api/doxygen/ir_2type_8h__dep__incl.svg | 96 +-
docs/reference/api/doxygen/map_8h__dep__incl.svg | 212 +--
.../api/doxygen/namespacemembers_func_r.html | 7 +-
docs/reference/api/doxygen/namespacemembers_r.html | 7 +-
docs/reference/api/doxygen/namespacemembers_s.html | 4 +-
.../reference/api/doxygen/namespacetvm_1_1tir.html | 30 +
docs/reference/api/doxygen/ndarray_8h.html | 2 +-
.../api/doxygen/ndarray_8h__dep__incl.svg | 897 ++++++------
docs/reference/api/doxygen/node_8h__dep__incl.svg | 132 +-
.../reference/api/doxygen/object_8h__dep__incl.svg | 100 +-
.../api/doxygen/optional_8h__dep__incl.svg | 188 +--
.../api/doxygen/packed__func_8h__dep__incl.svg | 140 +-
.../api/doxygen/reflection_8h__dep__incl.svg | 72 +-
.../api/doxygen/repr__printer_8h__dep__incl.svg | 120 +-
.../runtime_2container_2base_8h__dep__incl.svg | 116 +-
.../api/doxygen/runtime_2memory_8h__dep__incl.svg | 88 +-
.../api/doxygen/runtime_2module_8h__dep__incl.svg | 140 +-
docs/reference/api/doxygen/search/all_13.js | 1 +
docs/reference/api/doxygen/search/all_14.js | 2 +-
docs/reference/api/doxygen/search/functions_12.js | 1 +
docs/reference/api/doxygen/serializer_8h.html | 2 +-
.../api/doxygen/serializer_8h__dep__incl.svg | 891 ++++++-----
docs/reference/api/doxygen/shape__tuple_8h.html | 2 +-
.../api/doxygen/shape__tuple_8h__dep__incl.svg | 926 ++++++------
docs/reference/api/doxygen/stmt_8h.html | 2 +-
docs/reference/api/doxygen/stmt_8h__dep__incl.svg | 814 ++++++-----
docs/reference/api/doxygen/stmt__functor_8h.html | 6 +-
.../api/doxygen/stmt__functor_8h__dep__incl.svg | 12 +-
.../api/doxygen/stmt__functor_8h__incl.svg | 1544 +++++++++++---------
.../api/doxygen/stmt__functor_8h_source.html | 79 +-
.../reference/api/doxygen/string_8h__dep__incl.svg | 92 +-
.../doxygen/structural__equal_8h__dep__incl.svg | 300 ++--
.../api/doxygen/structural__hash_8h__dep__incl.svg | 300 ++--
docs/reference/api/doxygen/tir_2expr_8h.html | 2 +-
.../api/doxygen/tir_2expr_8h__dep__incl.svg | 716 ++++-----
docs/reference/api/doxygen/tir_2function_8h.html | 2 +-
.../api/doxygen/tir_2function_8h__dep__incl.svg | 577 ++++----
docs/reference/api/doxygen/var_8h.html | 2 +-
docs/reference/api/doxygen/var_8h__dep__incl.svg | 736 +++++-----
docs/reference/api/python/auto_scheduler.html | 4 +-
docs/reference/api/python/tir.html | 22 +
.../api/typedoc/classes/bytestreamreader.html | 12 +-
.../api/typedoc/classes/cachedcallstack.html | 34 +-
docs/reference/api/typedoc/classes/dldatatype.html | 12 +-
docs/reference/api/typedoc/classes/dldevice.html | 10 +-
.../reference/api/typedoc/classes/environment.html | 12 +-
docs/reference/api/typedoc/classes/ffilibrary.html | 20 +-
.../api/typedoc/classes/graphexecutor.html | 16 +-
docs/reference/api/typedoc/classes/instance.html | 40 +-
docs/reference/api/typedoc/classes/memory.html | 34 +-
docs/reference/api/typedoc/classes/module.html | 10 +-
docs/reference/api/typedoc/classes/ndarray.html | 22 +-
.../api/typedoc/classes/packedfunccell.html | 6 +-
docs/reference/api/typedoc/classes/rpcserver.html | 14 +-
docs/reference/api/typedoc/classes/scalar.html | 6 +-
.../api/typedoc/classes/webgpucontext.html | 12 +-
docs/reference/api/typedoc/enums/argtypecode.html | 30 +-
.../api/typedoc/enums/aynccallbackcode.html | 4 +-
.../api/typedoc/enums/dldatatypecode.html | 8 +-
.../api/typedoc/enums/rpcserverstate.html | 12 +-
docs/reference/api/typedoc/enums/sizeof.html | 18 +-
docs/reference/api/typedoc/index.html | 112 +-
.../api/typedoc/interfaces/disposable.html | 2 +-
.../api/typedoc/interfaces/functioninfo.html | 6 +-
.../api/typedoc/interfaces/libraryprovider.html | 4 +-
docs/searchindex.js | 2 +-
.../vta/tutorials/autotvm/sg_execution_times.html | 6 +-
.../tutorials/frontend/deploy_classification.html | 2 +-
.../vta/tutorials/frontend/deploy_detection.html | 2 +-
.../vta/tutorials/frontend/sg_execution_times.html | 6 +-
.../vta/tutorials/optimize/sg_execution_times.html | 6 +-
docs/topic/vta/tutorials/sg_execution_times.html | 6 +-
docs/tutorial/auto_scheduler_matmul_x86.html | 2 +-
docs/tutorial/autotvm_relay_x86.html | 174 +--
docs/tutorial/cross_compilation_and_rpc.html | 2 +-
docs/tutorial/intro_topi.html | 2 +-
docs/tutorial/sg_execution_times.html | 26 +-
docs/tutorial/tensor_expr_get_started.html | 42 +-
167 files changed, 7907 insertions(+), 9222 deletions(-)
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 363ff3f52..89dc88528 100644
--- a/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_mxnet.rst.txt
@@ -98,7 +98,7 @@ In this section, we download a pretrained imagenet model and classify an image.
.. code-block:: none
- Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zipf4aa77d4-0448-4ae1-b003-b1bad7ca0c21 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+ Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip1295d9df-221c-49dc-ad75-1f57fdf19cd0 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_paddle.rst.txt b/docs/_sources/how_to/compile_models/from_paddle.rst.txt
index 1873ea26f..d5cbb2cf8 100644
--- a/docs/_sources/how_to/compile_models/from_paddle.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_paddle.rst.txt
@@ -201,7 +201,7 @@ Look up prediction top 1 index in 1000 class synset.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 10.141 seconds)
+ **Total running time of the script:** ( 1 minutes 5.850 seconds)
.. _sphx_glr_download_how_to_compile_models_from_paddle.py:
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 b1e73f1d3..1b296ead1 100644
--- a/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_pytorch.rst.txt
@@ -79,7 +79,7 @@ Load a pretrained PyTorch model
.. code-block:: none
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]
43%|####2 | 19.0M/44.7M [00:00<00:00, 200MB/s]
96%|#########6| 43.0M/44.7M [00:00<00:00, 230MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 227MB/s]
+
0%| | 0.00/44.7M [00:00<?, ?B/s]
39%|###9 | 17.4M/44.7M [00:00<00:00, 183MB/s]
93%|#########3| 41.6M/44.7M [00:00<00:00, 224MB/s]
100%|##########| 44.7M/44.7M [00:00<00:00, 220MB/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 68dfea487..bdb97161a 100644
--- a/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
+++ b/docs/_sources/how_to/compile_models/from_tensorflow.rst.txt
@@ -372,7 +372,7 @@ Run the corresponding model on tensorflow
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 4.535 seconds)
+ **Total running time of the script:** ( 1 minutes 0.778 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 25c8d1d8e..77ae95593 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
@@ -5,14 +5,14 @@
Computation times
=================
-**04:52.614** total execution time for **how_to_compile_models** files:
+**04:50.120** total execution time for **how_to_compile_models** files:
-- **01:10.141**: :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)
-- **01:04.535**: :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``)
-- **00:55.666**: :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)
-- **00:25.857**: :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)
-- **00:21.078**: :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)
-- **00:20.828**: :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)
-- **00:18.698**: :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)
-- **00:13.337**: :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)
-- **00:02.474**: :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)
+- **01:05.850**: :ref:`sphx_glr_how_to_compile_models_from_paddle.py` (``from_paddle.py``)
+- **01:00.778**: :ref:`sphx_glr_how_to_compile_models_from_tensorflow.py` (``from_tensorflow.py``)
+- **00:58.534**: :ref:`sphx_glr_how_to_compile_models_from_darknet.py` (``from_darknet.py``)
+- **00:25.625**: :ref:`sphx_glr_how_to_compile_models_from_tflite.py` (``from_tflite.py``)
+- **00:22.643**: :ref:`sphx_glr_how_to_compile_models_from_coreml.py` (``from_coreml.py``)
+- **00:21.349**: :ref:`sphx_glr_how_to_compile_models_from_mxnet.py` (``from_mxnet.py``)
+- **00:18.848**: :ref:`sphx_glr_how_to_compile_models_from_pytorch.py` (``from_pytorch.py``)
+- **00:14.094**: :ref:`sphx_glr_how_to_compile_models_from_keras.py` (``from_keras.py``)
+- **00:02.399**: :ref:`sphx_glr_how_to_compile_models_from_onnx.py` (``from_onnx.py``)
diff --git a/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt b/docs/_sources/how_to/deploy_models/deploy_model_on_android.rst.txt
index 4344cbb52..b85c83c52 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
@@ -393,7 +393,7 @@ Execute on TVM
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 15.9295 15.9324 16.0288 15.8394 0.0625
+ 16.2107 16.1555 16.6169 16.0543 0.1617
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 69d23b898..bf89a93d3 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
@@ -108,7 +108,7 @@ Load pre-trained maskrcnn from torchvision and do tracing
.. code-block:: none
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]
9%|8 | 14.8M/170M [00:00<00:01, 155MB/s]
21%|## | 34.9M/170M [00:00<00:00, 188MB/s]
32%|###2 | 55.0M/170M [00:00<00:00, 198MB/s]
45%|####5 | 76.7M/170M [00:00<00:00, 210MB/s]
58%|#####7 | 98.4M/170M [00:00<00:00, 216MB/s]
70%|####### | 120M/170M [00:00<00:00, 218MB/s]
83%|########2 | 140M/170M [00:00<00:00, 215MB/s]
95%|#########5| 162M/170M [00:00<00:00, 220MB/s]
100%|##########| 170M/170M [00:00<00:00, 213MB/s]
+
0%| | 0.00/170M [00:00<?, ?B/s]
9%|8 | 15.1M/170M [00:00<00:01, 158MB/s]
22%|##1 | 37.1M/170M [00:00<00:00, 201MB/s]
36%|###6 | 61.7M/170M [00:00<00:00, 227MB/s]
50%|##### | 85.1M/170M [00:00<00:00, 234MB/s]
64%|######3 | 109M/170M [00:00<00:00, 239MB/s]
78%|#######7 | 132M/170M [00:00<00:00, 242MB/s]
92%|#########2| 156M/170M [00:00<00:00, 245MB/s]
100%|##########| 170M/170M [00:00<00:00, 236MB/s]
/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:3878: 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)
/usr/local/lib/python3.7/dist-packages/torchvision/models/detection/anchor_utils.py:127: 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').
@@ -253,7 +253,7 @@ Get boxes with score larger than 0.9
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 3 minutes 2.115 seconds)
+ **Total running time of the script:** ( 3 minutes 9.451 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 1c7f628f6..b2a0b904c 100644
--- a/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_prequantized.rst.txt
@@ -187,7 +187,7 @@ training. Other models require a full post training calibration.
.. code-block:: none
Downloading: "https://download.pytorch.org/models/mobilenet_v2-b0353104.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2-b0353104.pth
-
0%| | 0.00/13.6M [00:00<?, ?B/s]
100%|#########9| 13.5M/13.6M [00:00<00:00, 142MB/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 141MB/s]
+
0%| | 0.00/13.6M [00:00<?, ?B/s]
100%|##########| 13.6M/13.6M [00:00<00:00, 170MB/s]
@@ -344,7 +344,7 @@ 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)
- 90.2516 90.1702 92.3845 89.9664 0.3181
+ 90.2704 90.2249 91.4279 90.0486 0.2188
@@ -384,7 +384,7 @@ TODO
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 4.674 seconds)
+ **Total running time of the script:** ( 1 minutes 6.788 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 b424daa81..3b7bed528 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
@@ -351,7 +351,7 @@ Here we give an example of how to measure performance of TVM compiled models.
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 119.8836 119.8378 121.7985 118.9666 0.4667
+ 120.2926 120.3084 123.3358 119.3317 0.5461
@@ -385,7 +385,7 @@ 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:** ( 1 minutes 59.418 seconds)
+ **Total running time of the script:** ( 1 minutes 52.122 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 bc1b88b54..691498d15 100644
--- a/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
+++ b/docs/_sources/how_to/deploy_models/deploy_quantized.rst.txt
@@ -221,7 +221,7 @@ We create a Relay VM to build and execute the model.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 13.044 seconds)
+ **Total running time of the script:** ( 1 minutes 13.302 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 930f3e497..03af3863d 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
@@ -137,7 +137,7 @@ Convert and compile model for CPU.
data: None
input_sym_arg_type = in_param.infer_type()[0]
Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/ssd_512_resnet50_v1_voc-9c8b225a.zip...
-
0%| | 0/132723 [00:00<?, ?KB/s]
5%|5 | 6737/132723 [00:00<00:01, 67344.21KB/s]
12%|#1 | 15461/132723 [00:00<00:01, 79039.28KB/s]
18%|#8 | 24208/132723 [00:00<00:01, 82883.95KB/s]
25%|##4 | 32905/132723 [00:00<00:01, 84491.20KB/s]
31%|###1 | 41689/132723 [00:00<00:01, 85694.27KB/s]
38%|###7 | 50259/132723 [00:00<00:01, 79912.01KB/s]
44%|####3 | 58390/132723 [00:00<00:00, 80343.15KB/s]
50%|##### | 66472/132723 [00:00<00:01, 63064.99KB/s]
57%|#####6 | 75197/132723 [00:01<00:00, 69263.39KB/s]
62%|######2 | 82644/132723 [00:01<00:00, 56738.18KB/s]
69%|######8 | 91366/132723 [00:01<00:00, 63906.78KB/s]
75%|#######4 | 98990/132723 [00:01<00:00, 67006.84KB/s]
80%|######## | 106749/132723 [00:01<00:00, 69795.59KB/s]
86%|########6 | 114680/132723 [00:01<00:00, 64651.96KB/s]
93%|#########2| 122941/132723 [00:01<00:00, 69187.98KB/s]
99%|########
#9| 131712/132723 [00:01<00:00, 74183.60KB/s]
100%|##########| 132723/132723 [00:01<00:00, 71327.11KB/s]
+
0%| | 0/132723 [00:00<?, ?KB/s]
5%|4 | 6575/132723 [00:00<00:01, 65745.99KB/s]
11%|#1 | 15153/132723 [00:00<00:01, 77527.40KB/s]
18%|#7 | 23675/132723 [00:00<00:01, 81036.04KB/s]
24%|##4 | 32190/132723 [00:00<00:01, 82652.12KB/s]
31%|### | 40693/132723 [00:00<00:01, 83504.31KB/s]
37%|###6 | 49044/132723 [00:00<00:01, 77717.03KB/s]
43%|####2 | 56885/132723 [00:00<00:01, 69497.90KB/s]
49%|####9 | 65433/132723 [00:00<00:00, 74029.85KB/s]
56%|#####5 | 73983/132723 [00:00<00:00, 77337.45KB/s]
62%|######2 | 82669/132723 [00:01<00:00, 80115.62KB/s]
69%|######8 | 91303/132723 [00:01<00:00, 81945.29KB/s]
75%|#######5 | 99933/132723 [00:01<00:00, 83232.82KB/s]
82%|########1 | 108509/132723 [00:01<00:00, 83980.08KB/s]
88%|########8 | 117150/132723 [00:01<00:00, 84701.27KB/s]
95%|#########4| 125785/132723 [00:01<00:00, 85192.01KB/s]
100%|########
##| 132723/132723 [00:01<00:00, 80813.26KB/s]
@@ -202,7 +202,7 @@ Display result
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 2 minutes 23.362 seconds)
+ **Total running time of the script:** ( 2 minutes 27.384 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 63edd9e7a..55d4d290c 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
@@ -5,13 +5,13 @@
Computation times
=================
-**10:32.073** total execution time for **how_to_deploy_models** files:
+**10:40.220** total execution time for **how_to_deploy_models** files:
-- **03:02.115**: :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``)
-- **02:23.362**: :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)
-- **01:59.418**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)
-- **01:13.044**: :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)
-- **01:04.674**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)
-- **00:27.290**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)
-- **00:21.977**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)
-- **00:00.194**: :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)
+- **03:09.451**: :ref:`sphx_glr_how_to_deploy_models_deploy_object_detection_pytorch.py` (``deploy_object_detection_pytorch.py``)
+- **02:27.384**: :ref:`sphx_glr_how_to_deploy_models_deploy_ssd_gluoncv.py` (``deploy_ssd_gluoncv.py``)
+- **01:52.122**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized_tflite.py` (``deploy_prequantized_tflite.py``)
+- **01:13.302**: :ref:`sphx_glr_how_to_deploy_models_deploy_quantized.py` (``deploy_quantized.py``)
+- **01:06.788**: :ref:`sphx_glr_how_to_deploy_models_deploy_prequantized.py` (``deploy_prequantized.py``)
+- **00:29.123**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_android.py` (``deploy_model_on_android.py``)
+- **00:21.846**: :ref:`sphx_glr_how_to_deploy_models_deploy_model_on_rasp.py` (``deploy_model_on_rasp.py``)
+- **00:00.203**: :ref:`sphx_glr_how_to_deploy_models_deploy_sparse.py` (``deploy_sparse.py``)
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 fb186284f..2f4261709 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
@@ -423,7 +423,7 @@ 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.zip37fe282f-da91-43a1-a990-86309c5d85cb from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/mobilenet0.25-9f83e440.zip...
+ Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip3a5965b5-17fb-4b77-9335-e415a17b7a43 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 dc1aa06de..ebf44b5ae 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
@@ -5,9 +5,9 @@
Computation times
=================
-**00:38.319** total execution time for **how_to_extend_tvm** files:
+**00:39.355** total execution time for **how_to_extend_tvm** files:
-- **00:34.801**: :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``)
-- **00:02.256**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)
-- **00:01.061**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)
-- **00:00.201**: :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)
+- **00:35.731**: :ref:`sphx_glr_how_to_extend_tvm_bring_your_own_datatypes.py` (``bring_your_own_datatypes.py``)
+- **00:02.326**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_instrument.py` (``use_pass_instrument.py``)
+- **00:01.093**: :ref:`sphx_glr_how_to_extend_tvm_use_pass_infra.py` (``use_pass_infra.py``)
+- **00:00.205**: :ref:`sphx_glr_how_to_extend_tvm_low_level_custom_pass.py` (``low_level_custom_pass.py``)
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 1884280f3..354c48bc3 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
@@ -199,10 +199,10 @@ profile the execution time of each passes.
.. code-block:: none
Printing results of timing profile...
- InferType: 6001us [6001us] (44.02%; 44.02%)
- FoldScaleAxis: 7630us [2us] (55.98%; 55.98%)
- FoldConstant: 7628us [1470us] (55.96%; 99.97%)
- InferType: 6157us [6157us] (45.17%; 80.72%)
+ InferType: 6220us [6220us] (46.06%; 46.06%)
+ FoldScaleAxis: 7284us [2us] (53.94%; 53.94%)
+ FoldConstant: 7282us [1502us] (53.92%; 99.97%)
+ InferType: 5780us [5780us] (42.80%; 79.38%)
@@ -239,10 +239,10 @@ Refer to following sections and :py:func:`tvm.instrument.pass_instrument` for th
.. code-block:: none
Printing results of timing profile...
- InferType: 5829us [5829us] (44.86%; 44.86%)
- FoldScaleAxis: 7165us [2us] (55.14%; 55.14%)
- FoldConstant: 7163us [1503us] (55.13%; 99.98%)
- InferType: 5660us [5660us] (43.56%; 79.02%)
+ InferType: 5894us [5894us] (44.19%; 44.19%)
+ FoldScaleAxis: 7445us [2us] (55.81%; 55.81%)
+ FoldConstant: 7443us [1533us] (55.80%; 99.97%)
+ InferType: 5910us [5910us] (44.30%; 79.40%)
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 ddcbe0492..3e740fcdc 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
@@ -295,7 +295,7 @@ latency of convolution.
.. code-block:: none
- Convolution: 54.107188 ms
+ Convolution: 54.191826 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 ca3a8e298..ea83ed415 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
@@ -628,7 +628,7 @@ be able to run on our build server
.. code-block:: none
- conv2d with tensor core: 6.470923 ms
+ conv2d with tensor core: 6.846378 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 6d08d5f31..6713cb7ea 100644
--- a/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
+++ b/docs/_sources/how_to/optimize_operators/opt_gemm.rst.txt
@@ -118,8 +118,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
.. code-block:: none
- Numpy running time: 0.019322
- Baseline: 3.339645
+ Numpy running time: 0.019626
+ Baseline: 3.279458
@@ -210,7 +210,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
.. code-block:: none
- Opt1: 0.305210
+ Opt1: 0.315032
@@ -309,7 +309,7 @@ In this tutorial, we chose to vectorize the inner loop row data since it is cach
.. code-block:: none
- Opt2: 0.339233
+ Opt2: 0.342817
@@ -401,7 +401,7 @@ the access pattern for A matrix is more cache friendly.
.. code-block:: none
- Opt3: 0.115594
+ Opt3: 0.118973
@@ -520,7 +520,7 @@ flattening.
.. code-block:: none
- Opt4: 0.112289
+ Opt4: 0.110845
@@ -638,7 +638,7 @@ write to C when all the block results are ready.
.. code-block:: none
- Opt5: 0.111148
+ Opt5: 0.111547
@@ -759,7 +759,7 @@ Futhermore, we can also utilize multi-core processors to do the thread-level par
.. code-block:: none
- Opt6: 0.144759
+ Opt6: 0.145354
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 d5be1a5cf..2d5b292d9 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
@@ -5,8 +5,8 @@
Computation times
=================
-**00:34.778** total execution time for **how_to_optimize_operators** files:
+**00:35.098** total execution time for **how_to_optimize_operators** files:
-- **00:32.321**: :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)
-- **00:01.308**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``)
-- **00:01.150**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)
+- **00:32.437**: :ref:`sphx_glr_how_to_optimize_operators_opt_gemm.py` (``opt_gemm.py``)
+- **00:01.406**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_tensorcore.py` (``opt_conv_tensorcore.py``)
+- **00:01.255**: :ref:`sphx_glr_how_to_optimize_operators_opt_conv_cuda.py` (``opt_conv_cuda.py``)
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 16dac0f8b..be1e9d29e 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
@@ -5,11 +5,11 @@
Computation times
=================
-**05:02.810** total execution time for **how_to_tune_with_autoscheduler** files:
-
-- **02:17.757**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``)
-- **01:20.507**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)
-- **00:40.706**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)
-- **00:26.654**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)
-- **00:08.615**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)
-- **00:08.571**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)
+**04:57.684** total execution time for **how_to_tune_with_autoscheduler** files:
+
+- **02:20.501**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_conv2d_layer_cuda.py` (``tune_conv2d_layer_cuda.py``)
+- **01:21.450**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_x86.py` (``tune_network_x86.py``)
+- **00:41.139**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_cuda.py` (``tune_network_cuda.py``)
+- **00:16.500**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_sparse_x86.py` (``tune_sparse_x86.py``)
+- **00:09.239**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_mali.py` (``tune_network_mali.py``)
+- **00:08.855**: :ref:`sphx_glr_how_to_tune_with_autoscheduler_tune_network_arm.py` (``tune_network_arm.py``)
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 fe873d701..d626f17ef 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
@@ -222,484 +222,101 @@ cooperative fetching, unrolling and operator fusion.
compute: Buffer(compute_2: Pointer(float32), float32, [25088], [])}
buffer_map = {data_1: data, kernel_1: kernel, bias_1: bias, compute_1: compute}
preflattened_buffer_map = {data_1: data_3: Buffer(data_2, float32, [1, 512, 7, 7], []), kernel_1: kernel_3: Buffer(kernel_2, float32, [512, 512, 3, 3], []), bias_1: bias_3: Buffer(bias_2, float32, [1, 512, 1, 1], []), compute_1: compute_3: Buffer(compute_2, float32, [1, 512, 7, 7], [])} {
- attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 28;
- allocate(conv2d_nchw: Pointer(local float32), float32, [14]), storage_scope = local;
- allocate(pad_temp.shared: Pointer(shared float32), float32, [72]), storage_scope = shared;
- allocate(kernel.shared: Pointer(shared float32), float32, [3072]), storage_scope = shared;
- attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64 {
- conv2d_nchw_1: Buffer(conv2d_nchw, float32, [14], [], scope="local", align=32)[0] = 0f32
+ attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 112;
+ allocate(conv2d_nchw: Pointer(local float32), float32, [7]), storage_scope = local;
+ allocate(pad_temp.shared: Pointer(shared float32), float32, [432]), storage_scope = shared;
+ allocate(kernel.shared: Pointer(shared float32), float32, [4608]), storage_scope = shared;
+ attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32 {
+ conv2d_nchw_1: Buffer(conv2d_nchw, float32, [1], [], scope="local", align=4)[0] = 0f32
conv2d_nchw_1[1] = 0f32
conv2d_nchw_1[2] = 0f32
conv2d_nchw_1[3] = 0f32
conv2d_nchw_1[4] = 0f32
conv2d_nchw_1[5] = 0f32
conv2d_nchw_1[6] = 0f32
- conv2d_nchw_1[7] = 0f32
- conv2d_nchw_1[8] = 0f32
- conv2d_nchw_1[9] = 0f32
- conv2d_nchw_1[10] = 0f32
- conv2d_nchw_1[11] = 0f32
- conv2d_nchw_1[12] = 0f32
- conv2d_nchw_1[13] = 0f32
- for (rc.outer.outer: int32, 0, 64) {
- for (ry.outer.outer: int32, 0, 3) {
- let cse_var_2: int32 = (rc.outer.outer*72)
- let cse_var_1: int32 = (ry.outer.outer*3)
- {
- attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64 {
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1: Buffer(pad_temp.shared, float32, [72], [], scope="shared")[(threadIdx.x_1*4)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1*4), 9))) && (floormod((threadIdx.x_1*4), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv((threadIdx.x_1*4), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod((threadIdx.x_1*4), 9)) - 8)], 0f3 [...]
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 1)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 1), 9))) && (floormod(((threadIdx.x_1*4) + 1), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 1), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 1), 9)) - 8)], 0f32, dtype=float32)
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 2)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 2), 9))) && (floormod(((threadIdx.x_1*4) + 2), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 2), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 2), 9)) - 8)], 0f32, dtype=float32)
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 3)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 3), 9))) && (floormod(((threadIdx.x_1*4) + 3), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 3), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 3), 9)) - 8)], 0f32, dtype=float32)
+ for (rc.outer.outer: int32, 0, 32) {
+ let cse_var_1: int32 = (rc.outer.outer*784)
+ {
+ attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1: Buffer(pad_temp.shared, float32, [432], [], scope="shared")[threadIdx.x_1] = @tir.if_then_else(((((1 <= (floordiv(floormod(threadIdx.x_1, 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod(threadIdx.x_1, 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[(((((cse_var_1 + (floordiv(threadIdx.x_1, 27)*49)) + (floordiv(floormod(threadIdx.x_1, 27), 9)*7)) + (floormod(blockIdx.x, 7)*7 [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 32)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 32), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 32), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 5), 9))) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 32), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 32), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floormod( [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 64)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 64), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 64), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 64), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 64), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floormod( [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 96)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 96), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 96), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 96), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 96), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floormod( [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 128)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 128), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 128), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 128), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 128), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 160)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 160), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 160), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 7), 9))) && (floormod((threadIdx.x_1 + 7), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 160), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 160), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 192)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 192), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 192), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 3), 9))) && (floormod((threadIdx.x_1 + 3), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 192), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 192), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 224), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 224), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 8), 9))) && (floormod((threadIdx.x_1 + 8), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 224), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 224), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 256)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 256), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 256), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 4), 9))) && (floormod((threadIdx.x_1 + 4), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 256), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 256), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 288)] = @tir.if_then_else(((((1 <= (floormod(blockIdx.x, 7) + floormod((floordiv(threadIdx.x_1, 9) + 2), 3))) && ((floormod(blockIdx.x, 7) + floormod((floordiv(threadIdx.x_1, 9) + 2), 3)) < 8)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 288), 27)*49)) + (floormod(blockIdx.x, 7)*7)) + (floormod((floordiv(threadIdx.x_1, 9) + 2), 3)*7)) + floormod(threadIdx.x_1, 9 [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 320)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 320), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 320), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 5), 9))) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 320), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 320), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 352)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 352), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 352), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 352), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 352), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 384)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 384), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 384), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 384), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 384), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ if @tir.likely((threadIdx.x_1 < 16), dtype=bool) {
+ pad_temp.shared_1[(threadIdx.x_1 + 416)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 416), 27), 9) + floormod(blockIdx.x, 7)) < 8) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 416), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 416), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
+ }
+ for (ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer: int32, 0, 144) {
+ attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ kernel.shared_1: Buffer(kernel.shared, float32, [4608], [], scope="shared")[((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*32) + threadIdx.x_2)] = kernel[((((floordiv(blockIdx.x, 7)*147456) + (floordiv(((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*2) + floordiv(threadIdx.x_2, 16)), 9)*4608)) + (rc.outer.outer*144)) + floormod(((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*32) + threadIdx.x_2), 144))]
+ }
+ for (rc.outer.inner: int32, 0, 4) {
+ for (ry.outer.inner: int32, 0, 3) {
+ for (rc.inner: int32, 0, 4) {
+ let cse_var_9: int32 = (((rc.outer.inner*108) + (rc.inner*27)) + (ry.outer.inner*9))
+ let cse_var_8: int32 = (cse_var_9 + 7)
+ let cse_var_7: int32 = (cse_var_9 + 6)
+ let cse_var_6: int32 = (cse_var_9 + 5)
+ let cse_var_5: int32 = (cse_var_9 + 4)
+ let cse_var_4: int32 = (cse_var_9 + 3)
+ let cse_var_3: int32 = (cse_var_9 + 2)
+ let cse_var_2: int32 = (cse_var_9 + 1)
+ {
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_9]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_2]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_2]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[cse_var_8]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_8]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(cse_var_9 + 8)]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ }
}
}
- attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1: Buffer(kernel.shared, float32, [3072], [], scope="shared")[threadIdx.x_2] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(threadIdx.x_2, 24)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 64)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 8), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 64), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 128)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 16), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 128), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 192)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 36864)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 256)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 32), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 256), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 320)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 40), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 320), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 384)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 73728)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 448)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 56), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 448), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 512)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 64), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 512), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 576)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 110592)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 640)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 80), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 640), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 704)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 88), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 704), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 768)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 147456)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 832)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 104), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 832), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 896)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 112), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 896), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 960)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 184320)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1024)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 128), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1024), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1088)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 136), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1088), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1152)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 221184)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1216)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 152), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1216), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1280)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 160), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1280), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1344)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 258048)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1408)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 176), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1408), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1472)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 184), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1472), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1536)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 294912)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1600)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 200), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1600), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1664)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 208), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1664), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1728)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 331776)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1792)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 224), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1792), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1856)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 232), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1856), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1920)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 368640)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1984)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 248), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1984), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2048)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 256), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2048), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2112)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 405504)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2176)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 272), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2176), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2240)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 280), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2240), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2304)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 442368)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2368)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 296), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2368), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2432)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 304), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2432), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2496)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 479232)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2560)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 320), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2560), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2624)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 328), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2624), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2688)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 516096)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2752)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 344), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2752), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2816)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 352), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2816), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2880)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 552960)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2944)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 368), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2944), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 3008)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 376), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 3008), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[0]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[9]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[1]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[2]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[3]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[4]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[5]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[6]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[0]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[9]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[8]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[17]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[8]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[17]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[18]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[27]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[18]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[27]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[26]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[35]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[26]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[35]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[36]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[45]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[36]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[45]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[44]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[53]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[44]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[53]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[54]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[63]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[54]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[63]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[62]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[71]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[62]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[71]*kernel.shared_1[((threadIdx.x*48) + 47)]))
}
}
}
- for (i1.inner: int32, 0, 2) {
- for (i3.inner: int32, 0, 7) {
- compute[(((((floordiv(blockIdx.x, 7)*6272) + (threadIdx.x*98)) + (i1.inner*49)) + (floormod(blockIdx.x, 7)*7)) + i3.inner)] = max((conv2d_nchw_1[((i1.inner*7) + i3.inner)] + bias[(((floordiv(blockIdx.x, 7)*128) + (threadIdx.x*2)) + i1.inner)]), 0f32)
- }
- }
+ compute[(((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7))] = max((conv2d_nchw_1[0] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 1)] = max((conv2d_nchw_1[1] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 2)] = max((conv2d_nchw_1[2] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 3)] = max((conv2d_nchw_1[3] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 4)] = max((conv2d_nchw_1[4] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 5)] = max((conv2d_nchw_1[5] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 6)] = max((conv2d_nchw_1[6] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
}
}
@@ -751,7 +368,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 0.362 ms
+ Execution time of this operator: 0.343 ms
@@ -796,36 +413,36 @@ 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=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_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=32)
conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=1)
conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=1)
conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
- conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
+ conv2d_nchw_xx_o_o_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=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=2)
+ 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=7)
+ conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=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=3)
+ conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
+ conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=3)
+ conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2 [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
- compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
- compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=64)
+ 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=32)
compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
- compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=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=1)
- compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
+ compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=7)
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)
kernel_shared = s.cache_read(kernel, "shared", [conv2d_nchw])
@@ -844,14 +461,14 @@ 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=64)
+ 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=32)
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=4)
+ 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)
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=64)
+ 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=32)
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", 512)
+ 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, "unroll_explicit", True)
CUDA source code:
@@ -869,10 +486,10 @@ 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__(64) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
- float conv2d_nchw[14];
- __shared__ float pad_temp_shared[72];
- __shared__ float kernel_shared[3072];
+ extern "C" __global__ void __launch_bounds__(32) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+ float conv2d_nchw[7];
+ __shared__ float pad_temp_shared[432];
+ __shared__ float kernel_shared[4608];
conv2d_nchw[0] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
@@ -880,420 +497,63 @@ They can be used for debugging and learning the behavior of the auto-scheduler.
conv2d_nchw[4] = 0.000000e+00f;
conv2d_nchw[5] = 0.000000e+00f;
conv2d_nchw[6] = 0.000000e+00f;
- conv2d_nchw[7] = 0.000000e+00f;
- conv2d_nchw[8] = 0.000000e+00f;
- conv2d_nchw[9] = 0.000000e+00f;
- conv2d_nchw[10] = 0.000000e+00f;
- conv2d_nchw[11] = 0.000000e+00f;
- conv2d_nchw[12] = 0.000000e+00f;
- conv2d_nchw[13] = 0.000000e+00f;
- for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
- for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
- __syncthreads();
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) * 4) % 9))) && (((((int)threadIdx.x) * 4) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 1) % 9))) && ((((((int)threadIdx.x) * 4) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 2) % 9))) && ((((((int)threadIdx.x) * 4) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 3) % 9))) && ((((((int)threadIdx.x) * 4) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
- }
- kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
- kernel_shared[(((int)threadIdx.x) + 64)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 128)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
- kernel_shared[(((int)threadIdx.x) + 256)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 320)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 512)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
- kernel_shared[(((int)threadIdx.x) + 640)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 704)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
- kernel_shared[(((int)threadIdx.x) + 832)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
- kernel_shared[(((int)threadIdx.x) + 1024)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1088)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
- kernel_shared[(((int)threadIdx.x) + 1216)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1280)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
- kernel_shared[(((int)threadIdx.x) + 1408)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1472)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
- kernel_shared[(((int)threadIdx.x) + 1600)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1664)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
- kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1856)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
- kernel_shared[(((int)threadIdx.x) + 1984)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2048)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
- kernel_shared[(((int)threadIdx.x) + 2176)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
- kernel_shared[(((int)threadIdx.x) + 2368)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2432)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
- kernel_shared[(((int)threadIdx.x) + 2560)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2624)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
- kernel_shared[(((int)threadIdx.x) + 2752)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2816)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
- kernel_shared[(((int)threadIdx.x) + 2944)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 3008)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- __syncthreads();
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[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 rc_outer_outer = 0; rc_outer_outer < 32; ++rc_outer_outer) {
+ __syncthreads();
+ pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 27) / 9) + (((int)blockIdx.x) % 7))) && ((((((int)threadIdx.x) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 27) * 49)) + (((((int)threadIdx.x) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
+ pad_temp_shared[(((int)threadIdx.x) + 32)] = (((((1 <= ((((((int)threadIdx.x) + 5) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 5) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 5) % 9))) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 32) / 27) * 49)) + ((((((int)threadIdx.x) + 5) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] : 0.0 [...]
+ pad_temp_shared[(((int)threadIdx.x) + 64)] = (((((1 <= ((((((int)threadIdx.x) + 10) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 10) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 64) / 27) * 49)) + ((((((int)threadIdx.x) + 10) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : [...]
+ pad_temp_shared[(((int)threadIdx.x) + 96)] = (((((1 <= ((((((int)threadIdx.x) + 15) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 15) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 96) / 27) * 49)) + ((((((int)threadIdx.x) + 15) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : [...]
+ pad_temp_shared[(((int)threadIdx.x) + 128)] = (((((1 <= ((((((int)threadIdx.x) + 20) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 20) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 128) / 27) * 49)) + ((((((int)threadIdx.x) + 20) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] [...]
+ pad_temp_shared[(((int)threadIdx.x) + 160)] = (((((1 <= ((((((int)threadIdx.x) + 25) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 25) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 160) / 27) * 49)) + ((((((int)threadIdx.x) + 25) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 7) % 9)) - 8)] [...]
+ pad_temp_shared[(((int)threadIdx.x) + 192)] = (((((1 <= ((((((int)threadIdx.x) + 3) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 3) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 3) % 9))) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 192) / 27) * 49)) + ((((((int)threadIdx.x) + 3) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 3) % 9)) - 8)] : 0 [...]
+ pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= ((((((int)threadIdx.x) + 8) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 8) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 224) / 27) * 49)) + ((((((int)threadIdx.x) + 8) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 8) % 9)) - 8)] : 0 [...]
+ pad_temp_shared[(((int)threadIdx.x) + 256)] = (((((1 <= ((((((int)threadIdx.x) + 13) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 13) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 256) / 27) * 49)) + ((((((int)threadIdx.x) + 13) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 4) % 9)) - 8)] [...]
+ pad_temp_shared[(((int)threadIdx.x) + 288)] = (((((1 <= ((((int)blockIdx.x) % 7) + (((((int)threadIdx.x) / 9) + 2) % 3))) && (((((int)blockIdx.x) % 7) + (((((int)threadIdx.x) / 9) + 2) % 3)) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 288) / 27) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + ((((((int)threadIdx.x) / 9) + 2) % 3) * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
+ pad_temp_shared[(((int)threadIdx.x) + 320)] = (((((1 <= ((((((int)threadIdx.x) + 23) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 23) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 5) % 9))) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 320) / 27) * 49)) + ((((((int)threadIdx.x) + 23) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 5) % 9)) - 8)] [...]
+ pad_temp_shared[(((int)threadIdx.x) + 352)] = (((((1 <= ((((((int)threadIdx.x) + 1) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 1) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 352) / 27) * 49)) + ((((((int)threadIdx.x) + 1) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 1) % 9)) - 8)] : 0 [...]
+ pad_temp_shared[(((int)threadIdx.x) + 384)] = (((((1 <= ((((((int)threadIdx.x) + 6) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 6) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 384) / 27) * 49)) + ((((((int)threadIdx.x) + 6) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 6) % 9)) - 8)] : 0 [...]
+ if (((int)threadIdx.x) < 16) {
+ pad_temp_shared[(((int)threadIdx.x) + 416)] = ((((((((((int)threadIdx.x) + 11) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 416) / 27) * 49)) + ((((((int)threadIdx.x) + 11) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
}
- }
- for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
- for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[((((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 98)) + (i1_inner * 49)) + ((((int)blockIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[((((((int)blockIdx.x) / 7) * 128) + (((int)threadIdx.x) * 2)) + i1_inner)]), 0.000000e+00f);
+ for (int ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer = 0; ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer < 144; ++ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer) {
+ kernel_shared[((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 32) + ((int)threadIdx.x))] = kernel[(((((((int)blockIdx.x) / 7) * 147456) + ((((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 2) + (((int)threadIdx.x) >> 4)) / 9) * 4608)) + (rc_outer_outer * 144)) + (((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 32) + ((int)threadIdx.x)) % 144))];
+ }
+ __syncthreads();
+ for (int rc_outer_inner = 0; rc_outer_inner < 4; ++rc_outer_inner) {
+ for (int ry_outer_inner = 0; ry_outer_inner < 3; ++ry_outer_inner) {
+ for (int rc_inner = 0; rc_inner < 4; ++rc_inner) {
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9))] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 1)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 1)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 7)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 7)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 8)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ }
+ }
}
}
+ compute[((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7))] = max((conv2d_nchw[0] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 1)] = max((conv2d_nchw[1] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 2)] = max((conv2d_nchw[2] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 3)] = max((conv2d_nchw[3] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 4)] = max((conv2d_nchw[4] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 5)] = max((conv2d_nchw[5] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 6)] = max((conv2d_nchw[6] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
}
@@ -1351,7 +611,7 @@ In the example below we resume the status and do more 5 trials.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 2 minutes 17.757 seconds)
+ **Total running time of the script:** ( 2 minutes 20.501 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 40cc6ce91..e936e9dea 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
@@ -614,7 +614,7 @@ 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)
- 9.9415 9.9341 9.9663 9.9240 0.0180
+ 9.7607 9.7678 9.8201 9.6943 0.0516
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 72165732c..ef791576a 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
@@ -633,7 +633,7 @@ 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)
- 764.1780 765.6010 766.2926 760.6404 2.5173
+ 766.9284 768.4643 770.4900 761.8308 3.6982
@@ -658,7 +658,7 @@ Other Tips
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 20.507 seconds)
+ **Total running time of the script:** ( 1 minutes 21.450 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 9a0741a54..8bd0f30b4 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
@@ -362,74 +362,80 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [65536], []),
compute: Buffer(compute_2: Pointer(float32), float32, [65536], [])}
buffer_map = {placeholder_5: placeholder, placeholder_6: placeholder_1, placeholder_7: placeholder_2, placeholder_8: placeholder_3, placeholder_9: placeholder_4, compute_1: compute}
- preflattened_buffer_map = {placeholder_7: placeholder_15: Buffer(placeholder_12, int32, [4916], []), placeholder_5: placeholder_16: Buffer(placeholder_10, float32, [128, 256], []), placeholder_6: placeholder_17: Buffer(placeholder_11, float32, [4916, 16, 1], []), placeholder_8: placeholder_18: Buffer(placeholder_13, int32, [33], []), compute_1: compute_3: Buffer(compute_2, float32, [128, 512], []), placeholder_9: placeholder_19: Buffer(placeholder_14, float32, [128, 512], [])} {
- for (i0.outer.i1.outer.fused: int32, 0, 512) "parallel" {
- allocate(compute_4: Pointer(global float32), float32, [128]), storage_scope = global {
- for (i.inner.init: int32, 0, 8) {
- let cse_var_1: int32 = (i.inner.init*16)
- {
- compute_5: Buffer(compute_4, float32, [128], [])[cse_var_1] = 0f32
- compute_5[(cse_var_1 + 1)] = 0f32
- compute_5[(cse_var_1 + 2)] = 0f32
- compute_5[(cse_var_1 + 3)] = 0f32
- compute_5[(cse_var_1 + 4)] = 0f32
- compute_5[(cse_var_1 + 5)] = 0f32
- compute_5[(cse_var_1 + 6)] = 0f32
- compute_5[(cse_var_1 + 7)] = 0f32
- compute_5[(cse_var_1 + 8)] = 0f32
- compute_5[(cse_var_1 + 9)] = 0f32
- compute_5[(cse_var_1 + 10)] = 0f32
- compute_5[(cse_var_1 + 11)] = 0f32
- compute_5[(cse_var_1 + 12)] = 0f32
- compute_5[(cse_var_1 + 13)] = 0f32
- compute_5[(cse_var_1 + 14)] = 0f32
- compute_5[(cse_var_1 + 15)] = 0f32
- }
- }
- for (elem_idx: int32, 0, let cse_var_2: int32 = floormod(i0.outer.i1.outer.fused, 32) in (placeholder_3[(cse_var_2 + 1)] - placeholder_3[cse_var_2])) {
- for (i.inner: int32, 0, 8) {
- let cse_var_21: int32 = floormod(i0.outer.i1.outer.fused, 32)
- let cse_var_20: int32 = (i.inner*16)
- let cse_var_19: int32 = (elem_idx*16)
- let cse_var_18: int32 = (cse_var_20 + 10)
- let cse_var_17: int32 = (cse_var_20 + 11)
- let cse_var_16: int32 = (cse_var_20 + 12)
- let cse_var_15: int32 = (cse_var_20 + 13)
- let cse_var_14: int32 = (cse_var_20 + 14)
- let cse_var_13: int32 = (cse_var_20 + 15)
- let cse_var_12: int32 = (cse_var_20 + 2)
- let cse_var_11: int32 = (cse_var_20 + 3)
- let cse_var_10: int32 = (cse_var_20 + 4)
- let cse_var_9: int32 = (cse_var_20 + 5)
- let cse_var_8: int32 = (cse_var_20 + 6)
- let cse_var_7: int32 = (cse_var_20 + 7)
- let cse_var_6: int32 = (cse_var_20 + 8)
- let cse_var_5: int32 = (cse_var_20 + 9)
- let cse_var_4: int32 = (cse_var_20 + 1)
- let cse_var_3: int32 = ((floordiv(i0.outer.i1.outer.fused, 32)*2048) + (i.inner*256))
- {
- compute_5[cse_var_20] = (compute_5[cse_var_20] + (placeholder_1[((placeholder_3[cse_var_21]*16) + cse_var_19)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_4] = (compute_5[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 1)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_12] = (compute_5[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 2)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_11] = (compute_5[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 3)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_10] = (compute_5[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 4)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_9] = (compute_5[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 5)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_8] = (compute_5[cse_var_8] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 6)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_7] = (compute_5[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 7)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_6] = (compute_5[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 8)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_5] = (compute_5[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 9)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_18] = (compute_5[cse_var_18] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 10)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_17] = (compute_5[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 11)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_16] = (compute_5[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 12)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_15] = (compute_5[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 13)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_14] = (compute_5[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 14)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_13] = (compute_5[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 15)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
+ preflattened_buffer_map = {placeholder_8: placeholder_15: Buffer(placeholder_13, int32, [33], []), compute_1: compute_3: Buffer(compute_2, float32, [128, 512], []), placeholder_9: placeholder_16: Buffer(placeholder_14, float32, [128, 512], []), placeholder_6: placeholder_17: Buffer(placeholder_11, float32, [4916, 16, 1], []), placeholder_7: placeholder_18: Buffer(placeholder_12, int32, [4916], []), placeholder_5: placeholder_19: Buffer(placeholder_10, float32, [128, 256], [])} {
+ for (i0.outer.i1.outer.fused: int32, 0, 16) "parallel" {
+ allocate(compute_4: Pointer(global float32), float32, [4096]), storage_scope = global {
+ for (i.outer.inner: int32, 0, 4) {
+ for (nb_j.inner: int32, 0, 2) {
+ for (i.inner.init: int32, 0, 32) {
+ let cse_var_1: int32 = (((i.outer.inner*1024) + (i.inner.init*32)) + (nb_j.inner*16))
+ {
+ compute_5: Buffer(compute_4, float32, [4096], [])[cse_var_1] = 0f32
+ compute_5[(cse_var_1 + 1)] = 0f32
+ compute_5[(cse_var_1 + 2)] = 0f32
+ compute_5[(cse_var_1 + 3)] = 0f32
+ compute_5[(cse_var_1 + 4)] = 0f32
+ compute_5[(cse_var_1 + 5)] = 0f32
+ compute_5[(cse_var_1 + 6)] = 0f32
+ compute_5[(cse_var_1 + 7)] = 0f32
+ compute_5[(cse_var_1 + 8)] = 0f32
+ compute_5[(cse_var_1 + 9)] = 0f32
+ compute_5[(cse_var_1 + 10)] = 0f32
+ compute_5[(cse_var_1 + 11)] = 0f32
+ compute_5[(cse_var_1 + 12)] = 0f32
+ compute_5[(cse_var_1 + 13)] = 0f32
+ compute_5[(cse_var_1 + 14)] = 0f32
+ compute_5[(cse_var_1 + 15)] = 0f32
+ }
+ }
+ for (elem_idx: int32, 0, let cse_var_2: int32 = ((i0.outer.i1.outer.fused*2) + nb_j.inner) in (placeholder_3[(cse_var_2 + 1)] - placeholder_3[cse_var_2])) {
+ for (i.inner: int32, 0, 32) {
+ let cse_var_21: int32 = (elem_idx*16)
+ let cse_var_20: int32 = ((i0.outer.i1.outer.fused*2) + nb_j.inner)
+ let cse_var_19: int32 = ((i.outer.inner*8192) + (i.inner*256))
+ let cse_var_18: int32 = (((i.outer.inner*1024) + (i.inner*32)) + (nb_j.inner*16))
+ let cse_var_17: int32 = (cse_var_18 + 1)
+ let cse_var_16: int32 = (cse_var_18 + 11)
+ let cse_var_15: int32 = (cse_var_18 + 12)
+ let cse_var_14: int32 = (cse_var_18 + 13)
+ let cse_var_13: int32 = (cse_var_18 + 14)
+ let cse_var_12: int32 = (cse_var_18 + 15)
+ let cse_var_11: int32 = (cse_var_18 + 2)
+ let cse_var_10: int32 = (cse_var_18 + 3)
+ let cse_var_9: int32 = (cse_var_18 + 4)
+ let cse_var_8: int32 = (cse_var_18 + 5)
+ let cse_var_7: int32 = (cse_var_18 + 6)
+ let cse_var_6: int32 = (cse_var_18 + 7)
+ let cse_var_5: int32 = (cse_var_18 + 8)
+ let cse_var_4: int32 = (cse_var_18 + 9)
+ let cse_var_3: int32 = (cse_var_18 + 10)
+ {
+ compute_5[cse_var_18] = (compute_5[cse_var_18] + (placeholder_1[((placeholder_3[cse_var_20]*16) + cse_var_21)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_17] = (compute_5[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 1)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_11] = (compute_5[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 2)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_10] = (compute_5[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 3)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_9] = (compute_5[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 4)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_8] = (compute_5[cse_var_8] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 5)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_7] = (compute_5[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 6)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_6] = (compute_5[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 7)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_5] = (compute_5[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 8)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_4] = (compute_5[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 9)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_3] = (compute_5[cse_var_3] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 10)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_16] = (compute_5[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 11)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_15] = (compute_5[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 12)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_14] = (compute_5[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 13)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_13] = (compute_5[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 14)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_12] = (compute_5[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 15)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ }
+ }
}
}
}
- for (i0.inner: int32, 0, 8) {
- let cse_var_22: int32 = (((floordiv(i0.outer.i1.outer.fused, 32)*4096) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 32)*16))
- compute[ramp(cse_var_22, 1, 16)] = max((compute_5[ramp((i0.inner*16), 1, 16)] + placeholder_4[ramp(cse_var_22, 1, 16)]), broadcast(0f32, 16))
+ for (i0.inner: int32, 0, 128) {
+ for (i1.inner: int32, 0, 32) {
+ let cse_var_22: int32 = (((i0.inner*512) + (i0.outer.i1.outer.fused*32)) + i1.inner)
+ compute[cse_var_22] = max((compute_5[((i0.inner*32) + i1.inner)] + placeholder_4[cse_var_22]), 0f32)
+ }
}
}
}
@@ -483,7 +489,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 1.877 ms
+ Execution time of this operator: 1.718 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 16e4f63ac..b7ded2705 100644
--- a/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
+++ b/docs/_sources/how_to/tune_with_autotvm/sg_execution_times.rst.txt
@@ -5,10 +5,10 @@
Computation times
=================
-**00:43.302** total execution time for **how_to_tune_with_autotvm** files:
+**00:44.660** total execution time for **how_to_tune_with_autotvm** files:
-- **00:42.463**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)
-- **00:00.222**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)
-- **00:00.208**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)
-- **00:00.208**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``)
-- **00:00.202**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``)
+- **00:43.786**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_conv2d_cuda.py` (``tune_conv2d_cuda.py``)
+- **00:00.230**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_x86.py` (``tune_relay_x86.py``)
+- **00:00.217**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_cuda.py` (``tune_relay_cuda.py``)
+- **00:00.214**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_arm.py` (``tune_relay_arm.py``)
+- **00:00.213**: :ref:`sphx_glr_how_to_tune_with_autotvm_tune_relay_mobile_gpu.py` (``tune_relay_mobile_gpu.py``)
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 ad39e2963..d8367b28a 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
@@ -859,8 +859,8 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 4, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2885496
- No: 6 GFLOPS: 64.10/64.10 result: MeasureResult(costs=(0.0036112859666666665,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5755383968353271, timestamp=1650564545.7379923) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
- No: 7 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 6 GFLOPS: 103.76/103.76 result: MeasureResult(costs=(0.002231081375,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5949790477752686, timestamp=1650567454.3841782) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
+ No: 7 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -983,7 +983,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 16, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6225319
- No: 8 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 8 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1106,7 +1106,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,943546
- No: 9 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 9 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1229,7 +1229,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2868708
- No: 10 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 10 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 142, in build
res = future.result()
File "/usr/lib/python3.7/concurrent/futures/_base.py", line 435, in result
@@ -1247,7 +1247,7 @@ for this template
TimeoutError
[('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4691833
- No: 11 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 11 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1370,7 +1370,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 2, 64]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1042124
- No: 12 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 12 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1493,7 +1493,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10013405
- No: 13 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 13 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1616,7 +1616,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6732082
- No: 14 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 14 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1739,7 +1739,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7536735
- No: 15 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 15 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1862,7 +1862,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,482121
- No: 16 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 16 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1985,7 +1985,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2824525
- No: 17 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 17 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2108,7 +2108,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4559286
- No: 18 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 18 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2231,7 +2231,7 @@ for this template
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 32, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 512]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9677544
- No: 19 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+ No: 19 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 721, in __call__
yield remote, remote.load_module(os.path.split(build_result.filename)[1])
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 685, in run_through_rpc
@@ -2319,7 +2319,7 @@ for this template
15: _PyEval_EvalFrameDefault
14: 0x0000000000537c30
13: _PyObject_FastCallKeywords
- 12: 0x00007fa3d62befa2
+ 12: 0x00007f23d9b6ffa2
11: _ctypes_callproc
10: ffi_call
9: ffi_call_unix64
@@ -2384,7 +2384,7 @@ for this template
21: _PyFunction_FastCallKeywords
20: _PyEval_EvalFrameDefault
19: _PyFunction_FastCall [('tile_f', [-1, 8, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6390073
- No: 20 GFLOPS: 144.30/144.30 result: MeasureResult(costs=(0.00160430567,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4274165630340576, timestamp=1650564571.984725) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
+ No: 20 GFLOPS: 144.39/144.39 result: MeasureResult(costs=(0.0016032561200000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4371635913848877, timestamp=1650567480.202208) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
@@ -2437,7 +2437,7 @@ and measure running time.
Best config:
[('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
- Time cost of this operator: 0.001981
+ Time cost of this operator: 0.001958
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 35bc003a3..4492922d5 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
@@ -292,10 +292,10 @@ Timing the untuned program
########## Build without Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs
--------- --- -------- ------- ----- ------ -------
- tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 314.2 98.755 (1, 2, 10, 10, 3) 2 1
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.061 0.962 (1, 6, 10, 10) 1 1
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.901 0.283 (1, 1, 10, 10, 3) 1 1
- Total_time - 318.162 - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 321.4 98.767 (1, 2, 10, 10, 3) 2 1
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.094 0.951 (1, 6, 10, 10) 1 1
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.917 0.282 (1, 1, 10, 10, 3) 1 1
+ Total_time - 325.411 - - - -
@@ -357,10 +357,10 @@ Timing the tuned program
########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs
--------- --- -------- ------- ----- ------ -------
- tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 79.9 96.727 (1, 6, 10, 10, 1) 2 1
- tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.789 2.166 (1, 6, 10, 10) 1 1
- tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.915 1.107 (1, 1, 10, 10, 3) 1 1
- Total_time - 82.604 - - - -
+ tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 124.5 97.883 (1, 6, 10, 10, 1) 2 1
+ tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.77 1.392 (1, 6, 10, 10) 1 1
+ tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.922 0.725 (1, 1, 10, 10, 3) 1 1
+ Total_time - 127.192 - - - -
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 3b51f0580..f9faea394 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
@@ -5,10 +5,10 @@
Computation times
=================
-**00:43.796** total execution time for **how_to_work_with_microtvm** files:
+**00:45.418** total execution time for **how_to_work_with_microtvm** files:
-- **00:39.782**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)
-- **00:03.440**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)
-- **00:00.195**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)
-- **00:00.191**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``)
-- **00:00.189**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``)
+- **00:41.261**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_autotune.py` (``micro_autotune.py``)
+- **00:03.559**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tflite.py` (``micro_tflite.py``)
+- **00:00.202**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_ethosu.py` (``micro_ethosu.py``)
+- **00:00.198**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_reference_vm.py` (``micro_reference_vm.py``)
+- **00:00.198**: :ref:`sphx_glr_how_to_work_with_microtvm_micro_tvmc.py` (``micro_tvmc.py``)
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 49b601751..ad0039df4 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
@@ -5,8 +5,8 @@
Computation times
=================
-**00:05.545** total execution time for **how_to_work_with_relay** files:
+**00:09.219** total execution time for **how_to_work_with_relay** files:
-- **00:03.926**: :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)
-- **00:01.410**: :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)
-- **00:00.209**: :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)
+- **00:07.083**: :ref:`sphx_glr_how_to_work_with_relay_using_external_lib.py` (``using_external_lib.py``)
+- **00:01.922**: :ref:`sphx_glr_how_to_work_with_relay_build_gcn.py` (``build_gcn.py``)
+- **00:00.215**: :ref:`sphx_glr_how_to_work_with_relay_using_relay_viz.py` (``using_relay_viz.py``)
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 11e6bb12a..73fac8922 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
@@ -5,13 +5,13 @@
Computation times
=================
-**00:05.373** total execution time for **how_to_work_with_schedules** files:
+**00:05.598** total execution time for **how_to_work_with_schedules** files:
-- **00:01.951**: :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)
-- **00:01.064**: :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)
-- **00:00.690**: :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)
-- **00:00.686**: :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)
-- **00:00.307**: :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)
-- **00:00.233**: :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``)
-- **00:00.229**: :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)
-- **00:00.213**: :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)
+- **00:02.057**: :ref:`sphx_glr_how_to_work_with_schedules_intrin_math.py` (``intrin_math.py``)
+- **00:01.109**: :ref:`sphx_glr_how_to_work_with_schedules_tensorize.py` (``tensorize.py``)
+- **00:00.723**: :ref:`sphx_glr_how_to_work_with_schedules_reduction.py` (``reduction.py``)
+- **00:00.705**: :ref:`sphx_glr_how_to_work_with_schedules_scan.py` (``scan.py``)
+- **00:00.308**: :ref:`sphx_glr_how_to_work_with_schedules_extern_op.py` (``extern_op.py``)
+- **00:00.242**: :ref:`sphx_glr_how_to_work_with_schedules_schedule_primitives.py` (``schedule_primitives.py``)
+- **00:00.234**: :ref:`sphx_glr_how_to_work_with_schedules_tedd.py` (``tedd.py``)
+- **00:00.220**: :ref:`sphx_glr_how_to_work_with_schedules_tuple_inputs.py` (``tuple_inputs.py``)
diff --git a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
index e8c9bb7db..2ffab1315 100644
--- a/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
+++ b/docs/_sources/how_to/work_with_schedules/tensorize.rst.txt
@@ -318,7 +318,7 @@ The importing needs to happen before the tensorized GEMV being executed.
C: Buffer(C_2: Pointer(float32), float32, [524288], [])}
buffer_map = {A_1: A, B_1: B, C_1: C}
preflattened_buffer_map = {A_1: A_3: Buffer(A_2, float32, [1024, 64], []), B_1: B_3: Buffer(B_2, float32, [512, 64], []), C_1: C_3: Buffer(C_2, float32, [1024, 512], [])} {
- attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmpsv82ynx1/input0.cc'\nsource_filename = \"/tmp/tmpsv82ynx1/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca float*, align 8\n %8 = alloca float*, align 8\n %9 = alloca floa [...]
+ attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmp0hu7m0hl/input0.cc'\nsource_filename = \"/tmp/tmp0hu7m0hl/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = alloca float*, align 8\n %8 = alloca float*, align 8\n %9 = alloca floa [...]
for (i, 0, 1024) {
for (j.outer: int32, 0, 32) {
@tir.call_extern("gemv_update", @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), C_2, ((i*512) + (j.outer*16)), 16, 2, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), A_2, (i*64), 64, 1, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), B_2, (j.outer*1024), 1024, 1, dtype=handle), 16, 64, 64, dtype=int32)
diff --git a/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt b/docs/_sources/topic/vta/tutorials/autotvm/sg_execution_times.rst.txt
index da63e5d92..12dd53b59 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
@@ -5,7 +5,7 @@
Computation times
=================
-**00:20.491** total execution time for **topic_vta_tutorials_autotvm** files:
+**00:21.327** total execution time for **topic_vta_tutorials_autotvm** files:
-- **00:20.291**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``)
-- **00:00.200**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)
+- **00:21.119**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_relay_vta.py` (``tune_relay_vta.py``)
+- **00:00.207**: :ref:`sphx_glr_topic_vta_tutorials_autotvm_tune_alu_vta.py` (``tune_alu_vta.py``)
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 67ddffcf1..c655b9284 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_classification.rst.txt
@@ -265,7 +265,7 @@ 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 21.68s!
+ resnet18_v1 inference graph built in 22.58s!
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 60037fcc5..5ed19397f 100644
--- a/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/frontend/deploy_detection.rst.txt
@@ -301,7 +301,7 @@ The compilation steps are:
/workspace/python/tvm/relay/build_module.py:439: DeprecationWarning: Please use input parameter mod (tvm.IRModule) instead of deprecated parameter mod (tvm.relay.function.Function)
DeprecationWarning,
- yolov3-tiny inference graph built in 14.97s!
+ yolov3-tiny inference graph built in 15.52s!
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 09c4c2078..68771aa6f 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
@@ -5,7 +5,7 @@
Computation times
=================
-**01:28.893** total execution time for **topic_vta_tutorials_frontend** files:
+**01:29.340** total execution time for **topic_vta_tutorials_frontend** files:
-- **00:47.096**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)
-- **00:41.797**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``)
+- **00:46.975**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_detection.py` (``deploy_detection.py``)
+- **00:42.365**: :ref:`sphx_glr_topic_vta_tutorials_frontend_deploy_classification.py` (``deploy_classification.py``)
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 982f16914..ed24e664e 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
@@ -5,7 +5,7 @@
Computation times
=================
-**00:03.485** total execution time for **topic_vta_tutorials_optimize** files:
+**00:03.522** total execution time for **topic_vta_tutorials_optimize** files:
-- **00:02.968**: :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)
-- **00:00.516**: :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``)
+- **00:02.980**: :ref:`sphx_glr_topic_vta_tutorials_optimize_convolution_opt.py` (``convolution_opt.py``)
+- **00:00.541**: :ref:`sphx_glr_topic_vta_tutorials_optimize_matrix_multiply_opt.py` (``matrix_multiply_opt.py``)
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 65928edd6..ba426ff40 100644
--- a/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
+++ b/docs/_sources/topic/vta/tutorials/sg_execution_times.rst.txt
@@ -5,7 +5,7 @@
Computation times
=================
-**00:00.924** total execution time for **topic_vta_tutorials** files:
+**00:00.977** total execution time for **topic_vta_tutorials** files:
-- **00:00.468**: :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``)
-- **00:00.456**: :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``)
+- **00:00.496**: :ref:`sphx_glr_topic_vta_tutorials_matrix_multiply.py` (``matrix_multiply.py``)
+- **00:00.481**: :ref:`sphx_glr_topic_vta_tutorials_vta_get_started.py` (``vta_get_started.py``)
diff --git a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
index 5a34190f5..4feb510f8 100644
--- a/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
+++ b/docs/_sources/tutorial/auto_scheduler_matmul_x86.rst.txt
@@ -306,7 +306,7 @@ We build the binary and check its correctness and performance.
.. code-block:: none
- Execution time of this operator: 93.265 ms
+ Execution time of this operator: 93.368 ms
diff --git a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
index 35604c233..eaa25e02e 100644
--- a/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
+++ b/docs/_sources/tutorial/autotvm_relay_x86.rst.txt
@@ -268,7 +268,7 @@ standard deviation.
.. code-block:: none
- {'mean': 494.8987510900406, 'median': 494.9888973500492, 'std': 0.4855438254512492}
+ {'mean': 497.7622343500014, 'median': 497.80023144999745, 'std': 0.6134268998812399}
@@ -482,31 +482,31 @@ the tuning data to.
.. code-block:: none
-
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 1/25] Current/Best: 23.11/ 23.80 GFLOPS | Progress: (4/10) | 4.59 s
[Task 1/25] Current/Best: 17.10/ 23.80 GFLOPS | Progress: (8/10) | 7.19 s
[Task 1/25] Current/Best: 5.42/ 23.80 GFLOPS | Progress: (10/10) | 8.84 s Done.
-
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 2/25] Current/Best: 6.75/ 22.43 GFLOPS | Progress: (4/10) | 2.57 s
[Task 2/25] Current/Best: 12.87/ 22.43 GFLOPS | Progress: (8/10) | 4.23 s
[Task 2/25] Current/Best: 6.68/ 22.43 GFLOPS | Progress: (10/10) | 4.81 s Done.
-
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 3/25] Current/Best: 16.43/ 20.80 GFLOPS | Progress: (4/10) | 2.61 s
[Task 3/25] Current/Best: 9.54/ 20.80 GFLOPS | Progress: (8/10) | 4.70 s
[Task 3/25] Current/Best: 17.74/ 20.80 GFLOPS | Progress: (10/10) | 5.80 s Done.
-
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 4/25] Current/Best: 15.33/ 21.80 GFLOPS | Progress: (4/10) | 2.35 s
[Task 4/25] Current/Best: 13.01/ 21.80 GFLOPS | Progress: (8/10) | 4.01 s
[Task 4/25] Current/Best: 17.05/ 21.80 GFLOPS | Progress: (10/10) | 4.64 s Done.
-
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 5/25] Current/Best: 15.09/ 19.25 GFLOPS | Progress: (4/10) | 2.23 s
[Task 5/25] Current/Best: 15.43/ 19.25 GFLOPS | Progress: (8/10) | 3.80 s
[Task 5/25] Current/Best: 3.15/ 19.25 GFLOPS | Progress: (10/10) | 4.65 s Done.
-
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 6/25] Current/Best: 8.61/ 20.20 GFLOPS | Progress: (4/10) | 3.10 s
[Task 6/25] Current/Best: 8.07/ 20.20 GFLOPS | Progress: (8/10) | 5.50 s
[Task 6/25] Current/Best: 12.54/ 20.20 GFLOPS | Progress: (10/10) | 8.05 s Done.
-
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 7/25] Current/Best: 15.71/ 15.71 GFLOPS | Progress: (4/10) | 3.85 s
[Task 7/25] Current/Best: 10.80/ 16.99 GFLOPS | Progress: (8/10) | 5.62 s
[Task 7/25] Current/Best: 21.63/ 21.63 GFLOPS | Progress: (10/10) | 6.71 s Done.
-
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 8/25] Current/Best: 2.98/ 20.32 GFLOPS | Progress: (4/10) | 3.44 s
[Task 8/25] Current/Best: 9.84/ 20.32 GFLOPS | Progress: (8/10) | 6.60 s
[Task 8/25] Current/Best: 17.18/ 20.32 GFLOPS | Progress: (10/10) | 9.10 s Done.
-
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 9/25] Current/Best: 11.43/ 11.43 GFLOPS | Progress: (4/10) | 10.80 s
[Task 9/25] Current/Best: 3.30/ 17.60 GFLOPS | Progress: (8/10) | 22.06 s
[Task 9/25] Current/Best: 15.34/ 17.60 GFLOPS | Progress: (10/10) | 23.77 s
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 10/25] Current/Best: 12.11/ 13.22 GFLOPS | Progress: (4/10) | 3.03 s
[Task 10/25] Current/Best: 17.35/ 17.35 GFLOPS | Progress: (8/10) | 4.97 s
[Task 10/25] Current/Best: 14.19/ 17.35 GFLOPS | Progress: (10/10) | 6.09 s Done.
-
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 11/25] Current/Best: 3.10/ 22.93 GFLOPS | Progress: (4/10) | 3.39 s
[Task 11/25] Current/Best: 14.29/ 22.93 GFLOPS | Progress: (8/10) | 5.96 s
[Task 11/25] Current/Best: 13.36/ 22.93 GFLOPS | Progress: (10/10) | 6.90 s Done.
-
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 12/25] Current/Best: 12.57/ 16.49 GFLOPS | Progress: (4/10) | 2.98 s
[Task 12/25] Current/Best: 5.69/ 16.49 GFLOPS | Progress: (8/10) | 6.78 s
[Task 12/25] Current/Best: 13.88/ 16.49 GFLOPS | Progress: (10/10) | 8.77 s Done.
-
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 13/25] Current/Best: 7.01/ 17.62 GFLOPS | Progress: (4/10) | 4.88 s
[Task 13/25] Current/Best: 6.22/ 21.23 GFLOPS | Progress: (8/10) | 7.19 s
[Task 13/25] Current/Best: 12.46/ 21.23 GFLOPS | Progress: (10/10) | 8.02 s Done.
-
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 14/25] Current/Best: 11.68/ 20.69 GFLOPS | Progress: (4/10) | 4.11 s
[Task 14/25] Current/Best: 11.76/ 20.69 GFLOPS | Progress: (8/10) | 7.40 s
[Task 14/25] Current/Best: 6.40/ 20.69 GFLOPS | Progress: (10/10) | 8.69 s Done.
-
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 15/25] Current/Best: 7.10/ 17.98 GFLOPS | Progress: (4/10) | 2.43 s
[Task 15/25] Current/Best: 16.31/ 17.98 GFLOPS | Progress: (8/10) | 9.02 s
[Task 15/25] Current/Best: 23.72/ 23.72 GFLOPS | Progress: (10/10) | 9.74 s Done.
-
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 16/25] Current/Best: 14.70/ 21.33 GFLOPS | Progress: (4/10) | 2.19 s
[Task 16/25] Current/Best: 16.84/ 21.33 GFLOPS | Progress: (8/10) | 4.98 s
[Task 16/25] Current/Best: 19.36/ 21.33 GFLOPS | Progress: (10/10) | 5.59 s Done.
-
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 17/25] Current/Best: 10.00/ 22.92 GFLOPS | Progress: (4/10) | 2.69 s
[Task 17/25] Current/Best: 13.89/ 22.92 GFLOPS | Progress: (8/10) | 4.78 s
[Task 17/25] Current/Best: 12.34/ 23.36 GFLOPS | Progress: (10/10) | 5.74 s Done.
-
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 18/25] Current/Best: 10.49/ 18.97 GFLOPS | Progress: (4/10) | 4.55 s
[Task 18/25] Current/Best: 14.89/ 18.97 GFLOPS | Progress: (8/10) | 6.96 s
[Task 18/25] Current/Best: 17.07/ 18.97 GFLOPS | Progress: (10/10) | 7.86 s Done.
-
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 19/25] Current/Best: 11.01/ 20.44 GFLOPS | Progress: (4/10) | 4.41 s
[Task 19/25] Current/Best: 19.33/ 20.44 GFLOPS | Progress: (8/10) | 7.64 s
[Task 19/25] Current/Best: 2.70/ 20.44 GFLOPS | Progress: (10/10) | 10.62 s Done.
-
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 20/25] Current/Best: 10.21/ 15.74 GFLOPS | Progress: (4/10) | 3.34 s Done.
-
[Task 20/25] Current/Best: 6.78/ 16.39 GFLOPS | Progress: (8/10) | 5.25 s
[Task 20/25] Current/Best: 18.72/ 18.72 GFLOPS | Progress: (10/10) | 7.96 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 21/25] Current/Best: 15.47/ 19.17 GFLOPS | Progress: (4/10) | 2.81 s
[Task 21/25] Current/Best: 15.62/ 20.00 GFLOPS | Progress: (8/10) | 3.99 s
[Task 21/25] Current/Best: 12.13/ 20.00 GFLOPS | Progress: (10/10) | 4.78 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 22/25] Current/Best: 9.15/ 12.00 GFLOPS | Progress: (4/10) | 3.86 s
[Task 22/25] Current/Best: 11.68/ 14.83 GFLOPS | Progress: (8/10) | 6.25 s
[Task 22/25] Current/Best: 6.19/ 14.83 GFLOPS | Progress: (10/10) | 7.05 s Done.
-
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 23/25] Current/Best: 11.03/ 19.15 GFLOPS | Progress: (4/10) | 4.09 s
[Task 23/25] Current/Best: 15.30/ 19.15 GFLOPS | Progress: (8/10) | 6.09 s
[Task 23/25] Current/Best: 11.14/ 19.15 GFLOPS | Progress: (10/10) | 7.35 s Done.
-
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 24/25] Current/Best: 3.47/ 10.37 GFLOPS | Progress: (4/10) | 10.11 s
[Task 24/25] Current/Best: 9.05/ 10.37 GFLOPS | Progress: (8/10) | 16.49 s
[Task 24/25] Current/Best: 8.78/ 10.37 GFLOPS | Progress: (10/10) | 17.03 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
+
[Task 1/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 1/25] Current/Best: 15.17/ 15.31 GFLOPS | Progress: (4/10) | 6.03 s
[Task 1/25] Current/Best: 5.37/ 15.31 GFLOPS | Progress: (8/10) | 10.98 s
[Task 1/25] Current/Best: 13.56/ 15.31 GFLOPS | Progress: (10/10) | 12.01 s Done.
+
[Task 2/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 2/25] Current/Best: 15.49/ 15.49 GFLOPS | Progress: (4/10) | 2.55 s
[Task 2/25] Current/Best: 16.35/ 16.35 GFLOPS | Progress: (8/10) | 4.21 s
[Task 2/25] Current/Best: 9.12/ 20.08 GFLOPS | Progress: (10/10) | 4.77 s Done.
+
[Task 3/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 3/25] Current/Best: 11.13/ 24.25 GFLOPS | Progress: (4/10) | 2.95 s
[Task 3/25] Current/Best: 10.66/ 24.25 GFLOPS | Progress: (8/10) | 4.71 s
[Task 3/25] Current/Best: 17.74/ 24.25 GFLOPS | Progress: (10/10) | 5.46 s Done.
+
[Task 4/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 4/25] Current/Best: 6.33/ 12.87 GFLOPS | Progress: (4/10) | 5.77 s
[Task 4/25] Current/Best: 6.46/ 20.42 GFLOPS | Progress: (8/10) | 10.79 s
[Task 4/25] Current/Best: 11.02/ 20.42 GFLOPS | Progress: (10/10) | 12.79 s Done.
+
[Task 5/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 5/25] Current/Best: 14.23/ 14.23 GFLOPS | Progress: (4/10) | 2.99 s
[Task 5/25] Current/Best: 19.45/ 19.45 GFLOPS | Progress: (8/10) | 4.24 s
[Task 5/25] Current/Best: 9.78/ 19.45 GFLOPS | Progress: (10/10) | 5.52 s Done.
+
[Task 6/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 6/25] Current/Best: 8.00/ 17.90 GFLOPS | Progress: (4/10) | 3.91 s
[Task 6/25] Current/Best: 14.58/ 18.45 GFLOPS | Progress: (8/10) | 6.15 s
[Task 6/25] Current/Best: 5.25/ 18.45 GFLOPS | Progress: (10/10) | 7.56 s Done.
+
[Task 7/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 7/25] Current/Best: 11.37/ 11.37 GFLOPS | Progress: (4/10) | 3.04 s
[Task 7/25] Current/Best: 6.33/ 21.78 GFLOPS | Progress: (8/10) | 5.16 s
[Task 7/25] Current/Best: 6.34/ 21.78 GFLOPS | Progress: (10/10) | 6.17 s Done.
+
[Task 8/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 8/25] Current/Best: 2.89/ 14.79 GFLOPS | Progress: (4/10) | 23.50 s
[Task 8/25] Current/Best: 3.49/ 14.79 GFLOPS | Progress: (8/10) | 28.71 s
[Task 8/25] Current/Best: 13.60/ 14.79 GFLOPS | Progress: (10/10) | 29.80 s
[Task 9/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 9/25] Current/Best: 16.25/ 16.25 GFLOPS | Progress: (4/10) | 2.80 s
[Task 9/25] Current/Best: 11.89/ 23.08 GFLOPS | Progress: (8/10) | 5.38 s
[Task 9/25] Current/Best: 15.19/ 23.08 GFLOPS | Progress: (10/10) | 6.05 s Done.
+
[Task 10/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 10/25] Current/Best: 15.14/ 15.14 GFLOPS | Progress: (4/10) | 3.09 s
[Task 10/25] Current/Best: 4.21/ 22.70 GFLOPS | Progress: (8/10) | 4.58 s
[Task 10/25] Current/Best: 17.25/ 22.70 GFLOPS | Progress: (10/10) | 5.32 s Done.
+
[Task 11/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 11/25] Current/Best: 20.00/ 20.00 GFLOPS | Progress: (4/10) | 3.07 s
[Task 11/25] Current/Best: 16.31/ 20.00 GFLOPS | Progress: (8/10) | 5.05 s
[Task 11/25] Current/Best: 10.80/ 20.96 GFLOPS | Progress: (10/10) | 6.02 s Done.
+
[Task 12/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 12/25] Current/Best: 10.20/ 18.05 GFLOPS | Progress: (4/10) | 3.81 s
[Task 12/25] Current/Best: 19.13/ 19.13 GFLOPS | Progress: (8/10) | 5.48 s
[Task 12/25] Current/Best: 6.41/ 19.13 GFLOPS | Progress: (10/10) | 6.97 s Done.
+
[Task 13/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 13/25] Current/Best: 12.10/ 18.39 GFLOPS | Progress: (4/10) | 4.04 s
[Task 13/25] Current/Best: 8.81/ 18.39 GFLOPS | Progress: (8/10) | 6.33 s
[Task 13/25] Current/Best: 9.41/ 22.56 GFLOPS | Progress: (10/10) | 7.22 s Done.
+
[Task 14/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 14/25] Current/Best: 13.04/ 18.41 GFLOPS | Progress: (4/10) | 2.68 s
[Task 14/25] Current/Best: 7.03/ 18.41 GFLOPS | Progress: (8/10) | 5.56 s
[Task 14/25] Current/Best: 16.82/ 18.41 GFLOPS | Progress: (10/10) | 6.50 s
[Task 15/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
Done.
+
[Task 15/25] Current/Best: 6.39/ 20.88 GFLOPS | Progress: (4/10) | 2.55 s
[Task 15/25] Current/Best: 10.16/ 23.81 GFLOPS | Progress: (8/10) | 4.13 s
[Task 15/25] Current/Best: 3.25/ 23.81 GFLOPS | Progress: (10/10) | 5.03 s
[Task 16/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 16/25] Current/Best: 17.50/ 17.50 GFLOPS | Progress: (4/10) | 3.93 s
[Task 16/25] Current/Best: 10.09/ 20.84 GFLOPS | Progress: (8/10) | 5.22 s
[Task 16/25] Current/Best: 10.18/ 20.84 GFLOPS | Progress: (10/10) | 6.03 s Done.
+
[Task 17/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 17/25] Current/Best: 13.40/ 20.11 GFLOPS | Progress: (4/10) | 3.20 s
[Task 17/25] Current/Best: 18.22/ 24.35 GFLOPS | Progress: (8/10) | 5.12 s
[Task 17/25] Current/Best: 1.56/ 24.35 GFLOPS | Progress: (10/10) | 9.28 s Done.
+
[Task 18/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 18/25] Current/Best: 6.19/ 15.40 GFLOPS | Progress: (4/10) | 3.20 s
[Task 18/25] Current/Best: 11.47/ 17.33 GFLOPS | Progress: (8/10) | 6.52 s
[Task 18/25] Current/Best: 20.82/ 21.22 GFLOPS | Progress: (10/10) | 7.16 s Done.
+
[Task 19/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 19/25] Current/Best: 10.79/ 19.76 GFLOPS | Progress: (4/10) | 3.96 s
[Task 19/25] Current/Best: 10.96/ 19.76 GFLOPS | Progress: (8/10) | 7.35 s
[Task 19/25] Current/Best: 16.17/ 19.76 GFLOPS | Progress: (10/10) | 8.28 s Done.
+
[Task 20/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 20/25] Current/Best: 13.01/ 18.92 GFLOPS | Progress: (4/10) | 3.00 s
[Task 20/25] Current/Best: 12.96/ 18.92 GFLOPS | Progress: (8/10) | 6.77 s
[Task 20/25] Current/Best: 10.37/ 22.65 GFLOPS | Progress: (10/10) | 8.12 s
[Task 21/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 21/25] Current/Best: 11.45/ 15.34 GFLOPS | Progress: (4/10) | 3.02 s
[Task 21/25] Current/Best: 10.64/ 15.34 GFLOPS | Progress: (8/10) | 4.57 s
[Task 21/25] Current/Best: 13.14/ 15.34 GFLOPS | Progress: (10/10) | 5.19 s
[Task 22/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 22/25] Current/Best: 2.69/ 21.18 GFLOPS | Progress: (4/10) | 3.05 s
[Task 22/25] Current/Best: 5.42/ 21.18 GFLOPS | Progress: (8/10) | 4.68 s
[Task 22/25] Current/Best: 9.54/ 21.18 GFLOPS | Progress: (10/10) | 7.33
s Done.
+
[Task 23/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 23/25] Current/Best: 19.47/ 19.47 GFLOPS | Progress: (4/10) | 2.73 s
[Task 23/25] Current/Best: 17.68/ 19.47 GFLOPS | Progress: (8/10) | 6.52 s
[Task 23/25] Current/Best: 13.35/ 19.47 GFLOPS | Progress: (10/10) | 7.56 s Done.
+
[Task 24/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s
[Task 24/25] Current/Best: 1.99/ 4.26 GFLOPS | Progress: (4/10) | 53.95 s
[Task 24/25] Current/Best: 1.26/ 4.26 GFLOPS | Progress: (8/10) | 66.62 s
[Task 24/25] Current/Best: 3.42/ 4.26 GFLOPS | Progress: (10/10) | 71.60 s
[Task 25/25] Current/Best: 0.00/ 0.00 GFLOPS | Progress: (0/10) | 0.00 s Done.
Done.
-
[Task 25/25] Current/Best: 1.56/ 9.26 GFLOPS | Progress: (4/10) | 6.22 s
[Task 25/25] Current/Best: 7.90/ 9.26 GFLOPS | Progress: (8/10) | 34.74 s
[Task 25/25] Current/Best: 9.22/ 9.26 GFLOPS | Progress: (10/10) | 325.12 s
+ Done.
+ Done.
+
[Task 25/25] Current/Best: 5.86/ 9.24 GFLOPS | Progress: (4/10) | 19.36 s
[Task 25/25] Current/Best: 9.03/ 9.24 GFLOPS | Progress: (8/10) | 36.21 s
[Task 25/25] Current/Best: 7.87/ 9.24 GFLOPS | Progress: (10/10) | 37.14 s
The output from this tuning process will look something like this:
@@ -564,6 +564,14 @@ model using optimized operators to speed up our computations.
+.. rst-class:: sphx-glr-script-out
+
+ Out:
+
+ .. code-block:: none
+
+ Done.
+
Verify that the optimized model runs and produces the same results:
@@ -594,8 +602,8 @@ Verify that the optimized model runs and produces the same results:
.. code-block:: none
- class='n02123045 tabby, tabby cat' with probability=0.621102
- class='n02123159 tiger cat' with probability=0.356379
+ class='n02123045 tabby, tabby cat' with probability=0.621104
+ class='n02123159 tiger cat' with probability=0.356377
class='n02124075 Egyptian cat' with probability=0.019712
class='n02129604 tiger, Panthera tigris' with probability=0.001215
class='n04040759 radiator' with probability=0.000262
@@ -648,8 +656,8 @@ improvement in comparing the optimized model to the unoptimized model.
.. code-block:: none
- optimized: {'mean': 443.6937599999874, 'median': 443.03481865008507, 'std': 1.9977954528235273}
- unoptimized: {'mean': 494.8987510900406, 'median': 494.9888973500492, 'std': 0.4855438254512492}
+ optimized: {'mean': 426.8101321900008, 'median': 426.68636005000735, 'std': 1.0729700128891184}
+ unoptimized: {'mean': 497.7622343500014, 'median': 497.80023144999745, 'std': 0.6134268998812399}
@@ -669,7 +677,7 @@ profiling/benchmarking.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 11 minutes 59.050 seconds)
+ **Total running time of the script:** ( 8 minutes 6.836 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 af4d4a424..627766b2c 100644
--- a/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
+++ b/docs/_sources/tutorial/cross_compilation_and_rpc.rst.txt
@@ -235,7 +235,7 @@ device and returns the measured cost. Network overhead is excluded.
.. code-block:: none
- 1.245e-07 secs/op
+ 1.257e-07 secs/op
diff --git a/docs/_sources/tutorial/intro_topi.rst.txt b/docs/_sources/tutorial/intro_topi.rst.txt
index 9c0b6f2c5..4e8b4255a 100644
--- a/docs/_sources/tutorial/intro_topi.rst.txt
+++ b/docs/_sources/tutorial/intro_topi.rst.txt
@@ -233,7 +233,7 @@ As you can see, scheduled stages of computation have been accumulated and we can
.. code-block:: none
- [stage(a, placeholder(a, 0x20db2320)), stage(b, placeholder(b, 0x21f2df70)), stage(T_add, compute(T_add, body=[(a[ax0, ax1, ax2] + b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(min=0, ext=10))], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[(a[ax0, ax1, ax2]*b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(mi [...]
+ [stage(a, placeholder(a, 0x21e763c0)), stage(b, placeholder(b, 0x2260b2f0)), stage(T_add, compute(T_add, body=[(a[ax0, ax1, ax2] + b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(min=0, ext=10))], reduce_axis=[], tag=broadcast, attrs={})), stage(T_multiply, compute(T_multiply, body=[(a[ax0, ax1, ax2]*b[ax1, ax2])], axis=[iter_var(ax0, range(min=0, ext=100)), iter_var(ax1, range(min=0, ext=10)), iter_var(ax2, range(mi [...]
diff --git a/docs/_sources/tutorial/sg_execution_times.rst.txt b/docs/_sources/tutorial/sg_execution_times.rst.txt
index 8df21b9b5..73093ceb8 100644
--- a/docs/_sources/tutorial/sg_execution_times.rst.txt
+++ b/docs/_sources/tutorial/sg_execution_times.rst.txt
@@ -5,17 +5,17 @@
Computation times
=================
-**14:37.976** total execution time for **tutorial** files:
+**10:50.960** total execution time for **tutorial** files:
-- **11:59.050**: :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)
-- **01:00.120**: :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)
-- **00:48.278**: :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``)
-- **00:26.048**: :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)
-- **00:22.854**: :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)
-- **00:00.716**: :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)
-- **00:00.561**: :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)
-- **00:00.214**: :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``)
-- **00:00.039**: :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)
-- **00:00.034**: :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)
-- **00:00.032**: :ref:`sphx_glr_tutorial_install.py` (``install.py``)
-- **00:00.032**: :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)
+- **08:06.836**: :ref:`sphx_glr_tutorial_autotvm_relay_x86.py` (``autotvm_relay_x86.py``)
+- **01:00.139**: :ref:`sphx_glr_tutorial_tensor_expr_get_started.py` (``tensor_expr_get_started.py``)
+- **00:58.950**: :ref:`sphx_glr_tutorial_auto_scheduler_matmul_x86.py` (``auto_scheduler_matmul_x86.py``)
+- **00:26.377**: :ref:`sphx_glr_tutorial_relay_quick_start.py` (``relay_quick_start.py``)
+- **00:16.327**: :ref:`sphx_glr_tutorial_autotvm_matmul_x86.py` (``autotvm_matmul_x86.py``)
+- **00:01.168**: :ref:`sphx_glr_tutorial_tensor_ir_blitz_course.py` (``tensor_ir_blitz_course.py``)
+- **00:00.727**: :ref:`sphx_glr_tutorial_intro_topi.py` (``intro_topi.py``)
+- **00:00.226**: :ref:`sphx_glr_tutorial_cross_compilation_and_rpc.py` (``cross_compilation_and_rpc.py``)
+- **00:00.054**: :ref:`sphx_glr_tutorial_tvmc_python.py` (``tvmc_python.py``)
+- **00:00.053**: :ref:`sphx_glr_tutorial_introduction.py` (``introduction.py``)
+- **00:00.052**: :ref:`sphx_glr_tutorial_install.py` (``install.py``)
+- **00:00.049**: :ref:`sphx_glr_tutorial_tvmc_command_line_driver.py` (``tvmc_command_line_driver.py``)
diff --git a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
index d058f4f17..3fe6f8424 100644
--- a/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
+++ b/docs/_sources/tutorial/tensor_expr_get_started.rst.txt
@@ -243,7 +243,7 @@ helper function to run a profile of the TVM generated code.
.. code-block:: none
- Numpy running time: 0.000008
+ Numpy running time: 0.000007
naive: 0.000006
@@ -438,10 +438,10 @@ We can now compare the different schedules
.. code-block:: none
Operator Timing Performance
- numpy 7.875289993535262e-06 1.0
- naive 5.8737999999999996e-06 0.7458518993994808
- parallel 6.062800000000001e-06 0.7698510156422032
- vector 2.45568e-05 3.118208982800431
+ numpy 7.109870000476803e-06 1.0
+ naive 6.0174999999999996e-06 0.8463586534769909
+ parallel 6.0704e-06 0.853799014552011
+ vector 2.47547e-05 3.481737359240028
@@ -830,7 +830,7 @@ matrix multiplication.
.. code-block:: none
- Numpy running time: 0.018999
+ Numpy running time: 0.019535
@@ -886,7 +886,7 @@ optimizations.
.. code-block:: none
- none: 3.340335
+ none: 3.299393
@@ -985,7 +985,7 @@ schedule.
.. code-block:: none
- blocking: 0.306033
+ blocking: 0.327687
@@ -1077,7 +1077,7 @@ already cache friendly from our previous optimizations.
.. code-block:: none
- vectorization: 0.341253
+ vectorization: 0.350142
@main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1149,7 +1149,7 @@ more cache friendly.
.. code-block:: none
- loop permutation: 0.114305
+ loop permutation: 0.120235
@main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1246,7 +1246,7 @@ optimized schedule.
.. code-block:: none
- array packing: 0.108074
+ array packing: 0.109858
@main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1337,7 +1337,7 @@ to `C` when all the block results are ready.
.. code-block:: none
- block caching: 0.110396
+ block caching: 0.110653
@main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1421,7 +1421,7 @@ of thread-level parallelization.
.. code-block:: none
- parallelization: 0.143809
+ parallelization: 0.142940
@main = primfn(A_1: handle, B_1: handle, C_1: handle) -> ()
attr = {"from_legacy_te_schedule": True, "global_symbol": "main", "tir.noalias": True}
buffers = {A: Buffer(A_2: Pointer(float32), float32, [1048576], []),
@@ -1500,13 +1500,13 @@ working, we can compare the results.
.. code-block:: none
Operator Timing Performance
- none 3.3403347894 1.0
- blocking 0.30603281769999996 0.09161740873134767
- vectorization 0.3412531421 0.10216135914966082
- loop permutation 0.11430457890000001 0.03421949777690748
- array packing 0.10807382680000002 0.03235419010781627
- block caching 0.11039616320000001 0.03304943071883812
- parallelization 0.1438085398 0.043052133653295056
+ none 3.2993928527999996 1.0
+ blocking 0.3276873224 0.09931746142988433
+ vectorization 0.3501415423 0.1061230226048576
+ loop permutation 0.120235096 0.03644158224382513
+ array packing 0.10985756399999999 0.03329629689497884
+ block caching 0.1106534393 0.033537515608696
+ parallelization 0.1429396157 0.043323005800505264
@@ -1543,7 +1543,7 @@ the computation for specific platforms.
.. rst-class:: sphx-glr-timing
- **Total running time of the script:** ( 1 minutes 0.120 seconds)
+ **Total running time of the script:** ( 1 minutes 0.139 seconds)
.. _sphx_glr_download_tutorial_tensor_expr_get_started.py:
diff --git a/docs/commit_hash b/docs/commit_hash
index 91a4fb4ee..cd2feeb74 100644
--- a/docs/commit_hash
+++ b/docs/commit_hash
@@ -1 +1 @@
-ba4cc6c1f26d64c37b0d3d58d18c670895fed2cb
+876e2532278052d27c55adf982c9848ef3a76d4b
diff --git a/docs/genindex.html b/docs/genindex.html
index a02e5ab7d..b02cf6f92 100644
--- a/docs/genindex.html
+++ b/docs/genindex.html
@@ -3145,6 +3145,8 @@
<li><a href="reference/api/python/ir.html#tvm.instrument.PassTimingInstrument.render">(tvm.instrument.PassTimingInstrument static method)</a>
</li>
</ul></li>
+ <li><a href="reference/api/python/tir.html#tvm.tir.stmt_functor.renew_defs">renew_defs() (in module tvm.tir.stmt_functor)</a>
+</li>
<li><a href="reference/api/python/tir.html#tvm.tir.transform.RenormalizeSplitPattern">RenormalizeSplitPattern() (in module tvm.tir.transform)</a>
</li>
<li><a href="reference/api/python/te.html#tvm.te.Stage.reorder">reorder() (tvm.te.Stage method)</a>
diff --git a/docs/how_to/compile_models/from_mxnet.html b/docs/how_to/compile_models/from_mxnet.html
index 87b2598ea..48ab126c5 100644
--- a/docs/how_to/compile_models/from_mxnet.html
+++ b/docs/how_to/compile_models/from_mxnet.html
@@ -400,7 +400,7 @@
</div>
<img alt="../../_images/sphx_glr_from_mxnet_001.png" class="sphx-glr-single-img" src="../../_images/sphx_glr_from_mxnet_001.png" />
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zipf4aa77d4-0448-4ae1-b003-b1bad7ca0c21 from https://apache-mxnet.s3-accelerate.dualstack.amazonaws.com/gluon/models/resnet18_v1-a0666292.zip...
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/resnet18_v1-a0666292.zip1295d9df-221c-49dc-ad75-1f57fdf19cd0 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_paddle.html b/docs/how_to/compile_models/from_paddle.html
index 29a3eebab..2ce687029 100644
--- a/docs/how_to/compile_models/from_paddle.html
+++ b/docs/how_to/compile_models/from_paddle.html
@@ -463,7 +463,7 @@ A quick solution is</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>TVM prediction top-1 id: 282, class name: 282: 'tiger cat',
</pre></div>
</div>
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 10.141 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 5.850 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-paddle-py">
<div class="sphx-glr-download docutils container">
<p><a class="reference download internal" download="" href="../../_downloads/16269b77359771348d507395692524cf/from_paddle.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_paddle.py</span></code></a></p>
diff --git a/docs/how_to/compile_models/from_pytorch.html b/docs/how_to/compile_models/from_pytorch.html
index 39b1e1ab3..467292217 100644
--- a/docs/how_to/compile_models/from_pytorch.html
+++ b/docs/how_to/compile_models/from_pytorch.html
@@ -386,9 +386,9 @@ be unstable.</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>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]
- 43%|####2 | 19.0M/44.7M [00:00<00:00, 200MB/s]
- 96%|#########6| 43.0M/44.7M [00:00<00:00, 230MB/s]
-100%|##########| 44.7M/44.7M [00:00<00:00, 227MB/s]
+ 39%|###9 | 17.4M/44.7M [00:00<00:00, 183MB/s]
+ 93%|#########3| 41.6M/44.7M [00:00<00:00, 224MB/s]
+100%|##########| 44.7M/44.7M [00:00<00:00, 220MB/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 655abf2fb..f2efa97ad 100644
--- a/docs/how_to/compile_models/from_tensorflow.html
+++ b/docs/how_to/compile_models/from_tensorflow.html
@@ -606,7 +606,7 @@ 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 4.535 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 0.778 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-compile-models-from-tensorflow-py">
<div class="sphx-glr-download 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 4c7bb6188..a5e56332d 100644
--- a/docs/how_to/compile_models/sg_execution_times.html
+++ b/docs/how_to/compile_models/sg_execution_times.html
@@ -300,17 +300,17 @@
<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>04:52.614</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
+<p><strong>04:50.120</strong> total execution time for <strong>how_to_compile_models</strong> files:</p>
<ul class="simple">
-<li><p><strong>01:10.141</strong>: <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></li>
-<li><p><strong>01:04.535</strong>: <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></li>
-<li><p><strong>00:55.666</strong>: <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></li>
-<li><p><strong>00:25.857</strong>: <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></li>
-<li><p><strong>00:21.078</strong>: <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></li>
-<li><p><strong>00:20.828</strong>: <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></li>
-<li><p><strong>00:18.698</strong>: <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></li>
-<li><p><strong>00:13.337</strong>: <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></li>
-<li><p><strong>00:02.474</strong>: <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></li>
+<li><p><strong>01:05.850</strong>: <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></li>
+<li><p><strong>01:00.778</strong>: <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></li>
+<li><p><strong>00:58.534</strong>: <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></li>
+<li><p><strong>00:25.625</strong>: <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></li>
+<li><p><strong>00:22.643</strong>: <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></li>
+<li><p><strong>00:21.349</strong>: <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></li>
+<li><p><strong>00:18.848</strong>: <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></li>
+<li><p><strong>00:14.094</strong>: <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></li>
+<li><p><strong>00:02.399</strong>: <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></li>
</ul>
</div>
diff --git a/docs/how_to/deploy_models/deploy_model_on_android.html b/docs/how_to/deploy_models/deploy_model_on_android.html
index 456f09eba..2c3ca3ec1 100644
--- a/docs/how_to/deploy_models/deploy_model_on_android.html
+++ b/docs/how_to/deploy_models/deploy_model_on_android.html
@@ -622,7 +622,7 @@ to the remote android device.</p>
Evaluate inference time cost...
Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 15.9295 15.9324 16.0288 15.8394 0.0625
+ 16.2107 16.1555 16.6169 16.0543 0.1617
</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 289859d9f..3528e2db4 100644
--- a/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
+++ b/docs/how_to/deploy_models/deploy_object_detection_pytorch.html
@@ -409,15 +409,14 @@ be unstable.</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>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]
- 9%|8 | 14.8M/170M [00:00<00:01, 155MB/s]
- 21%|## | 34.9M/170M [00:00<00:00, 188MB/s]
- 32%|###2 | 55.0M/170M [00:00<00:00, 198MB/s]
- 45%|####5 | 76.7M/170M [00:00<00:00, 210MB/s]
- 58%|#####7 | 98.4M/170M [00:00<00:00, 216MB/s]
- 70%|####### | 120M/170M [00:00<00:00, 218MB/s]
- 83%|########2 | 140M/170M [00:00<00:00, 215MB/s]
- 95%|#########5| 162M/170M [00:00<00:00, 220MB/s]
-100%|##########| 170M/170M [00:00<00:00, 213MB/s]
+ 9%|8 | 15.1M/170M [00:00<00:01, 158MB/s]
+ 22%|##1 | 37.1M/170M [00:00<00:00, 201MB/s]
+ 36%|###6 | 61.7M/170M [00:00<00:00, 227MB/s]
+ 50%|##### | 85.1M/170M [00:00<00:00, 234MB/s]
+ 64%|######3 | 109M/170M [00:00<00:00, 239MB/s]
+ 78%|#######7 | 132M/170M [00:00<00:00, 242MB/s]
+ 92%|#########2| 156M/170M [00:00<00:00, 245MB/s]
+100%|##########| 170M/170M [00:00<00:00, 236MB/s]
/usr/local/lib/python3.7/dist-packages/torch/nn/functional.py:3878: 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)
/usr/local/lib/python3.7/dist-packages/torchvision/models/detection/anchor_utils.py:127: 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').
@@ -510,7 +509,7 @@ 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 2.115 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 3 minutes 9.451 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-object-detection-pytorch-py">
<div class="sphx-glr-download 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 f1d9303c4..e7a36a818 100644
--- a/docs/how_to/deploy_models/deploy_prequantized.html
+++ b/docs/how_to/deploy_models/deploy_prequantized.html
@@ -450,8 +450,7 @@ training. Other models require a full post training calibration.</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading: "https://download.pytorch.org/models/mobilenet_v2-b0353104.pth" to /workspace/.cache/torch/hub/checkpoints/mobilenet_v2-b0353104.pth
0%| | 0.00/13.6M [00:00<?, ?B/s]
-100%|#########9| 13.5M/13.6M [00:00<00:00, 142MB/s]
-100%|##########| 13.6M/13.6M [00:00<00:00, 141MB/s]
+100%|##########| 13.6M/13.6M [00:00<00:00, 170MB/s]
</pre></div>
</div>
</div>
@@ -540,7 +539,7 @@ output values are identical out of 1000 outputs from mobilenet v2.</p>
<p class="sphx-glr-script-out">Out:</p>
<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)
- 90.2516 90.1702 92.3845 89.9664 0.3181
+ 90.2704 90.2249 91.4279 90.0486 0.2188
</pre></div>
</div>
<div class="admonition note">
@@ -579,7 +578,7 @@ 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 4.674 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 6.788 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-py">
<div class="sphx-glr-download 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 fce92e489..ef855304e 100644
--- a/docs/how_to/deploy_models/deploy_prequantized_tflite.html
+++ b/docs/how_to/deploy_models/deploy_prequantized_tflite.html
@@ -540,7 +540,7 @@ TFLite Top-5 labels: [387 102 386 341 349]
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time summary:
mean (ms) median (ms) max (ms) min (ms) std (ms)
- 119.8836 119.8378 121.7985 118.9666 0.4667
+ 120.2926 120.3084 123.3358 119.3317 0.5461
</pre></div>
</div>
<div class="admonition note">
@@ -568,7 +568,7 @@ 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> ( 1 minutes 59.418 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 52.122 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-prequantized-tflite-py">
<div class="sphx-glr-download 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 f2c225dfc..3c1ec5c2a 100644
--- a/docs/how_to/deploy_models/deploy_quantized.html
+++ b/docs/how_to/deploy_models/deploy_quantized.html
@@ -480,7 +480,7 @@ 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 13.044 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 13.302 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-quantized-py">
<div class="sphx-glr-download 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 e32c0a31b..5b0e41260 100644
--- a/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
+++ b/docs/how_to/deploy_models/deploy_ssd_gluoncv.html
@@ -415,23 +415,22 @@ 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]
- 5%|5 | 6737/132723 [00:00<00:01, 67344.21KB/s]
- 12%|#1 | 15461/132723 [00:00<00:01, 79039.28KB/s]
- 18%|#8 | 24208/132723 [00:00<00:01, 82883.95KB/s]
- 25%|##4 | 32905/132723 [00:00<00:01, 84491.20KB/s]
- 31%|###1 | 41689/132723 [00:00<00:01, 85694.27KB/s]
- 38%|###7 | 50259/132723 [00:00<00:01, 79912.01KB/s]
- 44%|####3 | 58390/132723 [00:00<00:00, 80343.15KB/s]
- 50%|##### | 66472/132723 [00:00<00:01, 63064.99KB/s]
- 57%|#####6 | 75197/132723 [00:01<00:00, 69263.39KB/s]
- 62%|######2 | 82644/132723 [00:01<00:00, 56738.18KB/s]
- 69%|######8 | 91366/132723 [00:01<00:00, 63906.78KB/s]
- 75%|#######4 | 98990/132723 [00:01<00:00, 67006.84KB/s]
- 80%|######## | 106749/132723 [00:01<00:00, 69795.59KB/s]
- 86%|########6 | 114680/132723 [00:01<00:00, 64651.96KB/s]
- 93%|#########2| 122941/132723 [00:01<00:00, 69187.98KB/s]
- 99%|#########9| 131712/132723 [00:01<00:00, 74183.60KB/s]
-100%|##########| 132723/132723 [00:01<00:00, 71327.11KB/s]
+ 5%|4 | 6575/132723 [00:00<00:01, 65745.99KB/s]
+ 11%|#1 | 15153/132723 [00:00<00:01, 77527.40KB/s]
+ 18%|#7 | 23675/132723 [00:00<00:01, 81036.04KB/s]
+ 24%|##4 | 32190/132723 [00:00<00:01, 82652.12KB/s]
+ 31%|### | 40693/132723 [00:00<00:01, 83504.31KB/s]
+ 37%|###6 | 49044/132723 [00:00<00:01, 77717.03KB/s]
+ 43%|####2 | 56885/132723 [00:00<00:01, 69497.90KB/s]
+ 49%|####9 | 65433/132723 [00:00<00:00, 74029.85KB/s]
+ 56%|#####5 | 73983/132723 [00:00<00:00, 77337.45KB/s]
+ 62%|######2 | 82669/132723 [00:01<00:00, 80115.62KB/s]
+ 69%|######8 | 91303/132723 [00:01<00:00, 81945.29KB/s]
+ 75%|#######5 | 99933/132723 [00:01<00:00, 83232.82KB/s]
+ 82%|########1 | 108509/132723 [00:01<00:00, 83980.08KB/s]
+ 88%|########8 | 117150/132723 [00:01<00:00, 84701.27KB/s]
+ 95%|#########4| 125785/132723 [00:01<00:00, 85192.01KB/s]
+100%|##########| 132723/132723 [00:01<00:00, 80813.26KB/s]
</pre></div>
</div>
<p>Create TVM runtime and do inference
@@ -471,7 +470,7 @@ Downloading /workspace/.mxnet/models/ssd_512_resnet50_v1_voc-9c8b225a.zip from h
</pre></div>
</div>
<img alt="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" class="sphx-glr-single-img" src="../../_images/sphx_glr_deploy_ssd_gluoncv_001.png" />
-<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 23.362 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 27.384 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-deploy-models-deploy-ssd-gluoncv-py">
<div class="sphx-glr-download 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 fd9ee341b..315aeb5b3 100644
--- a/docs/how_to/deploy_models/sg_execution_times.html
+++ b/docs/how_to/deploy_models/sg_execution_times.html
@@ -300,16 +300,16 @@
<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>10:32.073</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
+<p><strong>10:40.220</strong> total execution time for <strong>how_to_deploy_models</strong> files:</p>
<ul class="simple">
-<li><p><strong>03:02.115</strong>: <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></li>
-<li><p><strong>02:23.362</strong>: <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></li>
-<li><p><strong>01:59.418</strong>: <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></li>
-<li><p><strong>01:13.044</strong>: <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></li>
-<li><p><strong>01:04.674</strong>: <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></li>
-<li><p><strong>00:27.290</strong>: <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></li>
-<li><p><strong>00:21.977</strong>: <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></li>
-<li><p><strong>00:00.194</strong>: <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></li>
+<li><p><strong>03:09.451</strong>: <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></li>
+<li><p><strong>02:27.384</strong>: <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></li>
+<li><p><strong>01:52.122</strong>: <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></li>
+<li><p><strong>01:13.302</strong>: <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></li>
+<li><p><strong>01:06.788</strong>: <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></li>
+<li><p><strong>00:29.123</strong>: <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></li>
+<li><p><strong>00:21.846</strong>: <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></li>
+<li><p><strong>00:00.203</strong>: <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></li>
</ul>
</div>
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 0390edf91..7f12fa118 100644
--- a/docs/how_to/extend_tvm/bring_your_own_datatypes.html
+++ b/docs/how_to/extend_tvm/bring_your_own_datatypes.html
@@ -588,7 +588,7 @@ In this alpha state of the Bring Your Own Datatypes framework, we have not imple
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Downloading /workspace/.mxnet/models/mobilenet0.25-9f83e440.zip37fe282f-da91-43a1-a990-86309c5d85cb 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.zip3a5965b5-17fb-4b77-9335-e415a17b7a43 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 dbb2d2d31..fe3f538ce 100644
--- a/docs/how_to/extend_tvm/sg_execution_times.html
+++ b/docs/how_to/extend_tvm/sg_execution_times.html
@@ -300,12 +300,12 @@
<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>00:38.319</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
+<p><strong>00:39.355</strong> total execution time for <strong>how_to_extend_tvm</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:34.801</strong>: <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></li>
-<li><p><strong>00:02.256</strong>: <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></li>
-<li><p><strong>00:01.061</strong>: <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></li>
-<li><p><strong>00:00.201</strong>: <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></li>
+<li><p><strong>00:35.731</strong>: <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></li>
+<li><p><strong>00:02.326</strong>: <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></li>
+<li><p><strong>00:01.093</strong>: <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></li>
+<li><p><strong>00:00.205</strong>: <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></li>
</ul>
</div>
diff --git a/docs/how_to/extend_tvm/use_pass_instrument.html b/docs/how_to/extend_tvm/use_pass_instrument.html
index 5e28b7079..c45375ce8 100644
--- a/docs/how_to/extend_tvm/use_pass_instrument.html
+++ b/docs/how_to/extend_tvm/use_pass_instrument.html
@@ -486,10 +486,10 @@ profile the execution time of each passes.</p>
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 6001us [6001us] (44.02%; 44.02%)
-FoldScaleAxis: 7630us [2us] (55.98%; 55.98%)
- FoldConstant: 7628us [1470us] (55.96%; 99.97%)
- InferType: 6157us [6157us] (45.17%; 80.72%)
+InferType: 6220us [6220us] (46.06%; 46.06%)
+FoldScaleAxis: 7284us [2us] (53.94%; 53.94%)
+ FoldConstant: 7282us [1502us] (53.92%; 99.97%)
+ InferType: 5780us [5780us] (42.80%; 79.38%)
</pre></div>
</div>
</div>
@@ -512,10 +512,10 @@ Refer to following sections and <a class="reference internal" href="../../refere
</div>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Printing results of timing profile...
-InferType: 5829us [5829us] (44.86%; 44.86%)
-FoldScaleAxis: 7165us [2us] (55.14%; 55.14%)
- FoldConstant: 7163us [1503us] (55.13%; 99.98%)
- InferType: 5660us [5660us] (43.56%; 79.02%)
+InferType: 5894us [5894us] (44.19%; 44.19%)
+FoldScaleAxis: 7445us [2us] (55.81%; 55.81%)
+ FoldConstant: 7443us [1533us] (55.80%; 99.97%)
+ InferType: 5910us [5910us] (44.30%; 79.40%)
</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 db46cdf7e..5d2b19f57 100644
--- a/docs/how_to/optimize_operators/opt_conv_cuda.html
+++ b/docs/how_to/optimize_operators/opt_conv_cuda.html
@@ -534,7 +534,7 @@ latency of convolution.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.107188 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Convolution: 54.191826 ms
</pre></div>
</div>
<div class="sphx-glr-footer class 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 b5d1be63f..d769c4783 100644
--- a/docs/how_to/optimize_operators/opt_conv_tensorcore.html
+++ b/docs/how_to/optimize_operators/opt_conv_tensorcore.html
@@ -878,7 +878,7 @@ be able to run on our build server</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 6.470923 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>conv2d with tensor core: 6.846378 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 d05ccad59..6ead84e8c 100644
--- a/docs/how_to/optimize_operators/opt_gemm.html
+++ b/docs/how_to/optimize_operators/opt_gemm.html
@@ -431,8 +431,8 @@ Then we write a baseline implementation, the simplest way to write a matrix mult
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019322
-Baseline: 3.339645
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Numpy running time: 0.019626
+Baseline: 3.279458
</pre></div>
</div>
<p>In TVM, we can always inspect lower level IR to debug or optimize our schedule.
@@ -494,7 +494,7 @@ fill 32 * 32 * sizeof(float) which is 4KB in the cache whose total size is 32KB
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.305210
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt1: 0.315032
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -563,7 +563,7 @@ vastly.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.339233
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt2: 0.342817
</pre></div>
</div>
<p>Here is the generated IR after vectorization.</p>
@@ -626,7 +626,7 @@ the access pattern for A matrix is more cache friendly.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.115594
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt3: 0.118973
</pre></div>
</div>
<p>Here is the generated IR after loop permutation.</p>
@@ -711,7 +711,7 @@ flattening.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.112289
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt4: 0.110845
</pre></div>
</div>
<p>Here is the generated IR after array packing.</p>
@@ -799,7 +799,7 @@ write to C when all the block results are ready.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.111148
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt5: 0.111547
</pre></div>
</div>
<p>Here is the generated IR after blocking.</p>
@@ -891,7 +891,7 @@ write to C when all the block results are ready.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.144759
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Opt6: 0.145354
</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 d43728061..74cee6a72 100644
--- a/docs/how_to/optimize_operators/sg_execution_times.html
+++ b/docs/how_to/optimize_operators/sg_execution_times.html
@@ -300,11 +300,11 @@
<div class="section" id="computation-times">
<span id="sphx-glr-how-to-optimize-operators-sg-execution-times"></span><h1>Computation times<a class="headerlink" href="#computation-times" title="Permalink to this headline">¶</a></h1>
-<p><strong>00:34.778</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
+<p><strong>00:35.098</strong> total execution time for <strong>how_to_optimize_operators</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:32.321</strong>: <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></li>
-<li><p><strong>00:01.308</strong>: <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></li>
-<li><p><strong>00:01.150</strong>: <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></li>
+<li><p><strong>00:32.437</strong>: <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></li>
+<li><p><strong>00:01.406</strong>: <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></li>
+<li><p><strong>00:01.255</strong>: <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></li>
</ul>
</div>
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 3ed1618ea..e46a9ccfb 100644
--- a/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
+++ b/docs/how_to/tune_with_autoscheduler/sg_execution_times.html
@@ -300,14 +300,14 @@
<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>05:02.810</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
+<p><strong>04:57.684</strong> total execution time for <strong>how_to_tune_with_autoscheduler</strong> files:</p>
<ul class="simple">
-<li><p><strong>02:17.757</strong>: <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></li>
-<li><p><strong>01:20.507</strong>: <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></li>
-<li><p><strong>00:40.706</strong>: <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></li>
-<li><p><strong>00:26.654</strong>: <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></li>
-<li><p><strong>00:08.615</strong>: <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></li>
-<li><p><strong>00:08.571</strong>: <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></li>
+<li><p><strong>02:20.501</strong>: <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></li>
+<li><p><strong>01:21.450</strong>: <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></li>
+<li><p><strong>00:41.139</strong>: <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></li>
+<li><p><strong>00:16.500</strong>: <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></li>
+<li><p><strong>00:09.239</strong>: <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></li>
+<li><p><strong>00:08.855</strong>: <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></li>
</ul>
</div>
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 5f60bd29e..42c984891 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
@@ -470,484 +470,101 @@ cooperative fetching, unrolling and operator fusion.</p>
compute: Buffer(compute_2: Pointer(float32), float32, [25088], [])}
buffer_map = {data_1: data, kernel_1: kernel, bias_1: bias, compute_1: compute}
preflattened_buffer_map = {data_1: data_3: Buffer(data_2, float32, [1, 512, 7, 7], []), kernel_1: kernel_3: Buffer(kernel_2, float32, [512, 512, 3, 3], []), bias_1: bias_3: Buffer(bias_2, float32, [1, 512, 1, 1], []), compute_1: compute_3: Buffer(compute_2, float32, [1, 512, 7, 7], [])} {
- attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 28;
- allocate(conv2d_nchw: Pointer(local float32), float32, [14]), storage_scope = local;
- allocate(pad_temp.shared: Pointer(shared float32), float32, [72]), storage_scope = shared;
- allocate(kernel.shared: Pointer(shared float32), float32, [3072]), storage_scope = shared;
- attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64 {
- conv2d_nchw_1: Buffer(conv2d_nchw, float32, [14], [], scope="local", align=32)[0] = 0f32
+ attr [IterVar(blockIdx.x: int32, (nullptr), "ThreadIndex", "blockIdx.x")] "thread_extent" = 112;
+ allocate(conv2d_nchw: Pointer(local float32), float32, [7]), storage_scope = local;
+ allocate(pad_temp.shared: Pointer(shared float32), float32, [432]), storage_scope = shared;
+ allocate(kernel.shared: Pointer(shared float32), float32, [4608]), storage_scope = shared;
+ attr [IterVar(threadIdx.x: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32 {
+ conv2d_nchw_1: Buffer(conv2d_nchw, float32, [1], [], scope="local", align=4)[0] = 0f32
conv2d_nchw_1[1] = 0f32
conv2d_nchw_1[2] = 0f32
conv2d_nchw_1[3] = 0f32
conv2d_nchw_1[4] = 0f32
conv2d_nchw_1[5] = 0f32
conv2d_nchw_1[6] = 0f32
- conv2d_nchw_1[7] = 0f32
- conv2d_nchw_1[8] = 0f32
- conv2d_nchw_1[9] = 0f32
- conv2d_nchw_1[10] = 0f32
- conv2d_nchw_1[11] = 0f32
- conv2d_nchw_1[12] = 0f32
- conv2d_nchw_1[13] = 0f32
- for (rc.outer.outer: int32, 0, 64) {
- for (ry.outer.outer: int32, 0, 3) {
- let cse_var_2: int32 = (rc.outer.outer*72)
- let cse_var_1: int32 = (ry.outer.outer*3)
- {
- attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64 {
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1: Buffer(pad_temp.shared, float32, [72], [], scope="shared")[(threadIdx.x_1*4)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1*4), 9))) && (floormod((threadIdx.x_1*4), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv((threadIdx.x_1*4), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + [...]
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 1)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 1), 9))) && (floormod(((threadIdx.x_1*4) + 1), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 1), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 1), 9)) - 8)], 0 [...]
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 2)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 2), 9))) && (floormod(((threadIdx.x_1*4) + 2), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 2), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 2), 9)) - 8)], 0 [...]
- }
- if @tir.likely((threadIdx.x_1 < 18), dtype=bool) {
- pad_temp.shared_1[((threadIdx.x_1*4) + 3)] = @tir.if_then_else(((((1 <= (ry.outer.outer + floormod(blockIdx.x, 7))) && ((ry.outer.outer + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(((threadIdx.x_1*4) + 3), 9))) && (floormod(((threadIdx.x_1*4) + 3), 9) < 8)), data[((((((rc.outer.outer*392) + (floordiv(((threadIdx.x_1*4) + 3), 9)*49)) + (ry.outer.outer*7)) + (floormod(blockIdx.x, 7)*7)) + floormod(((threadIdx.x_1*4) + 3), 9)) - 8)], 0 [...]
+ for (rc.outer.outer: int32, 0, 32) {
+ let cse_var_1: int32 = (rc.outer.outer*784)
+ {
+ attr [IterVar(threadIdx.x_1: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1: Buffer(pad_temp.shared, float32, [432], [], scope="shared")[threadIdx.x_1] = @tir.if_then_else(((((1 <= (floordiv(floormod(threadIdx.x_1, 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod(threadIdx.x_1, 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[(((((cse_var_1 + (floordiv(threadIdx.x_1, 27)*49)) + (floordiv(floormod(threadIdx.x_1 [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 32)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 32), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 32), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 5), 9))) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 32), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 32), 27), 9)*7)) + (floormo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 64)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 64), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 64), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 64), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 64), 27), 9)*7)) + (floormo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 96)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 96), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 96), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 96), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 96), 27), 9)*7)) + (floormo [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 128)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 128), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 128), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 128), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 128), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 160)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 160), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 160), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 7), 9))) && (floormod((threadIdx.x_1 + 7), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 160), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 160), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 192)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 192), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 192), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 3), 9))) && (floormod((threadIdx.x_1 + 3), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 192), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 192), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 224)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 224), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 224), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 8), 9))) && (floormod((threadIdx.x_1 + 8), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 224), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 224), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 256)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 256), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 256), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 4), 9))) && (floormod((threadIdx.x_1 + 4), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 256), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 256), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 288)] = @tir.if_then_else(((((1 <= (floormod(blockIdx.x, 7) + floormod((floordiv(threadIdx.x_1, 9) + 2), 3))) && ((floormod(blockIdx.x, 7) + floormod((floordiv(threadIdx.x_1, 9) + 2), 3)) < 8)) && (1 <= floormod(threadIdx.x_1, 9))) && (floormod(threadIdx.x_1, 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 288), 27)*49)) + (floormod(blockIdx.x, 7)*7)) + (floormod((floordiv(threadIdx.x_1, 9) + 2), 3) [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 320)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 320), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 320), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 5), 9))) && (floormod((threadIdx.x_1 + 5), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 320), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 320), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 352)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 352), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 352), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 1), 9))) && (floormod((threadIdx.x_1 + 1), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 352), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 352), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ pad_temp.shared_1[(threadIdx.x_1 + 384)] = @tir.if_then_else(((((1 <= (floordiv(floormod((threadIdx.x_1 + 384), 27), 9) + floormod(blockIdx.x, 7))) && ((floordiv(floormod((threadIdx.x_1 + 384), 27), 9) + floormod(blockIdx.x, 7)) < 8)) && (1 <= floormod((threadIdx.x_1 + 6), 9))) && (floormod((threadIdx.x_1 + 6), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 384), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 384), 27), 9)*7)) + (fl [...]
+ attr [IterVar(threadIdx.x_1, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ if @tir.likely((threadIdx.x_1 < 16), dtype=bool) {
+ pad_temp.shared_1[(threadIdx.x_1 + 416)] = @tir.if_then_else(((((floordiv(floormod((threadIdx.x_1 + 416), 27), 9) + floormod(blockIdx.x, 7)) < 8) && (1 <= floormod((threadIdx.x_1 + 2), 9))) && (floormod((threadIdx.x_1 + 2), 9) < 8)), data[(((((cse_var_1 + (floordiv((threadIdx.x_1 + 416), 27)*49)) + (floordiv(floormod((threadIdx.x_1 + 416), 27), 9)*7)) + (floormod(blockIdx.x, 7)*7)) + floormod((threadIdx.x_1 + 2), 9)) - 8)], 0f32, dtype=float32)
+ }
+ for (ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer: int32, 0, 144) {
+ attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 32;
+ kernel.shared_1: Buffer(kernel.shared, float32, [4608], [], scope="shared")[((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*32) + threadIdx.x_2)] = kernel[((((floordiv(blockIdx.x, 7)*147456) + (floordiv(((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*2) + floordiv(threadIdx.x_2, 16)), 9)*4608)) + (rc.outer.outer*144)) + floormod(((ax0.ax1.fused.ax2.fused.ax3.fused.outer.outer*32) + threadIdx.x_2), 144))]
+ }
+ for (rc.outer.inner: int32, 0, 4) {
+ for (ry.outer.inner: int32, 0, 3) {
+ for (rc.inner: int32, 0, 4) {
+ let cse_var_9: int32 = (((rc.outer.inner*108) + (rc.inner*27)) + (ry.outer.inner*9))
+ let cse_var_8: int32 = (cse_var_9 + 7)
+ let cse_var_7: int32 = (cse_var_9 + 6)
+ let cse_var_6: int32 = (cse_var_9 + 5)
+ let cse_var_5: int32 = (cse_var_9 + 4)
+ let cse_var_4: int32 = (cse_var_9 + 3)
+ let cse_var_3: int32 = (cse_var_9 + 2)
+ let cse_var_2: int32 = (cse_var_9 + 1)
+ {
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_9]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_2]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3))]))
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_2]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[cse_var_8]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 1)]))
+ conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[cse_var_3]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[cse_var_4]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[cse_var_5]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[cse_var_6]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[cse_var_7]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[cse_var_8]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[(cse_var_9 + 8)]*kernel.shared_1[(((((threadIdx.x*144) + (rc.outer.inner*36)) + (rc.inner*9)) + (ry.outer.inner*3)) + 2)]))
+ }
}
}
- attr [IterVar(threadIdx.x_2: int32, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1: Buffer(kernel.shared, float32, [3072], [], scope="shared")[threadIdx.x_2] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(threadIdx.x_2, 24)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 64)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 8), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 64), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 128)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 16), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 128), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 192)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 36864)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 256)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 32), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 256), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 320)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 40), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 320), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 384)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 73728)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 448)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 56), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 448), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 512)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 64), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 512), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 576)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 110592)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 640)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 80), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 640), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 704)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 88), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 704), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 768)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 147456)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 832)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 104), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 832), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 896)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 112), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 896), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 960)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 184320)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1024)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 128), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1024), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1088)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 136), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1088), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1152)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 221184)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1216)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 152), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1216), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1280)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 160), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1280), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1344)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 258048)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1408)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 176), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1408), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1472)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 184), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1472), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1536)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 294912)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1600)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 200), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1600), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1664)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 208), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1664), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1728)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 331776)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1792)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 224), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1792), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1856)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 232), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1856), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1920)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 368640)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 1984)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 248), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 1984), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2048)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 256), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2048), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2112)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 405504)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2176)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 272), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2176), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2240)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 280), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2240), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2304)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 442368)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2368)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 296), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2368), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2432)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 304), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2432), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2496)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 479232)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2560)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 320), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2560), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2624)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 328), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2624), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2688)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 516096)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2752)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 344), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2752), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2816)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 352), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2816), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2880)] = kernel[(((((((floordiv(blockIdx.x, 7)*589824) + (floordiv(floordiv(threadIdx.x_2, 8), 3)*4608)) + cse_var_2) + (floordiv(floormod(threadIdx.x_2, 24), 3)*9)) + cse_var_1) + floormod(threadIdx.x_2, 3)) + 552960)]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 2944)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 368), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 2944), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 1), 3))]
- attr [IterVar(threadIdx.x_2, (nullptr), "ThreadIndex", "threadIdx.x")] "thread_extent" = 64;
- kernel.shared_1[(threadIdx.x_2 + 3008)] = kernel[((((((floordiv(blockIdx.x, 7)*589824) + (floordiv((floordiv(threadIdx.x_2, 8) + 376), 3)*4608)) + cse_var_2) + (floordiv(floormod((threadIdx.x_2 + 3008), 24), 3)*9)) + cse_var_1) + floormod((threadIdx.x_2 + 2), 3))]
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[0]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[9]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[1]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[2]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[3]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[4]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[5]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[6]*kernel.shared_1[(threadIdx.x*48)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 3)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[0]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[9]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 24)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 27)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 1)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 4)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[1]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[10]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 25)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 28)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[8]*kernel.shared_1[((threadIdx.x*48) + 2)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[17]*kernel.shared_1[((threadIdx.x*48) + 5)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[2]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[11]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[3]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[12]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[4]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[13]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[5]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[14]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[6]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[15]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[7]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[16]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[8]*kernel.shared_1[((threadIdx.x*48) + 26)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[17]*kernel.shared_1[((threadIdx.x*48) + 29)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[18]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[27]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 6)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 9)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[18]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[27]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 30)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 33)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 7)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 10)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[19]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[28]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 31)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 34)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[26]*kernel.shared_1[((threadIdx.x*48) + 8)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[35]*kernel.shared_1[((threadIdx.x*48) + 11)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[20]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[29]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[21]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[30]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[22]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[31]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[23]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[32]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[24]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[33]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[25]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[34]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[26]*kernel.shared_1[((threadIdx.x*48) + 32)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[35]*kernel.shared_1[((threadIdx.x*48) + 35)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[36]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[45]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 12)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 15)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[36]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[45]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 36)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 39)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 13)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 16)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[37]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[46]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 37)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 40)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[44]*kernel.shared_1[((threadIdx.x*48) + 14)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[53]*kernel.shared_1[((threadIdx.x*48) + 17)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[38]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[47]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[39]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[48]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[40]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[49]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[41]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[50]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[42]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[51]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[43]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[52]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[44]*kernel.shared_1[((threadIdx.x*48) + 38)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[53]*kernel.shared_1[((threadIdx.x*48) + 41)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[54]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[63]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 18)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 21)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[54]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[63]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 42)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 45)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 19)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 22)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[55]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[64]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 43)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 46)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[0] = (conv2d_nchw_1[0] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[1] = (conv2d_nchw_1[1] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[2] = (conv2d_nchw_1[2] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[3] = (conv2d_nchw_1[3] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[4] = (conv2d_nchw_1[4] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[5] = (conv2d_nchw_1[5] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[62]*kernel.shared_1[((threadIdx.x*48) + 20)]))
- conv2d_nchw_1[6] = (conv2d_nchw_1[6] + (pad_temp.shared_1[71]*kernel.shared_1[((threadIdx.x*48) + 23)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[56]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[7] = (conv2d_nchw_1[7] + (pad_temp.shared_1[65]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[57]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[8] = (conv2d_nchw_1[8] + (pad_temp.shared_1[66]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[58]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[9] = (conv2d_nchw_1[9] + (pad_temp.shared_1[67]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[59]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[10] = (conv2d_nchw_1[10] + (pad_temp.shared_1[68]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[60]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[11] = (conv2d_nchw_1[11] + (pad_temp.shared_1[69]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[61]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[12] = (conv2d_nchw_1[12] + (pad_temp.shared_1[70]*kernel.shared_1[((threadIdx.x*48) + 47)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[62]*kernel.shared_1[((threadIdx.x*48) + 44)]))
- conv2d_nchw_1[13] = (conv2d_nchw_1[13] + (pad_temp.shared_1[71]*kernel.shared_1[((threadIdx.x*48) + 47)]))
}
}
}
- for (i1.inner: int32, 0, 2) {
- for (i3.inner: int32, 0, 7) {
- compute[(((((floordiv(blockIdx.x, 7)*6272) + (threadIdx.x*98)) + (i1.inner*49)) + (floormod(blockIdx.x, 7)*7)) + i3.inner)] = max((conv2d_nchw_1[((i1.inner*7) + i3.inner)] + bias[(((floordiv(blockIdx.x, 7)*128) + (threadIdx.x*2)) + i1.inner)]), 0f32)
- }
- }
+ compute[(((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7))] = max((conv2d_nchw_1[0] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 1)] = max((conv2d_nchw_1[1] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 2)] = max((conv2d_nchw_1[2] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 3)] = max((conv2d_nchw_1[3] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 4)] = max((conv2d_nchw_1[4] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 5)] = max((conv2d_nchw_1[5] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
+ compute[((((floordiv(blockIdx.x, 7)*1568) + (threadIdx.x*49)) + (floormod(blockIdx.x, 7)*7)) + 6)] = max((conv2d_nchw_1[6] + bias[((floordiv(blockIdx.x, 7)*32) + threadIdx.x)]), 0f32)
}
}
</pre></div>
@@ -984,7 +601,7 @@ cooperative fetching, unrolling and operator fusion.</p>
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.362 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 0.343 ms
</pre></div>
</div>
</div>
@@ -1015,36 +632,36 @@ 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=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_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=32)
conv2d_nchw_ff_o_o_o_o, conv2d_nchw_ff_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_ff_o_o_o_i, factor=1)
conv2d_nchw_yy_o_i, conv2d_nchw_yy_i = s[conv2d_nchw].split(conv2d_nchw_yy, factor=1)
conv2d_nchw_yy_o_o_i, conv2d_nchw_yy_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_i, factor=1)
conv2d_nchw_yy_o_o_o_i, conv2d_nchw_yy_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_i, factor=1)
conv2d_nchw_yy_o_o_o_o, conv2d_nchw_yy_o_o_o_i = s[conv2d_nchw].split(conv2d_nchw_yy_o_o_o_i, factor=1)
conv2d_nchw_xx_o_i, conv2d_nchw_xx_i = s[conv2d_nchw].split(conv2d_nchw_xx, factor=1)
-conv2d_nchw_xx_o_o_i, conv2d_nchw_xx_o_i = s[conv2d_nchw].split(conv2d_nchw_xx_o_i, factor=7)
+conv2d_nchw_xx_o_o_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=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=2)
+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=7)
+conv2d_nchw_rc_o_i, conv2d_nchw_rc_i = s[conv2d_nchw].split(conv2d_nchw_rc, factor=4)
conv2d_nchw_rc_o_o, conv2d_nchw_rc_o_i = s[conv2d_nchw].split(conv2d_nchw_rc_o_i, factor=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=3)
+conv2d_nchw_ry_o_o, conv2d_nchw_ry_o_i = s[conv2d_nchw].split(conv2d_nchw_ry_o_i, factor=3)
+conv2d_nchw_rx_o_i, conv2d_nchw_rx_i = s[conv2d_nchw].split(conv2d_nchw_rx, factor=3)
+conv2d_nchw_rx_o_o, conv2d_nchw_rx_o_i = s[conv2d_nchw].split(conv2d_nchw_rx_o_i, factor=1)
s[conv2d_nchw].reorder(conv2d_nchw_nn_o_o_o_o, conv2d_nchw_ff_o_o_o_o, conv2d_nchw_yy_o_o_o_o, conv2d_nchw_xx_o_o_o_o, conv2d_nchw_nn_o_o_o_i, conv2d_nchw_ff_o_o_o_i, conv2d_nchw_yy_o_o_o_i, conv2d_nchw_xx_o_o_o_i, conv2d_nchw_nn_o_o_i, conv2d_nchw_ff_o_o_i, conv2d_nchw_yy_o_o_i, conv2d_nchw_xx_o_o_i, conv2d_nchw_rc_o_o, conv2d_nchw_ry_o_o, conv2d_nchw_rx_o_o, conv2d_nchw_rc_o_i, conv2d_nchw_ry_o_i, conv2d_nchw_rx_o_i, conv2d_nchw_nn_o_i, conv2d_nchw_ff_o_i, conv2d_nchw_yy_o_i, conv2d_nc [...]
compute_i0_o_i, compute_i0_i = s[compute].split(compute_i0, factor=1)
compute_i0_o_o_i, compute_i0_o_i = s[compute].split(compute_i0_o_i, factor=1)
compute_i0_o_o_o, compute_i0_o_o_i = s[compute].split(compute_i0_o_o_i, factor=1)
-compute_i1_o_i, compute_i1_i = s[compute].split(compute_i1, factor=2)
-compute_i1_o_o_i, compute_i1_o_i = s[compute].split(compute_i1_o_i, factor=64)
+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=32)
compute_i1_o_o_o, compute_i1_o_o_i = s[compute].split(compute_i1_o_o_i, factor=1)
compute_i2_o_i, compute_i2_i = s[compute].split(compute_i2, factor=1)
compute_i2_o_o_i, compute_i2_o_i = s[compute].split(compute_i2_o_i, factor=1)
compute_i2_o_o_o, compute_i2_o_o_i = s[compute].split(compute_i2_o_o_i, factor=1)
-compute_i3_o_i, compute_i3_i = s[compute].split(compute_i3, factor=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=1)
-compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=1)
+compute_i3_o_o_o, compute_i3_o_o_i = s[compute].split(compute_i3_o_o_i, factor=7)
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)
kernel_shared = s.cache_read(kernel, "shared", [conv2d_nchw])
@@ -1063,14 +680,14 @@ 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=64)
+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=32)
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=4)
+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)
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=64)
+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=32)
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", 512)
+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, "unroll_explicit", True)
CUDA source code:
@@ -1088,10 +705,10 @@ CUDA source code:
#define int64_t long long
#define uint64_t unsigned long long
#endif
-extern "C" __global__ void __launch_bounds__(64) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
- float conv2d_nchw[14];
- __shared__ float pad_temp_shared[72];
- __shared__ float kernel_shared[3072];
+extern "C" __global__ void __launch_bounds__(32) default_function_kernel0(float* __restrict__ data, float* __restrict__ kernel, float* __restrict__ compute, float* __restrict__ bias) {
+ float conv2d_nchw[7];
+ __shared__ float pad_temp_shared[432];
+ __shared__ float kernel_shared[4608];
conv2d_nchw[0] = 0.000000e+00f;
conv2d_nchw[1] = 0.000000e+00f;
conv2d_nchw[2] = 0.000000e+00f;
@@ -1099,420 +716,63 @@ extern "C" __global__ void __launch_bounds__(64) default_function_kern
conv2d_nchw[4] = 0.000000e+00f;
conv2d_nchw[5] = 0.000000e+00f;
conv2d_nchw[6] = 0.000000e+00f;
- conv2d_nchw[7] = 0.000000e+00f;
- conv2d_nchw[8] = 0.000000e+00f;
- conv2d_nchw[9] = 0.000000e+00f;
- conv2d_nchw[10] = 0.000000e+00f;
- conv2d_nchw[11] = 0.000000e+00f;
- conv2d_nchw[12] = 0.000000e+00f;
- conv2d_nchw[13] = 0.000000e+00f;
- for (int rc_outer_outer = 0; rc_outer_outer < 64; ++rc_outer_outer) {
- for (int ry_outer_outer = 0; ry_outer_outer < 3; ++ry_outer_outer) {
- __syncthreads();
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[(((int)threadIdx.x) * 4)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) * 4) % 9))) && (((((int)threadIdx.x) * 4) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + (((((int)threadIdx.x) * 4) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) * 4) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 1)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 1) % 9))) && ((((((int)threadIdx.x) * 4) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 1) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 1) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 2)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 2) % 9))) && ((((((int)threadIdx.x) * 4) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 2) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 2) % 9)) - 8)] : 0.000000e+00f);
- }
- if (((int)threadIdx.x) < 18) {
- pad_temp_shared[((((int)threadIdx.x) * 4) + 3)] = (((((1 <= (ry_outer_outer + (((int)blockIdx.x) % 7))) && ((ry_outer_outer + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((((int)threadIdx.x) * 4) + 3) % 9))) && ((((((int)threadIdx.x) * 4) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 392) + ((((((int)threadIdx.x) * 4) + 3) / 9) * 49)) + (ry_outer_outer * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((((int)threadIdx.x) * 4) + 3) % 9)) - 8)] : 0.000000e+00f);
- }
- kernel_shared[((int)threadIdx.x)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3))];
- kernel_shared[(((int)threadIdx.x) + 64)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 64) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 128)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 128) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 192)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 36864)];
- kernel_shared[(((int)threadIdx.x) + 256)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 256) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 320)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 320) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 384)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 73728)];
- kernel_shared[(((int)threadIdx.x) + 448)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 448) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 512)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 512) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 576)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 110592)];
- kernel_shared[(((int)threadIdx.x) + 640)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 640) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 704)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 704) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 768)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 147456)];
- kernel_shared[(((int)threadIdx.x) + 832)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 832) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 896)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 896) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 960)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 184320)];
- kernel_shared[(((int)threadIdx.x) + 1024)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1024) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1088)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1088) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1152)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 221184)];
- kernel_shared[(((int)threadIdx.x) + 1216)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1216) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1280)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1280) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1344)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 258048)];
- kernel_shared[(((int)threadIdx.x) + 1408)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1408) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1472)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1472) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1536)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 294912)];
- kernel_shared[(((int)threadIdx.x) + 1600)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1600) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1664)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1664) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1728)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 331776)];
- kernel_shared[(((int)threadIdx.x) + 1792)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1792) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1856)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1856) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 1920)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 368640)];
- kernel_shared[(((int)threadIdx.x) + 1984)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 1984) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2048)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2048) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2112)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 405504)];
- kernel_shared[(((int)threadIdx.x) + 2176)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2176) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2240)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2240) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2304)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 442368)];
- kernel_shared[(((int)threadIdx.x) + 2368)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2368) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2432)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2432) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2496)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 479232)];
- kernel_shared[(((int)threadIdx.x) + 2560)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2560) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2624)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2624) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2688)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 516096)];
- kernel_shared[(((int)threadIdx.x) + 2752)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2752) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2816)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2816) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- kernel_shared[(((int)threadIdx.x) + 2880)] = kernel[((((((((((int)blockIdx.x) / 7) * 589824) + ((((int)threadIdx.x) / 24) * 4608)) + (rc_outer_outer * 72)) + (((((int)threadIdx.x) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + (((int)threadIdx.x) % 3)) + 552960)];
- kernel_shared[(((int)threadIdx.x) + 2944)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 2944) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 16) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 1) % 3))];
- kernel_shared[(((int)threadIdx.x) + 3008)] = kernel[(((((((((int)blockIdx.x) / 7) * 589824) + (((((int)threadIdx.x) + 3008) / 24) * 4608)) + (rc_outer_outer * 72)) + ((((((int)threadIdx.x) + 8) % 24) / 3) * 9)) + (ry_outer_outer * 3)) + ((((int)threadIdx.x) + 2) % 3))];
- __syncthreads();
- conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[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 rc_outer_outer = 0; rc_outer_outer < 32; ++rc_outer_outer) {
+ __syncthreads();
+ pad_temp_shared[((int)threadIdx.x)] = (((((1 <= (((((int)threadIdx.x) % 27) / 9) + (((int)blockIdx.x) % 7))) && ((((((int)threadIdx.x) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + ((((int)threadIdx.x) / 27) * 49)) + (((((int)threadIdx.x) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + (((int)threadIdx.x) % 9)) - 8)] : 0.000000e+00f);
+ pad_temp_shared[(((int)threadIdx.x) + 32)] = (((((1 <= ((((((int)threadIdx.x) + 5) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 5) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 5) % 9))) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 32) / 27) * 49)) + ((((((int)threadIdx.x) + 5) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)thr [...]
+ pad_temp_shared[(((int)threadIdx.x) + 64)] = (((((1 <= ((((((int)threadIdx.x) + 10) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 10) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 64) / 27) * 49)) + ((((((int)threadIdx.x) + 10) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int) [...]
+ pad_temp_shared[(((int)threadIdx.x) + 96)] = (((((1 <= ((((((int)threadIdx.x) + 15) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 15) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 96) / 27) * 49)) + ((((((int)threadIdx.x) + 15) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int) [...]
+ pad_temp_shared[(((int)threadIdx.x) + 128)] = (((((1 <= ((((((int)threadIdx.x) + 20) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 20) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 128) / 27) * 49)) + ((((((int)threadIdx.x) + 20) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((in [...]
+ pad_temp_shared[(((int)threadIdx.x) + 160)] = (((((1 <= ((((((int)threadIdx.x) + 25) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 25) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 7) % 9))) && (((((int)threadIdx.x) + 7) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 160) / 27) * 49)) + ((((((int)threadIdx.x) + 25) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((in [...]
+ pad_temp_shared[(((int)threadIdx.x) + 192)] = (((((1 <= ((((((int)threadIdx.x) + 3) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 3) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 3) % 9))) && (((((int)threadIdx.x) + 3) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 192) / 27) * 49)) + ((((((int)threadIdx.x) + 3) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)t [...]
+ pad_temp_shared[(((int)threadIdx.x) + 224)] = (((((1 <= ((((((int)threadIdx.x) + 8) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 8) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 8) % 9))) && (((((int)threadIdx.x) + 8) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 224) / 27) * 49)) + ((((((int)threadIdx.x) + 8) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)t [...]
+ pad_temp_shared[(((int)threadIdx.x) + 256)] = (((((1 <= ((((((int)threadIdx.x) + 13) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 13) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 4) % 9))) && (((((int)threadIdx.x) + 4) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 256) / 27) * 49)) + ((((((int)threadIdx.x) + 13) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((in [...]
+ pad_temp_shared[(((int)threadIdx.x) + 288)] = (((((1 <= ((((int)blockIdx.x) % 7) + (((((int)threadIdx.x) / 9) + 2) % 3))) && (((((int)blockIdx.x) % 7) + (((((int)threadIdx.x) / 9) + 2) % 3)) < 8)) && (1 <= (((int)threadIdx.x) % 9))) && ((((int)threadIdx.x) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 288) / 27) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + ((((((int)threadIdx.x) / 9) + 2) % 3) * 7)) + (((int)threadIdx.x) % 9) [...]
+ pad_temp_shared[(((int)threadIdx.x) + 320)] = (((((1 <= ((((((int)threadIdx.x) + 23) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 23) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 5) % 9))) && (((((int)threadIdx.x) + 5) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 320) / 27) * 49)) + ((((((int)threadIdx.x) + 23) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((in [...]
+ pad_temp_shared[(((int)threadIdx.x) + 352)] = (((((1 <= ((((((int)threadIdx.x) + 1) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 1) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 1) % 9))) && (((((int)threadIdx.x) + 1) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 352) / 27) * 49)) + ((((((int)threadIdx.x) + 1) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)t [...]
+ pad_temp_shared[(((int)threadIdx.x) + 384)] = (((((1 <= ((((((int)threadIdx.x) + 6) % 27) / 9) + (((int)blockIdx.x) % 7))) && (((((((int)threadIdx.x) + 6) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8)) && (1 <= ((((int)threadIdx.x) + 6) % 9))) && (((((int)threadIdx.x) + 6) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 384) / 27) * 49)) + ((((((int)threadIdx.x) + 6) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)t [...]
+ if (((int)threadIdx.x) < 16) {
+ pad_temp_shared[(((int)threadIdx.x) + 416)] = ((((((((((int)threadIdx.x) + 11) % 27) / 9) + (((int)blockIdx.x) % 7)) < 8) && (1 <= ((((int)threadIdx.x) + 2) % 9))) && (((((int)threadIdx.x) + 2) % 9) < 8)) ? data[((((((rc_outer_outer * 784) + (((((int)threadIdx.x) + 416) / 27) * 49)) + ((((((int)threadIdx.x) + 11) % 27) / 9) * 7)) + ((((int)blockIdx.x) % 7) * 7)) + ((((int)threadIdx.x) + 2) % 9)) - 8)] : 0.000000e+00f);
}
- }
- for (int i1_inner = 0; i1_inner < 2; ++i1_inner) {
- for (int i3_inner = 0; i3_inner < 7; ++i3_inner) {
- compute[((((((((int)blockIdx.x) / 7) * 6272) + (((int)threadIdx.x) * 98)) + (i1_inner * 49)) + ((((int)blockIdx.x) % 7) * 7)) + i3_inner)] = max((conv2d_nchw[((i1_inner * 7) + i3_inner)] + bias[((((((int)blockIdx.x) / 7) * 128) + (((int)threadIdx.x) * 2)) + i1_inner)]), 0.000000e+00f);
+ for (int ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer = 0; ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer < 144; ++ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer) {
+ kernel_shared[((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 32) + ((int)threadIdx.x))] = kernel[(((((((int)blockIdx.x) / 7) * 147456) + ((((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 2) + (((int)threadIdx.x) >> 4)) / 9) * 4608)) + (rc_outer_outer * 144)) + (((ax0_ax1_fused_ax2_fused_ax3_fused_outer_outer * 32) + ((int)threadIdx.x)) % 144))];
+ }
+ __syncthreads();
+ for (int rc_outer_inner = 0; rc_outer_inner < 4; ++rc_outer_inner) {
+ for (int ry_outer_inner = 0; ry_outer_inner < 3; ++ry_outer_inner) {
+ for (int rc_inner = 0; rc_inner < 4; ++rc_inner) {
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[(((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9))] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 1)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3))]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 1)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 7)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 1)]));
+ conv2d_nchw[0] = (conv2d_nchw[0] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 2)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[1] = (conv2d_nchw[1] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 3)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[2] = (conv2d_nchw[2] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 4)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[3] = (conv2d_nchw[3] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 5)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[4] = (conv2d_nchw[4] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 6)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[5] = (conv2d_nchw[5] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 7)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ conv2d_nchw[6] = (conv2d_nchw[6] + (pad_temp_shared[((((rc_outer_inner * 108) + (rc_inner * 27)) + (ry_outer_inner * 9)) + 8)] * kernel_shared[(((((((int)threadIdx.x) * 144) + (rc_outer_inner * 36)) + (rc_inner * 9)) + (ry_outer_inner * 3)) + 2)]));
+ }
+ }
}
}
+ compute[((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7))] = max((conv2d_nchw[0] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 1)] = max((conv2d_nchw[1] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 2)] = max((conv2d_nchw[2] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 3)] = max((conv2d_nchw[3] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 4)] = max((conv2d_nchw[4] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 5)] = max((conv2d_nchw[5] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
+ compute[(((((((int)blockIdx.x) / 7) * 1568) + (((int)threadIdx.x) * 49)) + ((((int)blockIdx.x) % 7) * 7)) + 6)] = max((conv2d_nchw[6] + bias[(((((int)blockIdx.x) / 7) * 32) + ((int)threadIdx.x))]), 0.000000e+00f);
}
</pre></div>
</div>
@@ -1549,7 +809,7 @@ 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> ( 2 minutes 17.757 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 2 minutes 20.501 seconds)</p>
<div class="sphx-glr-footer class 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 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 53bcdc135..5acc0249d 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_cuda.html
@@ -876,7 +876,7 @@ 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)
- 9.9415 9.9341 9.9663 9.9240 0.0180
+ 9.7607 9.7678 9.8201 9.6943 0.0516
</pre></div>
</div>
</div>
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 0f0feb450..339af7c4f 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_network_x86.html
@@ -895,7 +895,7 @@ 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)
- 764.1780 765.6010 766.2926 760.6404 2.5173
+ 766.9284 768.4643 770.4900 761.8308 3.6982
</pre></div>
</div>
</div>
@@ -917,7 +917,7 @@ 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 20.507 seconds)</p>
+<p class="sphx-glr-timing"><strong>Total running time of the script:</strong> ( 1 minutes 21.450 seconds)</p>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-tune-with-autoscheduler-tune-network-x86-py">
<div class="sphx-glr-download 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 b2368abcd..fd5bc3845 100644
--- a/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
+++ b/docs/how_to/tune_with_autoscheduler/tune_sparse_x86.html
@@ -600,74 +600,80 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
placeholder_4: Buffer(placeholder_14: Pointer(float32), float32, [65536], []),
compute: Buffer(compute_2: Pointer(float32), float32, [65536], [])}
buffer_map = {placeholder_5: placeholder, placeholder_6: placeholder_1, placeholder_7: placeholder_2, placeholder_8: placeholder_3, placeholder_9: placeholder_4, compute_1: compute}
- preflattened_buffer_map = {placeholder_7: placeholder_15: Buffer(placeholder_12, int32, [4916], []), placeholder_5: placeholder_16: Buffer(placeholder_10, float32, [128, 256], []), placeholder_6: placeholder_17: Buffer(placeholder_11, float32, [4916, 16, 1], []), placeholder_8: placeholder_18: Buffer(placeholder_13, int32, [33], []), compute_1: compute_3: Buffer(compute_2, float32, [128, 512], []), placeholder_9: placeholder_19: Buffer(placeholder_14, float32, [128, 512], [])} {
- for (i0.outer.i1.outer.fused: int32, 0, 512) "parallel" {
- allocate(compute_4: Pointer(global float32), float32, [128]), storage_scope = global {
- for (i.inner.init: int32, 0, 8) {
- let cse_var_1: int32 = (i.inner.init*16)
- {
- compute_5: Buffer(compute_4, float32, [128], [])[cse_var_1] = 0f32
- compute_5[(cse_var_1 + 1)] = 0f32
- compute_5[(cse_var_1 + 2)] = 0f32
- compute_5[(cse_var_1 + 3)] = 0f32
- compute_5[(cse_var_1 + 4)] = 0f32
- compute_5[(cse_var_1 + 5)] = 0f32
- compute_5[(cse_var_1 + 6)] = 0f32
- compute_5[(cse_var_1 + 7)] = 0f32
- compute_5[(cse_var_1 + 8)] = 0f32
- compute_5[(cse_var_1 + 9)] = 0f32
- compute_5[(cse_var_1 + 10)] = 0f32
- compute_5[(cse_var_1 + 11)] = 0f32
- compute_5[(cse_var_1 + 12)] = 0f32
- compute_5[(cse_var_1 + 13)] = 0f32
- compute_5[(cse_var_1 + 14)] = 0f32
- compute_5[(cse_var_1 + 15)] = 0f32
- }
- }
- for (elem_idx: int32, 0, let cse_var_2: int32 = floormod(i0.outer.i1.outer.fused, 32) in (placeholder_3[(cse_var_2 + 1)] - placeholder_3[cse_var_2])) {
- for (i.inner: int32, 0, 8) {
- let cse_var_21: int32 = floormod(i0.outer.i1.outer.fused, 32)
- let cse_var_20: int32 = (i.inner*16)
- let cse_var_19: int32 = (elem_idx*16)
- let cse_var_18: int32 = (cse_var_20 + 10)
- let cse_var_17: int32 = (cse_var_20 + 11)
- let cse_var_16: int32 = (cse_var_20 + 12)
- let cse_var_15: int32 = (cse_var_20 + 13)
- let cse_var_14: int32 = (cse_var_20 + 14)
- let cse_var_13: int32 = (cse_var_20 + 15)
- let cse_var_12: int32 = (cse_var_20 + 2)
- let cse_var_11: int32 = (cse_var_20 + 3)
- let cse_var_10: int32 = (cse_var_20 + 4)
- let cse_var_9: int32 = (cse_var_20 + 5)
- let cse_var_8: int32 = (cse_var_20 + 6)
- let cse_var_7: int32 = (cse_var_20 + 7)
- let cse_var_6: int32 = (cse_var_20 + 8)
- let cse_var_5: int32 = (cse_var_20 + 9)
- let cse_var_4: int32 = (cse_var_20 + 1)
- let cse_var_3: int32 = ((floordiv(i0.outer.i1.outer.fused, 32)*2048) + (i.inner*256))
- {
- compute_5[cse_var_20] = (compute_5[cse_var_20] + (placeholder_1[((placeholder_3[cse_var_21]*16) + cse_var_19)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_4] = (compute_5[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 1)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_12] = (compute_5[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 2)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_11] = (compute_5[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 3)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_10] = (compute_5[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 4)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_9] = (compute_5[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 5)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_8] = (compute_5[cse_var_8] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 6)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_7] = (compute_5[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 7)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_6] = (compute_5[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 8)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_5] = (compute_5[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 9)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_18] = (compute_5[cse_var_18] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 10)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_17] = (compute_5[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 11)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_16] = (compute_5[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 12)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_15] = (compute_5[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 13)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_14] = (compute_5[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 14)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
- compute_5[cse_var_13] = (compute_5[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_21]*16) + cse_var_19) + 15)]*max(placeholder[(cse_var_3 + placeholder_2[(placeholder_3[cse_var_21] + elem_idx)])], 0f32)))
+ preflattened_buffer_map = {placeholder_8: placeholder_15: Buffer(placeholder_13, int32, [33], []), compute_1: compute_3: Buffer(compute_2, float32, [128, 512], []), placeholder_9: placeholder_16: Buffer(placeholder_14, float32, [128, 512], []), placeholder_6: placeholder_17: Buffer(placeholder_11, float32, [4916, 16, 1], []), placeholder_7: placeholder_18: Buffer(placeholder_12, int32, [4916], []), placeholder_5: placeholder_19: Buffer(placeholder_10, float32, [128, 256], [])} {
+ for (i0.outer.i1.outer.fused: int32, 0, 16) "parallel" {
+ allocate(compute_4: Pointer(global float32), float32, [4096]), storage_scope = global {
+ for (i.outer.inner: int32, 0, 4) {
+ for (nb_j.inner: int32, 0, 2) {
+ for (i.inner.init: int32, 0, 32) {
+ let cse_var_1: int32 = (((i.outer.inner*1024) + (i.inner.init*32)) + (nb_j.inner*16))
+ {
+ compute_5: Buffer(compute_4, float32, [4096], [])[cse_var_1] = 0f32
+ compute_5[(cse_var_1 + 1)] = 0f32
+ compute_5[(cse_var_1 + 2)] = 0f32
+ compute_5[(cse_var_1 + 3)] = 0f32
+ compute_5[(cse_var_1 + 4)] = 0f32
+ compute_5[(cse_var_1 + 5)] = 0f32
+ compute_5[(cse_var_1 + 6)] = 0f32
+ compute_5[(cse_var_1 + 7)] = 0f32
+ compute_5[(cse_var_1 + 8)] = 0f32
+ compute_5[(cse_var_1 + 9)] = 0f32
+ compute_5[(cse_var_1 + 10)] = 0f32
+ compute_5[(cse_var_1 + 11)] = 0f32
+ compute_5[(cse_var_1 + 12)] = 0f32
+ compute_5[(cse_var_1 + 13)] = 0f32
+ compute_5[(cse_var_1 + 14)] = 0f32
+ compute_5[(cse_var_1 + 15)] = 0f32
+ }
+ }
+ for (elem_idx: int32, 0, let cse_var_2: int32 = ((i0.outer.i1.outer.fused*2) + nb_j.inner) in (placeholder_3[(cse_var_2 + 1)] - placeholder_3[cse_var_2])) {
+ for (i.inner: int32, 0, 32) {
+ let cse_var_21: int32 = (elem_idx*16)
+ let cse_var_20: int32 = ((i0.outer.i1.outer.fused*2) + nb_j.inner)
+ let cse_var_19: int32 = ((i.outer.inner*8192) + (i.inner*256))
+ let cse_var_18: int32 = (((i.outer.inner*1024) + (i.inner*32)) + (nb_j.inner*16))
+ let cse_var_17: int32 = (cse_var_18 + 1)
+ let cse_var_16: int32 = (cse_var_18 + 11)
+ let cse_var_15: int32 = (cse_var_18 + 12)
+ let cse_var_14: int32 = (cse_var_18 + 13)
+ let cse_var_13: int32 = (cse_var_18 + 14)
+ let cse_var_12: int32 = (cse_var_18 + 15)
+ let cse_var_11: int32 = (cse_var_18 + 2)
+ let cse_var_10: int32 = (cse_var_18 + 3)
+ let cse_var_9: int32 = (cse_var_18 + 4)
+ let cse_var_8: int32 = (cse_var_18 + 5)
+ let cse_var_7: int32 = (cse_var_18 + 6)
+ let cse_var_6: int32 = (cse_var_18 + 7)
+ let cse_var_5: int32 = (cse_var_18 + 8)
+ let cse_var_4: int32 = (cse_var_18 + 9)
+ let cse_var_3: int32 = (cse_var_18 + 10)
+ {
+ compute_5[cse_var_18] = (compute_5[cse_var_18] + (placeholder_1[((placeholder_3[cse_var_20]*16) + cse_var_21)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_17] = (compute_5[cse_var_17] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 1)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_11] = (compute_5[cse_var_11] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 2)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_10] = (compute_5[cse_var_10] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 3)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_9] = (compute_5[cse_var_9] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 4)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_8] = (compute_5[cse_var_8] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 5)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_7] = (compute_5[cse_var_7] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 6)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_6] = (compute_5[cse_var_6] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 7)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_5] = (compute_5[cse_var_5] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 8)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_4] = (compute_5[cse_var_4] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 9)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_3] = (compute_5[cse_var_3] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 10)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_16] = (compute_5[cse_var_16] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 11)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_15] = (compute_5[cse_var_15] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 12)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_14] = (compute_5[cse_var_14] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 13)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_13] = (compute_5[cse_var_13] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 14)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ compute_5[cse_var_12] = (compute_5[cse_var_12] + (placeholder_1[(((placeholder_3[cse_var_20]*16) + cse_var_21) + 15)]*max(placeholder[(cse_var_19 + placeholder_2[(placeholder_3[cse_var_20] + elem_idx)])], 0f32)))
+ }
+ }
}
}
}
- for (i0.inner: int32, 0, 8) {
- let cse_var_22: int32 = (((floordiv(i0.outer.i1.outer.fused, 32)*4096) + (i0.inner*512)) + (floormod(i0.outer.i1.outer.fused, 32)*16))
- compute[ramp(cse_var_22, 1, 16)] = max((compute_5[ramp((i0.inner*16), 1, 16)] + placeholder_4[ramp(cse_var_22, 1, 16)]), broadcast(0f32, 16))
+ for (i0.inner: int32, 0, 128) {
+ for (i1.inner: int32, 0, 32) {
+ let cse_var_22: int32 = (((i0.inner*512) + (i0.outer.i1.outer.fused*32)) + i1.inner)
+ compute[cse_var_22] = max((compute_5[((i0.inner*32) + i1.inner)] + placeholder_4[cse_var_22]), 0f32)
+ }
}
}
}
@@ -706,7 +712,7 @@ layout transformation, parallelization, vectorization, unrolling, and operator f
</pre></div>
</div>
<p class="sphx-glr-script-out">Out:</p>
-<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.877 ms
+<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Execution time of this operator: 1.718 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 2e590a417..c4916b85b 100644
--- a/docs/how_to/tune_with_autotvm/sg_execution_times.html
+++ b/docs/how_to/tune_with_autotvm/sg_execution_times.html
@@ -300,13 +300,13 @@
<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>00:43.302</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
+<p><strong>00:44.660</strong> total execution time for <strong>how_to_tune_with_autotvm</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:42.463</strong>: <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></li>
-<li><p><strong>00:00.222</strong>: <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></li>
-<li><p><strong>00:00.208</strong>: <a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></li>
-<li><p><strong>00:00.208</strong>: <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></li>
-<li><p><strong>00:00.202</strong>: <a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></li>
+<li><p><strong>00:43.786</strong>: <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></li>
+<li><p><strong>00:00.230</strong>: <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></li>
+<li><p><strong>00:00.217</strong>: <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></li>
+<li><p><strong>00:00.214</strong>: <a class="reference internal" href="tune_relay_arm.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-arm-py"><span class="std std-ref">Auto-tuning a Convolutional Network for ARM CPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_arm.py</span></code>)</p></li>
+<li><p><strong>00:00.213</strong>: <a class="reference internal" href="tune_relay_mobile_gpu.html#sphx-glr-how-to-tune-with-autotvm-tune-relay-mobile-gpu-py"><span class="std std-ref">Auto-tuning a Convolutional Network for Mobile GPU</span></a> (<code class="docutils literal notranslate"><span class="pre">tune_relay_mobile_gpu.py</span></code>)</p></li>
</ul>
</div>
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 96a249c82..e2ab315a0 100644
--- a/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
+++ b/docs/how_to/tune_with_autotvm/tune_conv2d_cuda.html
@@ -1142,8 +1142,8 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 4, 4, 32]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 1, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2885496
-No: 6 GFLOPS: 64.10/64.10 result: MeasureResult(costs=(0.0036112859666666665,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5755383968353271, timestamp=1650564545.7379923) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
-No: 7 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 6 GFLOPS: 103.76/103.76 result: MeasureResult(costs=(0.002231081375,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.5949790477752686, timestamp=1650567454.3841782) [('tile_f', [-1, 1, 1, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,3754080
+No: 7 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1266,7 +1266,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 16, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 256, 1]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6225319
-No: 8 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 8 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1389,7 +1389,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 32]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 8, 64]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,943546
-No: 9 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 9 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1512,7 +1512,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 16, 32]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2868708
-No: 10 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 10 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 142, in build
res = future.result()
File "/usr/lib/python3.7/concurrent/futures/_base.py", line 435, in result
@@ -1530,7 +1530,7 @@ No: 10 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
TimeoutError
[('tile_f', [-1, 32, 2, 4]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 4, 2]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4691833
-No: 11 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 11 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1653,7 +1653,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 2, 64]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,1042124
-No: 12 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 12 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1776,7 +1776,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 32, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 32, 16]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,10013405
-No: 13 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 13 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -1899,7 +1899,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 8, 8, 2]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 1, 7, 1]), ('tile_rc', [-1, 4, 32]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6732082
-No: 14 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 14 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2022,7 +2022,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 2, 4, 32]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 1, 1, 1]), ('tile_rc', [-1, 4, 128]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 1)],None,7536735
-No: 15 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 15 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2145,7 +2145,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 4]), ('tile_y', [-1, 1, 1, 7]), ('tile_x', [-1, 1, 1, 7]), ('tile_rc', [-1, 128, 4]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 1, 1]), ('auto_unroll_max_step', 0), ('unroll_explicit', 0)],None,482121
-No: 16 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 16 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2268,7 +2268,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1, 16]), ('tile_y', [-1, 1, 7, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 32, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 512), ('unroll_explicit', 0)],None,2824525
-No: 17 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 17 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2391,7 +2391,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, 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, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 8, 8]), ('tile_ry', [-1, 1, 3]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 0)],None,4559286
-No: 18 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 18 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 571, in __call__
func, arg_info = _build_func_common(measure_input, self.runtime, **kwargs)
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 523, in _build_func_common
@@ -2514,7 +2514,7 @@ Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 854, in verify_pass
raise InstantiationError("Skipped because of invalid gpu kernel")
tvm.autotvm.task.space.InstantiationError: Skipped because of invalid gpu kernel [('tile_f', [-1, 1, 32, 16]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 512]), ('tile_ry', [-1, 3, 1]), ('tile_rx', [-1, 3, 1]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9677544
-No: 19 GFLOPS: 0.00/64.10 result: Traceback (most recent call last):
+No: 19 GFLOPS: 0.00/103.76 result: Traceback (most recent call last):
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 721, in __call__
yield remote, remote.load_module(os.path.split(build_result.filename)[1])
File "/workspace/python/tvm/autotvm/measure/measure_methods.py", line 685, in run_through_rpc
@@ -2602,7 +2602,7 @@ tvm._ffi.base.TVMError: Traceback (most recent call last):
15: _PyEval_EvalFrameDefault
14: 0x0000000000537c30
13: _PyObject_FastCallKeywords
- 12: 0x00007fa3d62befa2
+ 12: 0x00007f23d9b6ffa2
11: _ctypes_callproc
10: ffi_call
9: ffi_call_unix64
@@ -2667,7 +2667,7 @@ Traceback (most recent call last):
21: _PyFunction_FastCallKeywords
20: _PyEval_EvalFrameDefault
19: _PyFunction_FastCall [('tile_f', [-1, 8, 2, 16]), ('tile_y', [-1, 7, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 1, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 0), ('unroll_explicit', 1)],None,6390073
-No: 20 GFLOPS: 144.30/144.30 result: MeasureResult(costs=(0.00160430567,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4274165630340576, timestamp=1650564571.984725) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
+No: 20 GFLOPS: 144.39/144.39 result: MeasureResult(costs=(0.0016032561200000001,), error_no=MeasureErrorNo.NO_ERROR, all_cost=1.4371635913848877, timestamp=1650567480.202208) [('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
</pre></div>
</div>
<p>Finally we can inspect the best config from log file, check correctness,
@@ -2706,7 +2706,7 @@ and measure running time.</p>
<p class="sphx-glr-script-out">Out:</p>
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>Best config:
[('tile_f', [-1, 1, 4, 1]), ('tile_y', [-1, 1, 1, 1]), ('tile_x', [-1, 7, 1, 1]), ('tile_rc', [-1, 4, 1]), ('tile_ry', [-1, 1, 1]), ('tile_rx', [-1, 1, 3]), ('auto_unroll_max_step', 1500), ('unroll_explicit', 1)],None,9881539
-Time cost of this operator: 0.001981
+Time cost of this operator: 0.001958
</pre></div>
</div>
<div class="sphx-glr-footer class 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 dcb301a4a..46030755d 100644
--- a/docs/how_to/work_with_microtvm/micro_autotune.html
+++ b/docs/how_to/work_with_microtvm/micro_autotune.html
@@ -553,10 +553,10 @@ 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
--------- --- -------- ------- ----- ------ -------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 314.2 98.755 (1, 2, 10, 10, 3) 2 1
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.061 0.962 (1, 6, 10, 10) 1 1
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.901 0.283 (1, 1, 10, 10, 3) 1 1
-Total_time - 318.162 - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 321.4 98.767 (1, 2, 10, 10, 3) 2 1
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 3.094 0.951 (1, 6, 10, 10) 1 1
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.917 0.282 (1, 1, 10, 10, 3) 1 1
+Total_time - 325.411 - - - -
</pre></div>
</div>
</div>
@@ -608,10 +608,10 @@ Total_time -
<div class="sphx-glr-script-out highlight-none notranslate"><div class="highlight"><pre><span></span>########## Build with Autotuning ##########
Node Name Ops Time(us) Time(%) Shape Inputs Outputs
--------- --- -------- ------- ----- ------ -------
-tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 79.9 96.727 (1, 6, 10, 10, 1) 2 1
-tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.789 2.166 (1, 6, 10, 10) 1 1
-tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.915 1.107 (1, 1, 10, 10, 3) 1 1
-Total_time - 82.604 - - - -
+tvmgen_default_fused_nn_contrib_conv2d_NCHWc tvmgen_default_fused_nn_contrib_conv2d_NCHWc 124.5 97.883 (1, 6, 10, 10, 1) 2 1
+tvmgen_default_fused_layout_transform_1 tvmgen_default_fused_layout_transform_1 1.77 1.392 (1, 6, 10, 10) 1 1
+tvmgen_default_fused_layout_transform tvmgen_default_fused_layout_transform 0.922 0.725 (1, 1, 10, 10, 3) 1 1
+Total_time - 127.192 - - - -
</pre></div>
</div>
<div class="sphx-glr-footer class sphx-glr-footer-example docutils container" id="sphx-glr-download-how-to-work-with-microtvm-micro-autotune-py">
diff --git a/docs/how_to/work_with_microtvm/sg_execution_times.html b/docs/how_to/work_with_microtvm/sg_execution_times.html
index 67cfa4b88..8c02060bb 100644
--- a/docs/how_to/work_with_microtvm/sg_execution_times.html
+++ b/docs/how_to/work_with_microtvm/sg_execution_times.html
@@ -300,13 +300,13 @@
<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>00:43.796</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
+<p><strong>00:45.418</strong> total execution time for <strong>how_to_work_with_microtvm</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:39.782</strong>: <a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">Autotuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></li>
-<li><p><strong>00:03.440</strong>: <a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">microTVM with TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></li>
-<li><p><strong>00:00.195</strong>: <a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></li>
-<li><p><strong>00:00.191</strong>: <a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">Executing a Tiny Model with TVMC Micro</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></li>
-<li><p><strong>00:00.189</strong>: <a class="reference internal" href="micro_reference_vm.html#sphx-glr-how-to-work-with-microtvm-micro-reference-vm-py"><span class="std std-ref">microTVM Reference Virtual Machines</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_reference_vm.py</span></code>)</p></li>
+<li><p><strong>00:41.261</strong>: <a class="reference internal" href="micro_autotune.html#sphx-glr-how-to-work-with-microtvm-micro-autotune-py"><span class="std std-ref">Autotuning with microTVM</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_autotune.py</span></code>)</p></li>
+<li><p><strong>00:03.559</strong>: <a class="reference internal" href="micro_tflite.html#sphx-glr-how-to-work-with-microtvm-micro-tflite-py"><span class="std std-ref">microTVM with TFLite Models</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tflite.py</span></code>)</p></li>
+<li><p><strong>00:00.202</strong>: <a class="reference internal" href="micro_ethosu.html#sphx-glr-how-to-work-with-microtvm-micro-ethosu-py"><span class="std std-ref">Running TVM on bare metal Arm(R) Cortex(R)-M55 CPU and Ethos(TM)-U55 NPU</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_ethosu.py</span></code>)</p></li>
+<li><p><strong>00:00.198</strong>: <a class="reference internal" href="micro_reference_vm.html#sphx-glr-how-to-work-with-microtvm-micro-reference-vm-py"><span class="std std-ref">microTVM Reference Virtual Machines</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_reference_vm.py</span></code>)</p></li>
+<li><p><strong>00:00.198</strong>: <a class="reference internal" href="micro_tvmc.html#sphx-glr-how-to-work-with-microtvm-micro-tvmc-py"><span class="std std-ref">Executing a Tiny Model with TVMC Micro</span></a> (<code class="docutils literal notranslate"><span class="pre">micro_tvmc.py</span></code>)</p></li>
</ul>
</div>
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 b7434bf7f..abb1358fc 100644
--- a/docs/how_to/work_with_relay/sg_execution_times.html
+++ b/docs/how_to/work_with_relay/sg_execution_times.html
@@ -300,11 +300,11 @@
<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>00:05.545</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
+<p><strong>00:09.219</strong> total execution time for <strong>how_to_work_with_relay</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:03.926</strong>: <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></li>
-<li><p><strong>00:01.410</strong>: <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></li>
-<li><p><strong>00:00.209</strong>: <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></li>
+<li><p><strong>00:07.083</strong>: <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></li>
+<li><p><strong>00:01.922</strong>: <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></li>
+<li><p><strong>00:00.215</strong>: <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></li>
</ul>
</div>
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 f828b9fb2..7946b072e 100644
--- a/docs/how_to/work_with_schedules/sg_execution_times.html
+++ b/docs/how_to/work_with_schedules/sg_execution_times.html
@@ -300,16 +300,16 @@
<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>00:05.373</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
+<p><strong>00:05.598</strong> total execution time for <strong>how_to_work_with_schedules</strong> files:</p>
<ul class="simple">
-<li><p><strong>00:01.951</strong>: <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></li>
-<li><p><strong>00:01.064</strong>: <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></li>
-<li><p><strong>00:00.690</strong>: <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></li>
-<li><p><strong>00:00.686</strong>: <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></li>
-<li><p><strong>00:00.307</strong>: <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></li>
-<li><p><strong>00:00.233</strong>: <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></li>
-<li><p><strong>00:00.229</strong>: <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></li>
-<li><p><strong>00:00.213</strong>: <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></li>
+<li><p><strong>00:02.057</strong>: <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></li>
+<li><p><strong>00:01.109</strong>: <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></li>
+<li><p><strong>00:00.723</strong>: <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></li>
+<li><p><strong>00:00.705</strong>: <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></li>
+<li><p><strong>00:00.308</strong>: <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></li>
+<li><p><strong>00:00.242</strong>: <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></li>
+<li><p><strong>00:00.234</strong>: <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></li>
+<li><p><strong>00:00.220</strong>: <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></li>
</ul>
</div>
diff --git a/docs/how_to/work_with_schedules/tensorize.html b/docs/how_to/work_with_schedules/tensorize.html
index d043c2d0e..6c7484be7 100644
--- a/docs/how_to/work_with_schedules/tensorize.html
+++ b/docs/how_to/work_with_schedules/tensorize.html
@@ -552,7 +552,7 @@ The importing needs to happen before the tensorized GEMV being executed.</p>
C: Buffer(C_2: Pointer(float32), float32, [524288], [])}
buffer_map = {A_1: A, B_1: B, C_1: C}
preflattened_buffer_map = {A_1: A_3: Buffer(A_2, float32, [1024, 64], []), B_1: B_3: Buffer(B_2, float32, [512, 64], []), C_1: C_3: Buffer(C_2, float32, [1024, 512], [])} {
- attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmpsv82ynx1/input0.cc'\nsource_filename = \"/tmp/tmpsv82ynx1/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = allo [...]
+ attr [IterVar(i: int32, (nullptr), "DataPar", "")] "pragma_import_llvm" = "; ModuleID = '/tmp/tmp0hu7m0hl/input0.cc'\nsource_filename = \"/tmp/tmp0hu7m0hl/input0.cc\"\ntarget datalayout = \"e-m:e-i64:64-f80:128-n8:16:32:64-S128\"\ntarget triple = \"x86_64-pc-linux-gnu\"\n\n; Function Attrs: noinline nounwind optnone uwtable\ndefine dso_local i32 @gemv_update(float*, float*, float*, i32, i32, i32) #0 {\n %7 = allo [...]
for (i, 0, 1024) {
for (j.outer: int32, 0, 32) {
@tir.call_extern("gemv_update", @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), C_2, ((i*512) + (j.outer*16)), 16, 2, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), A_2, (i*64), 64, 1, dtype=handle), @tir.tvm_access_ptr(@tir.type_annotation(, dtype=float32), B_2, (j.outer*1024), 1024, 1, dtype=handle), 16, 64, 64, dtype=int32)
diff --git a/docs/objects.inv b/docs/objects.inv
index ed198c243..c52ff02b9 100644
Binary files a/docs/objects.inv and b/docs/objects.inv differ
diff --git a/docs/reference/api/doxygen/array_8h__dep__incl.svg b/docs/reference/api/doxygen/array_8h__dep__incl.svg
index ecbb1dba4..ca6f80c8a 100644
--- a/docs/reference/api/doxygen/array_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/array_8h__dep__incl.svg
@@ -76,48 +76,48 @@
<path fill="none" stroke="#191970" d="M1493.9135,-1014.7252C1371.7601,-1000.6702 1128,-963.146 1128,-887.5 1128,-887.5 1128,-887.5 1128,-764.5 1128,-689.9142 1074.5273,-611.1006 1054.4751,-584.1677"/>
<polygon fill="#191970" stroke="#191970" points="1493.5895,-1018.2108 1503.9192,-1015.8553 1494.3752,-1011.255 1493.5895,-1018.2108"/>
</g>
-<!-- Node153 -->
+<!-- Node154 -->
<g id="node34" class="node">
-<title>Node153</title>
+<title>Node154</title>
<g id="a_node34"><a xlink:href="ir_2type_8h.html" target="_top" xlink:title="IR/AST nodes for the unified type system in TVM. ">
<polygon fill="#ffffff" stroke="#000000" points="655,-755 655,-774 773,-774 773,-755 655,-755"/>
<text text-anchor="middle" x="714" y="-762" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/type.h</text>
</a>
</g>
</g>
-<!-- Node19->Node153 -->
+<!-- Node19->Node154 -->
<g id="edge71" class="edge">
-<title>Node19->Node153</title>
+<title>Node19->Node154</title>
<path fill="none" stroke="#191970" d="M1493.7806,-1016.6683C1386.4684,-1008.5613 1183.3117,-991.0252 1114,-970 952.2501,-920.9345 777.9995,-807.8823 728.0777,-774.1632"/>
<polygon fill="#191970" stroke="#191970" points="1493.6252,-1020.1663 1503.8587,-1017.4232 1494.1481,-1013.1859 1493.6252,-1020.1663"/>
</g>
-<!-- Node175 -->
+<!-- Node176 -->
<g id="node39" class="node">
-<title>Node175</title>
+<title>Node176</title>
<g id="a_node39"><a xlink:href="buffer_8h.html" target="_top" xlink:title="Symbolic n-dimensional array, to represent a memory buffer. ">
<polygon fill="#ffffff" stroke="#ff0000" points="2065,-565 2065,-584 2193,-584 2193,-565 2065,-565"/>
<text text-anchor="middle" x="2129" y="-572" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/buffer.h</text>
</a>
</g>
</g>
-<!-- Node19->Node175 -->
+<!-- Node19->Node176 -->
<g id="edge141" class="edge">
-<title>Node19->Node175</title>
+<title>Node19->Node176</title>
<path fill="none" stroke="#191970" d="M1630.4994,-1018.4843C1840.3197,-1008.3444 2460,-971.3034 2460,-887.5 2460,-887.5 2460,-887.5 2460,-703 2460,-644.4384 2277.6712,-602.021 2183.6579,-584.0419"/>
<polygon fill="#191970" stroke="#191970" points="1630.0499,-1015.0017 1620.2284,-1018.9749 1630.384,-1021.9937 1630.0499,-1015.0017"/>
</g>
-<!-- Node176 -->
+<!-- Node177 -->
<g id="node40" class="node">
-<title>Node176</title>
+<title>Node177</title>
<g id="a_node40"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
<polygon fill="#ffffff" stroke="#ff0000" points="2023.5,-447.5 2023.5,-466.5 2144.5,-466.5 2144.5,-447.5 2023.5,-447.5"/>
<text text-anchor="middle" x="2084" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
</a>
</g>
</g>
-<!-- Node19->Node176 -->
+<!-- Node19->Node177 -->
<g id="edge142" class="edge">
-<title>Node19->Node176</title>
+<title>Node19->Node177</title>
<path fill="none" stroke="#191970" d="M1630.1904,-1019.5326C1788.7778,-1014.5127 2178.5565,-999.4928 2306,-970 2396.4861,-949.0599 2498,-980.3775 2498,-887.5 2498,-887.5 2498,-887.5 2498,-574.5 2498,-501.9343 2259.635,-471.8408 2144.6425,-461.5451"/>
<polygon fill="#191970" stroke="#191970" points="1629.9697,-1016.0377 1620.0841,-1019.8488 1630.1887,-1023.0343 1629.9697,-1016.0377"/>
</g>
@@ -844,57 +844,57 @@
<path fill="none" stroke="#191970" d="M2167.1109,-381.8763C2192.5369,-370.742 2228.4418,-355.019 2254.6966,-343.5218"/>
<polygon fill="#191970" stroke="#191970" points="2165.6774,-378.6831 2157.9212,-385.9005 2168.4854,-385.0952 2165.6774,-378.6831"/>
</g>
-<!-- Node153->Node20 -->
+<!-- Node154->Node20 -->
<g id="edge72" class="edge">
-<title>Node153->Node20</title>
+<title>Node154->Node20</title>
<path fill="none" stroke="#191970" d="M714,-744.7305C714,-717.9497 714,-671.186 714,-651.1068"/>
<polygon fill="#191970" stroke="#191970" points="710.5001,-744.8484 714,-754.8484 717.5001,-744.8484 710.5001,-744.8484"/>
</g>
-<!-- Node153->Node21 -->
+<!-- Node154->Node21 -->
<g id="edge107" class="edge">
-<title>Node153->Node21</title>
+<title>Node154->Node21</title>
<path fill="none" stroke="#191970" d="M723.7821,-745.6465C739.7518,-716.4772 773.9988,-660.2118 817,-626 902.7214,-557.8001 1030.2622,-530.4238 1107.4339,-519.6771"/>
<polygon fill="#191970" stroke="#191970" points="720.5413,-744.2847 718.9113,-754.7534 726.7139,-747.5861 720.5413,-744.2847"/>
</g>
-<!-- Node153->Node109 -->
+<!-- Node154->Node109 -->
<g id="edge110" class="edge">
-<title>Node153->Node109</title>
+<title>Node154->Node109</title>
<path fill="none" stroke="#191970" d="M648.0216,-752.1171C623.2557,-744.5113 596.6257,-732.35 578,-713 554.754,-688.8499 554,-675.0202 554,-641.5 554,-641.5 554,-641.5 554,-513 554,-476.2161 557.4607,-433.0694 559.5155,-410.6929"/>
<polygon fill="#191970" stroke="#191970" points="647.1851,-755.5182 657.7625,-754.9113 649.1153,-748.7896 647.1851,-755.5182"/>
</g>
-<!-- Node153->Node110 -->
+<!-- Node154->Node110 -->
<g id="edge108" class="edge">
-<title>Node153->Node110</title>
+<title>Node154->Node110</title>
<path fill="none" stroke="#191970" d="M697.323,-747.3288C674.7779,-721.9957 638.5241,-672.5638 649,-626 676.3492,-504.4371 759.3667,-374.1763 783.4099,-338.1836"/>
<polygon fill="#191970" stroke="#191970" points="695.0043,-749.9774 704.3559,-754.9573 700.1509,-745.2326 695.0043,-749.9774"/>
</g>
-<!-- Node153->Node115 -->
+<!-- Node154->Node115 -->
<g id="edge114" class="edge">
-<title>Node153->Node115</title>
+<title>Node154->Node115</title>
<path fill="none" stroke="#191970" d="M783.0995,-760.6537C961.015,-748.8722 1421,-705.7234 1421,-574.5 1421,-574.5 1421,-574.5 1421,-513 1421,-461.0314 1346.8856,-310.7123 1326.9094,-271.1404"/>
<polygon fill="#191970" stroke="#191970" points="782.8148,-757.1647 773.0624,-761.3048 783.268,-764.15 782.8148,-757.1647"/>
</g>
-<!-- Node153->Node136 -->
+<!-- Node154->Node136 -->
<g id="edge111" class="edge">
-<title>Node153->Node136</title>
+<title>Node154->Node136</title>
<path fill="none" stroke="#191970" d="M673.2264,-751.271C653.5382,-743.038 630.9923,-730.6057 616,-713 594.2676,-687.4793 592,-675.0202 592,-641.5 592,-641.5 592,-641.5 592,-574.5 592,-500.4546 609.8401,-483.6788 624,-411 629.1337,-384.6502 635.1533,-353.5586 638.1333,-338.1539"/>
<polygon fill="#191970" stroke="#191970" points="672.0297,-754.5614 682.6163,-754.9796 674.6012,-748.0508 672.0297,-754.5614"/>
</g>
-<!-- Node153->Node137 -->
+<!-- Node154->Node137 -->
<g id="edge112" class="edge">
-<title>Node153->Node137</title>
+<title>Node154->Node137</title>
<path fill="none" stroke="#191970" d="M644.8023,-761.7168C511.5647,-755.8143 224.1564,-740.2958 129,-713 65.9901,-694.9255 0,-707.051 0,-641.5 0,-641.5 0,-641.5 0,-457 0,-392.9291 79.5566,-354.5768 124.3073,-338.0718"/>
<polygon fill="#191970" stroke="#191970" points="644.7521,-765.2179 654.8957,-762.1586 645.0583,-758.2246 644.7521,-765.2179"/>
</g>
-<!-- Node153->Node138 -->
+<!-- Node154->Node138 -->
<g id="edge113" class="edge">
-<title>Node153->Node138</title>
+<title>Node154->Node138</title>
<path fill="none" stroke="#191970" d="M644.8335,-760.5571C517.4279,-751.6945 253.5762,-725.3479 198,-657 94.9546,-530.2743 371.699,-377.2256 448.7295,-338.0104"/>
<polygon fill="#191970" stroke="#191970" points="644.6632,-764.0535 654.8774,-761.2391 645.1375,-757.0696 644.6632,-764.0535"/>
</g>
-<!-- Node154 -->
+<!-- Node155 -->
<g id="node35" class="node">
-<title>Node154</title>
+<title>Node155</title>
<g id="a_node35"><a xlink:href="affine__type_8h.html" target="_top" xlink:title="Quantized Tensor Types. ">
<polygon fill="#ffffff" stroke="#000000" points="826.5,-626.5 826.5,-656.5 941.5,-656.5 941.5,-626.5 826.5,-626.5"/>
<text text-anchor="start" x="834.5" y="-644.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/affine</text>
@@ -902,30 +902,30 @@
</a>
</g>
</g>
-<!-- Node153->Node154 -->
+<!-- Node154->Node155 -->
<g id="edge73" class="edge">
-<title>Node153->Node154</title>
+<title>Node154->Node155</title>
<path fill="none" stroke="#191970" d="M746.1503,-750.5912C765.9947,-741.2981 791.4897,-728.045 812,-713 835.0114,-696.1203 857.7315,-671.9717 871.3141,-656.5134"/>
<polygon fill="#191970" stroke="#191970" points="744.5422,-747.478 736.9091,-754.8256 747.4581,-753.8418 744.5422,-747.478"/>
</g>
-<!-- Node155 -->
+<!-- Node156 -->
<g id="node36" class="node">
-<title>Node155</title>
+<title>Node156</title>
<g id="a_node36"><a xlink:href="ir_2expr_8h.html" target="_top" xlink:title="Base expr nodes in TVM. ">
<polygon fill="#ffffff" stroke="#ff0000" points="944.5,-693.5 944.5,-712.5 1061.5,-712.5 1061.5,-693.5 944.5,-693.5"/>
<text text-anchor="middle" x="1003" y="-700.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/expr.h</text>
</a>
</g>
</g>
-<!-- Node153->Node155 -->
+<!-- Node154->Node156 -->
<g id="edge74" class="edge">
-<title>Node153->Node155</title>
+<title>Node154->Node156</title>
<path fill="none" stroke="#191970" d="M768.7317,-752.8529C823.568,-741.1836 906.9662,-723.4363 958.0556,-712.5643"/>
<polygon fill="#191970" stroke="#191970" points="767.8916,-749.4533 758.8392,-754.9581 769.3487,-756.2999 767.8916,-749.4533"/>
</g>
-<!-- Node172 -->
+<!-- Node173 -->
<g id="node38" class="node">
-<title>Node172</title>
+<title>Node173</title>
<g id="a_node38"><a xlink:href="tensor__type_8h.html" target="_top" xlink:title="Polymorphic tensor types. ">
<polygon fill="#ffffff" stroke="#000000" points="28,-626.5 28,-656.5 146,-656.5 146,-626.5 28,-626.5"/>
<text text-anchor="start" x="36" y="-644.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/tensor</text>
@@ -933,216 +933,216 @@
</a>
</g>
</g>
-<!-- Node153->Node172 -->
+<!-- Node154->Node173 -->
<g id="edge109" class="edge">
-<title>Node153->Node172</title>
+<title>Node154->Node173</title>
<path fill="none" stroke="#191970" d="M644.7617,-760.6597C505.9138,-752.5824 203.0202,-732.9311 160,-713 132.958,-700.4715 109.9702,-673.5942 97.3614,-656.5965"/>
<polygon fill="#191970" stroke="#191970" points="644.6855,-764.1611 654.8709,-761.2442 645.0897,-757.1727 644.6855,-764.1611"/>
</g>
-<!-- Node155->Node20 -->
+<!-- Node156->Node20 -->
<g id="edge75" class="edge">
-<title>Node155->Node20</title>
+<title>Node156->Node20</title>
<path fill="none" stroke="#191970" d="M948.2683,-691.3529C893.432,-679.6836 810.0338,-661.9363 758.9444,-651.0643"/>
<polygon fill="#191970" stroke="#191970" points="947.6513,-694.7999 958.1608,-693.4581 949.1084,-687.9533 947.6513,-694.7999"/>
</g>
-<!-- Node155->Node21 -->
+<!-- Node156->Node21 -->
<g id="edge87" class="edge">
-<title>Node155->Node21</title>
+<title>Node156->Node21</title>
<path fill="none" stroke="#191970" d="M1032.0876,-688.5259C1046.5617,-680.4676 1063.7483,-669.5352 1077,-657 1121.891,-614.5363 1157.0876,-547.7408 1169.3296,-522.8122"/>
<polygon fill="#191970" stroke="#191970" points="1030.1853,-685.5749 1023.0363,-693.3943 1033.5011,-691.7398 1030.1853,-685.5749"/>
</g>
-<!-- Node155->Node30 -->
+<!-- Node156->Node30 -->
<g id="edge94" class="edge">
-<title>Node155->Node30</title>
+<title>Node156->Node30</title>
<path fill="none" stroke="#191970" d="M1068.3018,-691.6305C1120.1985,-682.0656 1188.2063,-668.1973 1214,-657 1262.0621,-636.1358 1273.1643,-626.2456 1311,-590 1409.979,-495.1806 1439.0054,-468.1941 1497,-344 1507.3232,-321.8931 1512.2475,-293.7672 1514.4339,-276.7235"/>
<polygon fill="#191970" stroke="#191970" points="1067.3628,-688.2441 1058.1546,-693.4841 1068.6207,-695.1301 1067.3628,-688.2441"/>
</g>
-<!-- Node155->Node77 -->
+<!-- Node156->Node77 -->
<g id="edge92" class="edge">
-<title>Node155->Node77</title>
+<title>Node156->Node77</title>
<path fill="none" stroke="#191970" d="M934.1933,-700.3372C774.565,-693.8451 391.0488,-676.3122 371,-657 238.9649,-529.8158 284.7261,-383.4836 406,-246 437.1327,-210.706 445.6583,-199.545 488,-179 539.9671,-153.7846 604.4799,-140.7463 652.2747,-134.1077"/>
<polygon fill="#191970" stroke="#191970" points="934.2446,-703.842 944.3778,-700.7487 934.5273,-696.8477 934.2446,-703.842"/>
</g>
-<!-- Node155->Node110 -->
+<!-- Node156->Node110 -->
<g id="edge88" class="edge">
-<title>Node155->Node110</title>
+<title>Node156->Node110</title>
<path fill="none" stroke="#191970" d="M1014.8355,-684.6917C1023.7759,-668.3331 1032.816,-644.2083 1022,-626 1014.052,-612.6199 913.1195,-565.5449 899,-559 858.788,-540.3602 832.5318,-558.5036 806,-523 762.7995,-465.1912 780.4109,-368.9322 787.5461,-338.2261"/>
<polygon fill="#191970" stroke="#191970" points="1011.8094,-682.933 1009.7338,-693.3226 1017.8354,-686.495 1011.8094,-682.933"/>
</g>
-<!-- Node155->Node115 -->
+<!-- Node156->Node115 -->
<g id="edge106" class="edge">
-<title>Node155->Node115</title>
+<title>Node156->Node115</title>
<path fill="none" stroke="#191970" d="M1050.8196,-690.7882C1124.3181,-670.9679 1259.5643,-629.974 1288,-590 1359.429,-489.5874 1272.6868,-433.1615 1300,-313 1303.4911,-297.6412 1311.3737,-281.0976 1316.6918,-271.0192"/>
<polygon fill="#191970" stroke="#191970" points="1049.6352,-687.4817 1040.876,-693.4421 1051.4404,-694.2449 1049.6352,-687.4817"/>
</g>
-<!-- Node155->Node136 -->
+<!-- Node156->Node136 -->
<g id="edge91" class="edge">
-<title>Node155->Node136</title>
+<title>Node156->Node136</title>
<path fill="none" stroke="#191970" d="M1002.8441,-682.7919C1001.6833,-665.8071 997.5011,-641.8338 984,-626 957.3809,-594.7817 935.429,-608.8702 899,-590 848.8433,-564.0188 832.9808,-560.7932 791,-523 723.6452,-462.3639 664.2924,-368.77 645.8795,-338.3883"/>
<polygon fill="#191970" stroke="#191970" points="999.3642,-683.3887 1003.2423,-693.2482 1006.3591,-683.1222 999.3642,-683.3887"/>
</g>
-<!-- Node155->Node138 -->
+<!-- Node156->Node138 -->
<g id="edge93" class="edge">
-<title>Node155->Node138</title>
+<title>Node156->Node138</title>
<path fill="none" stroke="#191970" d="M994.5679,-684.2107C985.8071,-666.7314 970.5038,-641.168 950,-626 907.2847,-594.4007 885.0242,-610.4875 836,-590 678.7949,-524.3031 608.2539,-540.043 497,-411 478.2636,-389.2677 471.3859,-354.6571 469.0693,-338.0292"/>
<polygon fill="#191970" stroke="#191970" points="991.4987,-685.9063 998.9599,-693.4284 997.8181,-682.8952 991.4987,-685.9063"/>
</g>
-<!-- Node155->Node149 -->
+<!-- Node156->Node149 -->
<g id="edge86" class="edge">
-<title>Node155->Node149</title>
+<title>Node156->Node149</title>
<path fill="none" stroke="#191970" d="M1030.9715,-687.9168C1042.1701,-680.2694 1053.8552,-669.8551 1060,-657 1071.631,-632.6676 1059.4981,-600.142 1051.9933,-584.1739"/>
<polygon fill="#191970" stroke="#191970" points="1028.7959,-685.1512 1022.21,-693.4505 1032.5339,-691.0696 1028.7959,-685.1512"/>
</g>
-<!-- Node155->Node154 -->
+<!-- Node156->Node155 -->
<g id="edge76" class="edge">
-<title>Node155->Node154</title>
+<title>Node156->Node155</title>
<path fill="none" stroke="#191970" d="M975.504,-688.7899C956.816,-679.1318 932.1315,-666.3747 913.1107,-656.5446"/>
<polygon fill="#191970" stroke="#191970" points="973.9156,-691.9087 984.4063,-693.3906 977.1294,-685.6901 973.9156,-691.9087"/>
</g>
-<!-- Node160 -->
+<!-- Node161 -->
<g id="node37" class="node">
-<title>Node160</title>
+<title>Node161</title>
<g id="a_node37"><a xlink:href="ir_2attrs_8h.html" target="_top" xlink:title="Helpers for attribute objects. ">
<polygon fill="#ffffff" stroke="#ff0000" points="380.5,-632 380.5,-651 499.5,-651 499.5,-632 380.5,-632"/>
<text text-anchor="middle" x="440" y="-639" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/attrs.h</text>
</a>
</g>
</g>
-<!-- Node155->Node160 -->
+<!-- Node156->Node161 -->
<g id="edge77" class="edge">
-<title>Node155->Node160</title>
+<title>Node156->Node161</title>
<path fill="none" stroke="#191970" d="M934.234,-695.4883C822.9061,-683.3272 605.0807,-659.5328 499.7974,-648.032"/>
<polygon fill="#191970" stroke="#191970" points="934.0214,-698.9858 944.3424,-696.5925 934.7816,-692.0272 934.0214,-698.9858"/>
</g>
-<!-- Node155->Node172 -->
+<!-- Node156->Node173 -->
<g id="edge89" class="edge">
-<title>Node155->Node172</title>
+<title>Node156->Node173</title>
<path fill="none" stroke="#191970" d="M934.1877,-700.9628C794.9284,-696.4219 469.8265,-683.7004 198,-657 181.1528,-655.3452 162.9366,-653.025 146.2277,-650.6827"/>
<polygon fill="#191970" stroke="#191970" points="934.1821,-704.4643 944.2897,-701.2884 934.4077,-697.4679 934.1821,-704.4643"/>
</g>
-<!-- Node155->Node175 -->
+<!-- Node156->Node176 -->
<g id="edge95" class="edge">
-<title>Node155->Node175</title>
+<title>Node156->Node176</title>
<path fill="none" stroke="#191970" d="M1071.9482,-694.887C1143.7437,-686.2524 1259.4164,-671.817 1359,-657 1441.0273,-644.7952 1460.7406,-636.5273 1543,-626 1731.9283,-601.8216 1956.5929,-585.539 2064.654,-578.4801"/>
<polygon fill="#191970" stroke="#191970" points="1071.1931,-691.4524 1061.681,-696.1182 1072.0266,-698.4026 1071.1931,-691.4524"/>
</g>
-<!-- Node155->Node176 -->
+<!-- Node156->Node177 -->
<g id="edge104" class="edge">
-<title>Node155->Node176</title>
+<title>Node156->Node177</title>
<path fill="none" stroke="#191970" d="M1071.7743,-696.0649C1129.0379,-689.2569 1212.3337,-676.9432 1283,-657 1318.1083,-647.0919 1324.2387,-637.0643 1359,-626 1605.0012,-547.6995 1908.5053,-488.6989 2030.0043,-466.5512"/>
<polygon fill="#191970" stroke="#191970" points="1071.2778,-692.5989 1061.7497,-697.232 1072.0874,-699.552 1071.2778,-692.5989"/>
</g>
-<!-- Node155->Node182 -->
+<!-- Node156->Node182 -->
<g id="edge105" class="edge">
-<title>Node155->Node182</title>
+<title>Node156->Node182</title>
<path fill="none" stroke="#191970" d="M1072.0492,-699.7925C1174.2785,-694.4275 1371.734,-681.703 1538,-657 1542.4854,-656.3336 1547.1089,-655.5597 1551.7488,-654.7201"/>
<polygon fill="#191970" stroke="#191970" points="1071.6645,-696.3077 1061.8586,-700.3194 1072.026,-703.2983 1071.6645,-696.3077"/>
</g>
-<!-- Node160->Node77 -->
+<!-- Node161->Node77 -->
<g id="edge83" class="edge">
-<title>Node160->Node77</title>
+<title>Node161->Node77</title>
<path fill="none" stroke="#191970" d="M424.9773,-623.818C417.7668,-614.3756 409.6938,-602.2057 405,-590 360.672,-474.7308 320.1622,-419.3173 383,-313 446.7994,-205.0559 598.2813,-155.6684 674.5173,-137.0071"/>
<polygon fill="#191970" stroke="#191970" points="422.2569,-626.0204 431.235,-631.6457 427.7245,-621.6494 422.2569,-626.0204"/>
</g>
-<!-- Node160->Node79 -->
+<!-- Node161->Node79 -->
<g id="edge81" class="edge">
-<title>Node160->Node79</title>
+<title>Node161->Node79</title>
<path fill="none" stroke="#191970" d="M421.0079,-624.8304C410.9327,-615.3193 398.8444,-602.773 390,-590 299.6527,-459.5202 250,-420.206 250,-261.5 250,-261.5 250,-261.5 250,-194.5 250,-149.5161 262.6895,-96.8203 268.3403,-75.5843"/>
<polygon fill="#191970" stroke="#191970" points="418.8205,-627.5734 428.5559,-631.7536 423.5522,-622.4148 418.8205,-627.5734"/>
</g>
-<!-- Node160->Node109 -->
+<!-- Node161->Node109 -->
<g id="edge80" class="edge">
-<title>Node160->Node109</title>
+<title>Node161->Node109</title>
<path fill="none" stroke="#191970" d="M449.2002,-622.7955C472.2978,-575.8368 532.3898,-453.6661 553.5194,-410.7085"/>
<polygon fill="#191970" stroke="#191970" points="446.0164,-621.3384 444.7433,-631.8565 452.2977,-624.428 446.0164,-621.3384"/>
</g>
-<!-- Node160->Node110 -->
+<!-- Node161->Node110 -->
<g id="edge79" class="edge">
-<title>Node160->Node110</title>
+<title>Node161->Node110</title>
<path fill="none" stroke="#191970" d="M459.6178,-625.2921C496.3174,-594.846 577.6437,-526.7824 644,-467 696.4332,-419.7613 757.0227,-360.8383 780.2509,-338.0821"/>
<polygon fill="#191970" stroke="#191970" points="457.1507,-622.7909 451.6825,-631.8656 461.6163,-628.1815 457.1507,-622.7909"/>
</g>
-<!-- Node160->Node136 -->
+<!-- Node161->Node136 -->
<g id="edge82" class="edge">
-<title>Node160->Node136</title>
+<title>Node161->Node136</title>
<path fill="none" stroke="#191970" d="M438.312,-621.4798C435.408,-573.5313 434.8806,-450.8453 497,-380 516.3974,-357.8779 546.0251,-345.2566 573.2918,-338.0566"/>
<polygon fill="#191970" stroke="#191970" points="434.8416,-622.0196 439.045,-631.7449 441.8238,-621.521 434.8416,-622.0196"/>
</g>
-<!-- Node160->Node137 -->
+<!-- Node161->Node137 -->
<g id="edge84" class="edge">
-<title>Node160->Node137</title>
+<title>Node161->Node137</title>
<path fill="none" stroke="#191970" d="M370.5874,-632.5222C338.3241,-625.4135 301.1778,-612.7028 274,-590 189.4698,-519.3883 161.5298,-376.4131 155.4062,-338.1176"/>
<polygon fill="#191970" stroke="#191970" points="369.9257,-635.9594 380.4269,-634.5534 371.341,-629.1039 369.9257,-635.9594"/>
</g>
-<!-- Node160->Node138 -->
+<!-- Node161->Node138 -->
<g id="edge85" class="edge">
-<title>Node160->Node138</title>
+<title>Node161->Node138</title>
<path fill="none" stroke="#191970" d="M430.6555,-622.5346C419.404,-597.9097 402,-553.305 402,-513 402,-513 402,-513 402,-457 402,-421.154 409.0306,-411.575 426,-380 434.6295,-363.9431 448.6056,-348.089 458.0415,-338.3038"/>
<polygon fill="#191970" stroke="#191970" points="427.5944,-624.2489 435.0322,-631.7942 433.9231,-621.2575 427.5944,-624.2489"/>
</g>
-<!-- Node160->Node149 -->
+<!-- Node161->Node149 -->
<g id="edge78" class="edge">
-<title>Node160->Node149</title>
+<title>Node161->Node149</title>
<path fill="none" stroke="#191970" d="M509.529,-633.8255C626.779,-620.8835 862.3949,-594.8765 978.7665,-582.0315"/>
<polygon fill="#191970" stroke="#191970" points="509.1402,-630.347 499.5845,-634.9231 509.9082,-637.3048 509.1402,-630.347"/>
</g>
-<!-- Node172->Node137 -->
+<!-- Node173->Node137 -->
<g id="edge90" class="edge">
-<title>Node172->Node137</title>
+<title>Node173->Node137</title>
<path fill="none" stroke="#191970" d="M92.683,-616.1079C97.9578,-590.2919 105,-549.0711 105,-513 105,-513 105,-513 105,-457 105,-409.7256 134.3121,-358.7094 147.5941,-338.0193"/>
<polygon fill="#191970" stroke="#191970" points="89.2001,-615.6612 90.5562,-626.1688 96.0488,-617.1089 89.2001,-615.6612"/>
</g>
-<!-- Node175->Node150 -->
+<!-- Node176->Node150 -->
<g id="edge103" class="edge">
-<title>Node175->Node150</title>
+<title>Node176->Node150</title>
<path fill="none" stroke="#191970" d="M2137.0526,-555.5999C2146.638,-530.7559 2160.5723,-485.7235 2154,-447 2151.4326,-431.8727 2144.9064,-415.2677 2140.4599,-405.1129"/>
<polygon fill="#191970" stroke="#191970" points="2133.7824,-554.3509 2133.2834,-564.934 2140.2732,-556.9721 2133.7824,-554.3509"/>
</g>
-<!-- Node175->Node176 -->
+<!-- Node176->Node177 -->
<g id="edge96" class="edge">
-<title>Node175->Node176</title>
+<title>Node176->Node177</title>
<path fill="none" stroke="#191970" d="M2121.6279,-555.2507C2111.9047,-529.8622 2095.1597,-486.1393 2087.7334,-466.7484"/>
<polygon fill="#191970" stroke="#191970" points="2118.4724,-556.7977 2125.3175,-564.8845 2125.0094,-554.2941 2118.4724,-556.7977"/>
</g>
-<!-- Node176->Node54 -->
+<!-- Node177->Node54 -->
<g id="edge98" class="edge">
-<title>Node176->Node54</title>
+<title>Node177->Node54</title>
<path fill="none" stroke="#191970" d="M2154.4075,-445.5141C2220.1789,-431.5712 2317.6231,-402.279 2382,-344 2424.1037,-305.8844 2447.1878,-239.6829 2455.8996,-209.8523"/>
<polygon fill="#191970" stroke="#191970" points="2153.6286,-442.1009 2144.5376,-447.5419 2155.0374,-448.9577 2153.6286,-442.1009"/>
</g>
-<!-- Node176->Node92 -->
+<!-- Node177->Node92 -->
<g id="edge99" class="edge">
-<title>Node176->Node92</title>
+<title>Node177->Node92</title>
<path fill="none" stroke="#191970" d="M2068.3978,-439.7166C2043.7003,-412.3577 1996.3979,-359.9581 1976.82,-338.2705"/>
<polygon fill="#191970" stroke="#191970" points="2065.9823,-442.264 2075.2811,-447.3416 2071.1783,-437.5734 2065.9823,-442.264"/>
</g>
-<!-- Node176->Node108 -->
+<!-- Node177->Node108 -->
<g id="edge102" class="edge">
-<title>Node176->Node108</title>
+<title>Node177->Node108</title>
<path fill="none" stroke="#191970" d="M2069.3765,-438.7107C2058.4887,-422.7217 2047.082,-399.1347 2057,-380 2067.2779,-360.1709 2089.0909,-346.3546 2106.2099,-338.0608"/>
<polygon fill="#191970" stroke="#191970" points="2066.8294,-441.1565 2075.562,-447.1558 2072.4767,-437.0202 2066.8294,-441.1565"/>
</g>
-<!-- Node176->Node115 -->
+<!-- Node177->Node115 -->
<g id="edge101" class="edge">
-<title>Node176->Node115</title>
+<title>Node177->Node115</title>
<path fill="none" stroke="#191970" d="M2013.2181,-452.9313C1910.4634,-446.4062 1725.3384,-432.1489 1662,-411 1581.098,-383.9865 1575.1871,-347.0829 1497,-313 1453.7426,-294.1435 1401.5565,-279.791 1365.4891,-271.0874"/>
<polygon fill="#191970" stroke="#191970" points="2013.0022,-456.4245 2023.2017,-453.5574 2013.4404,-449.4383 2013.0022,-456.4245"/>
</g>
-<!-- Node176->Node137 -->
+<!-- Node177->Node137 -->
<g id="edge97" class="edge">
-<title>Node176->Node137</title>
+<title>Node177->Node137</title>
<path fill="none" stroke="#191970" d="M2013.3211,-452.3364C1780.363,-436.9585 1015.5058,-386.4149 383,-344 328.9283,-340.374 267.6779,-336.2261 222.1366,-333.1342"/>
<polygon fill="#191970" stroke="#191970" points="2013.147,-455.8325 2023.3559,-452.9988 2013.6082,-448.8477 2013.147,-455.8325"/>
</g>
-<!-- Node176->Node150 -->
+<!-- Node177->Node150 -->
<g id="edge100" class="edge">
-<title>Node176->Node150</title>
+<title>Node177->Node150</title>
<path fill="none" stroke="#191970" d="M2098.7338,-439.5744C2108.1384,-428.4517 2119.9811,-414.4454 2127.7604,-405.2449"/>
<polygon fill="#191970" stroke="#191970" points="2095.909,-437.4945 2092.125,-447.3906 2101.2544,-442.0142 2095.909,-437.4945"/>
</g>
-<!-- Node186->Node160 -->
+<!-- Node186->Node161 -->
<g id="edge116" class="edge">
-<title>Node186->Node160</title>
+<title>Node186->Node161</title>
<path fill="none" stroke="#191970" d="M1329.1524,-948.9452C1170.6619,-936.1456 824.0227,-896.1642 558,-780 510.9832,-759.4692 494.5438,-754.2211 464,-713 449.8124,-693.8528 443.6674,-665.7921 441.3003,-651.1694"/>
<polygon fill="#191970" stroke="#191970" points="1329.072,-952.4497 1339.317,-949.7495 1329.6243,-945.4715 1329.072,-952.4497"/>
</g>
@@ -1194,21 +1194,21 @@
<path fill="none" stroke="#191970" d="M1473.8761,-809.8772C1492.6344,-794.0134 1522.3709,-769.2894 1549,-749 1605.3022,-706.102 1644.3455,-718.719 1679,-657 1724.7826,-575.462 1726.6475,-533.8565 1692,-447 1687.6926,-436.2018 1572.5659,-318.8456 1530.8846,-276.5686"/>
<polygon fill="#191970" stroke="#191970" points="1471.4391,-807.3549 1466.0805,-816.4947 1475.9692,-812.6914 1471.4391,-807.3549"/>
</g>
-<!-- Node187->Node153 -->
+<!-- Node187->Node154 -->
<g id="edge121" class="edge">
-<title>Node187->Node153</title>
+<title>Node187->Node154</title>
<path fill="none" stroke="#191970" d="M1376.2645,-819.619C1267.8808,-810.8155 1066.6257,-794.3975 895,-780 854.3004,-776.5858 808.4606,-772.6594 773.0958,-769.6128"/>
<polygon fill="#191970" stroke="#191970" points="1376.2087,-823.1259 1386.4592,-820.4469 1376.7753,-816.1489 1376.2087,-823.1259"/>
</g>
-<!-- Node187->Node155 -->
+<!-- Node187->Node156 -->
<g id="edge120" class="edge">
-<title>Node187->Node155</title>
+<title>Node187->Node156</title>
<path fill="none" stroke="#191970" d="M1410.0022,-813.755C1319.9439,-789.248 1119.2348,-734.6303 1038.0692,-712.5432"/>
<polygon fill="#191970" stroke="#191970" points="1409.3403,-817.2021 1419.9084,-816.4507 1411.1784,-810.4477 1409.3403,-817.2021"/>
</g>
-<!-- Node187->Node176 -->
+<!-- Node187->Node177 -->
<g id="edge124" class="edge">
-<title>Node187->Node176</title>
+<title>Node187->Node177</title>
<path fill="none" stroke="#191970" d="M1484.5412,-812.1205C1543.4552,-784.4018 1671.5865,-723.9091 1691,-713 1845.3988,-626.2386 2021.1219,-502.1138 2070.484,-466.7416"/>
<polygon fill="#191970" stroke="#191970" points="1482.9358,-809.0077 1475.3762,-816.4308 1485.9149,-815.3421 1482.9358,-809.0077"/>
</g>
@@ -1252,9 +1252,9 @@
<path fill="none" stroke="#191970" d="M1716.2069,-951.3366C1850.5495,-944.6558 2144.8747,-927.668 2243,-903 2326.9906,-881.8853 2422,-912.604 2422,-826 2422,-826 2422,-826 2422,-703 2422,-604.5217 2536,-611.4783 2536,-513 2536,-513 2536,-513 2536,-328.5 2536,-280.0913 2498.0081,-233.1499 2475.7472,-209.8"/>
<polygon fill="#191970" stroke="#191970" points="1715.8548,-947.8496 1706.0393,-951.8377 1716.1995,-954.8411 1715.8548,-947.8496"/>
</g>
-<!-- Node196->Node160 -->
+<!-- Node196->Node161 -->
<g id="edge131" class="edge">
-<title>Node196->Node160</title>
+<title>Node196->Node161</title>
<path fill="none" stroke="#191970" d="M1579.6034,-946.246C1382.8699,-921.9307 818.5926,-848.0059 646,-780 563.0028,-747.297 479.5399,-676.9263 450.63,-651.1856"/>
<polygon fill="#191970" stroke="#191970" points="1579.4532,-949.7539 1589.8062,-947.5032 1580.3094,-942.8064 1579.4532,-949.7539"/>
</g>
diff --git a/docs/reference/api/doxygen/c__runtime__api_8h__dep__incl.svg b/docs/reference/api/doxygen/c__runtime__api_8h__dep__incl.svg
index af4fcaf82..c2fde2664 100644
--- a/docs/reference/api/doxygen/c__runtime__api_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/c__runtime__api_8h__dep__incl.svg
@@ -47,18 +47,18 @@
<path fill="none" stroke="#191970" d="M1093.1682,-721.5837C1145.8097,-693.3486 1233.7718,-636.0661 1264,-556 1285.2411,-499.7382 1220.661,-439.1829 1192.4712,-416.1395"/>
<polygon fill="#191970" stroke="#191970" points="1091.3131,-718.6046 1084.0883,-726.354 1094.5688,-724.8014 1091.3131,-718.6046"/>
</g>
-<!-- Node161 -->
+<!-- Node162 -->
<g id="node15" class="node">
-<title>Node161</title>
+<title>Node162</title>
<g id="a_node15"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
<polygon fill="#ffffff" stroke="#ff0000" points="973.5,-73 973.5,-92 1094.5,-92 1094.5,-73 973.5,-73"/>
<text text-anchor="middle" x="1034" y="-80" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
</a>
</g>
</g>
-<!-- Node4->Node161 -->
+<!-- Node4->Node162 -->
<g id="edge111" class="edge">
-<title>Node4->Node161</title>
+<title>Node4->Node162</title>
<path fill="none" stroke="#191970" d="M983.685,-734.2436C901.2221,-724.8989 767.7859,-707.7392 721,-690 706.6191,-684.5474 488,-555.8799 488,-540.5 488,-540.5 488,-540.5 488,-406.5 488,-246.5733 580.7362,-207.0585 723,-134 814.649,-86.9343 850.4815,-115.57 952,-98 962.707,-96.1469 974.1785,-94.0493 985.0475,-92.0094"/>
<polygon fill="#191970" stroke="#191970" points="983.3992,-737.7334 993.7274,-735.3715 984.1806,-730.7771 983.3992,-737.7334"/>
</g>
@@ -534,24 +534,24 @@
<path fill="none" stroke="#191970" d="M1119.0742,-395.0446C1058.6294,-383.2247 973.2691,-365.1279 961,-355 930.0093,-329.4179 924,-312.6854 924,-272.5 924,-272.5 924,-272.5 924,-211 924,-168.6872 906.1149,-121.2048 895.9149,-97.5656"/>
<polygon fill="#191970" stroke="#191970" points="1118.4365,-398.486 1128.9206,-396.9578 1119.7718,-391.6145 1118.4365,-398.486"/>
</g>
-<!-- Node13->Node161 -->
+<!-- Node13->Node162 -->
<g id="edge36" class="edge">
-<title>Node13->Node161</title>
+<title>Node13->Node162</title>
<path fill="none" stroke="#191970" d="M1120.2725,-394.8748C1064.881,-383.4896 989.7147,-366.2571 981,-355 916.0934,-271.1577 1001.8869,-130.1712 1027.2998,-92.1767"/>
<polygon fill="#191970" stroke="#191970" points="1119.8675,-398.364 1130.3648,-396.9298 1121.2643,-391.5048 1119.8675,-398.364"/>
</g>
-<!-- Node162 -->
+<!-- Node163 -->
<g id="node16" class="node">
-<title>Node162</title>
+<title>Node163</title>
<g id="a_node16"><a xlink:href="relay_2base_8h.html" target="_top" xlink:title="Base classes for the Relay IR. ">
<polygon fill="#ffffff" stroke="#ff0000" points="1608.5,-6 1608.5,-25 1745.5,-25 1745.5,-6 1608.5,-6"/>
<text text-anchor="middle" x="1677" y="-13" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/base.h</text>
</a>
</g>
</g>
-<!-- Node13->Node162 -->
+<!-- Node13->Node163 -->
<g id="edge31" class="edge">
-<title>Node13->Node162</title>
+<title>Node13->Node163</title>
<path fill="none" stroke="#191970" d="M1258.8622,-402.7711C1392.656,-395.8177 1652.9909,-379.3376 1687,-355 1744.2523,-314.0291 1753,-281.402 1753,-211 1753,-211 1753,-211 1753,-149.5 1753,-96.8121 1707.9944,-45.8623 1687.2497,-25.1761"/>
<polygon fill="#191970" stroke="#191970" points="1258.4764,-399.2862 1248.6693,-403.2951 1258.8358,-406.277 1258.4764,-399.2862"/>
</g>
@@ -660,9 +660,9 @@
<path fill="none" stroke="#191970" d="M1343.0628,-193.3133C1356.3842,-177.064 1377.3685,-152.7334 1398,-134 1412.7442,-120.6123 1430.8943,-107.3384 1445.0179,-97.6203"/>
<polygon fill="#191970" stroke="#191970" points="1340.0707,-191.4469 1336.5013,-201.4224 1345.5124,-195.8501 1340.0707,-191.4469"/>
</g>
-<!-- Node135->Node161 -->
+<!-- Node135->Node162 -->
<g id="edge18" class="edge">
-<title>Node135->Node161</title>
+<title>Node135->Node162</title>
<path fill="none" stroke="#191970" d="M1306.2487,-195.9241C1280.1987,-179.1441 1235.8428,-152.0228 1195,-134 1154.0557,-115.9324 1104.7598,-101.1153 1071.5922,-92.0875"/>
<polygon fill="#191970" stroke="#191970" points="1304.3723,-198.8788 1314.6637,-201.3962 1308.1884,-193.0105 1304.3723,-198.8788"/>
</g>
@@ -706,15 +706,15 @@
<path fill="none" stroke="#191970" d="M1537.5475,-70.1401C1608.8119,-57.4751 1718.4326,-37.9934 1785.7143,-26.0362"/>
<polygon fill="#191970" stroke="#191970" points="1536.5543,-66.7617 1527.321,-71.9575 1537.7791,-73.6537 1536.5543,-66.7617"/>
</g>
-<!-- Node161->Node162 -->
+<!-- Node162->Node163 -->
<g id="edge19" class="edge">
-<title>Node161->Node162</title>
+<title>Node162->Node163</title>
<path fill="none" stroke="#191970" d="M1104.9533,-75.1067C1229.3272,-62.1471 1485.1461,-35.491 1608.0748,-22.6819"/>
<polygon fill="#191970" stroke="#191970" points="1104.3648,-71.649 1094.7814,-76.1666 1105.0904,-78.6113 1104.3648,-71.649"/>
</g>
-<!-- Node169->Node161 -->
+<!-- Node169->Node162 -->
<g id="edge22" class="edge">
-<title>Node169->Node161</title>
+<title>Node169->Node162</title>
<path fill="none" stroke="#191970" d="M1106.268,-134.0102C1088.109,-121.067 1062.7886,-103.0195 1047.4429,-92.0817"/>
<polygon fill="#191970" stroke="#191970" points="1104.3574,-136.9464 1114.5321,-139.9005 1108.4204,-131.2462 1104.3574,-136.9464"/>
</g>
@@ -724,9 +724,9 @@
<path fill="none" stroke="#191970" d="M1612.2338,-320.0122C1605.5207,-300.8721 1592.5146,-272.3725 1571,-257 1542.2059,-236.4263 1449.4997,-223.2284 1387.552,-216.4682"/>
<polygon fill="#191970" stroke="#191970" points="1608.9889,-321.3525 1615.4032,-329.785 1615.6475,-319.193 1608.9889,-321.3525"/>
</g>
-<!-- Node171->Node162 -->
+<!-- Node171->Node163 -->
<g id="edge28" class="edge">
-<title>Node171->Node162</title>
+<title>Node171->Node163</title>
<path fill="none" stroke="#191970" d="M1632.0769,-321.8208C1649.5289,-298.0732 1677,-253.8816 1677,-211 1677,-211 1677,-211 1677,-149.5 1677,-102.9722 1677,-47.357 1677,-25.2517"/>
<polygon fill="#191970" stroke="#191970" points="1629.2197,-319.796 1625.957,-329.8759 1634.7935,-324.0307 1629.2197,-319.796"/>
</g>
@@ -863,9 +863,9 @@
<path fill="none" stroke="#191970" d="M230,-582.3415C230,-554.8131 230,-511.5714 230,-488.7614"/>
<polygon fill="#191970" stroke="#191970" points="226.5001,-582.3889 230,-592.389 233.5001,-582.389 226.5001,-582.3889"/>
</g>
-<!-- Node188->Node161 -->
+<!-- Node188->Node162 -->
<g id="edge87" class="edge">
-<title>Node188->Node161</title>
+<title>Node188->Node162</title>
<path fill="none" stroke="#191970" d="M763.9792,-659.7518C721.3691,-649.6038 671.1996,-635.7131 653,-623 541.8143,-545.3327 522.0464,-440.1936 592,-324 653.2727,-222.2255 675.4843,-190.47 780,-134 784.5338,-131.5504 912.7261,-106.2192 985.2193,-92.0198"/>
<polygon fill="#191970" stroke="#191970" points="763.2449,-663.1745 773.7802,-662.0527 764.8448,-656.3598 763.2449,-663.1745"/>
</g>
diff --git a/docs/reference/api/doxygen/data__type_8h__dep__incl.svg b/docs/reference/api/doxygen/data__type_8h__dep__incl.svg
index 1e0bef8a3..bc2f688ab 100644
--- a/docs/reference/api/doxygen/data__type_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/data__type_8h__dep__incl.svg
@@ -31,18 +31,18 @@
<path fill="none" stroke="#191970" d="M1342.627,-794.5859C1197.3876,-788.001 862.3041,-770.6068 751,-746 648.6289,-723.3681 530,-768.343 530,-663.5 530,-663.5 530,-663.5 530,-596.5 530,-549.1045 514.1309,-493.9774 507.1881,-472.1272"/>
<polygon fill="#191970" stroke="#191970" points="1342.6032,-798.0883 1352.7505,-795.0413 1342.9179,-791.0954 1342.6032,-798.0883"/>
</g>
-<!-- Node165 -->
+<!-- Node166 -->
<g id="node22" class="node">
-<title>Node165</title>
+<title>Node166</title>
<g id="a_node22"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
<polygon fill="#ffffff" stroke="#ff0000" points="1329.5,-263 1329.5,-282 1450.5,-282 1450.5,-263 1329.5,-263"/>
<text text-anchor="middle" x="1390" y="-270" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
</a>
</g>
</g>
-<!-- Node8->Node165 -->
+<!-- Node8->Node166 -->
<g id="edge128" class="edge">
-<title>Node8->Node165</title>
+<title>Node8->Node166</title>
<path fill="none" stroke="#191970" d="M1479.1306,-795.1756C1679.0419,-786.7787 2257.0715,-750.7137 2370,-612 2472.1019,-486.5849 2721.0387,-752.5927 2099,-447 1987.1406,-392.0462 1966.8294,-358.2543 1847,-324 1697.0713,-281.1415 1652.7624,-307.08 1498,-288 1482.6782,-286.111 1466.166,-283.8538 1450.7742,-281.6543"/>
<polygon fill="#191970" stroke="#191970" points="1478.8986,-791.6821 1469.0504,-795.5889 1479.1855,-798.6762 1478.8986,-791.6821"/>
</g>
@@ -308,9 +308,9 @@
<path fill="none" stroke="#191970" d="M573.205,-454.9537C685.3019,-442.7304 904.5837,-418.8193 1009.4365,-407.3859"/>
<polygon fill="#191970" stroke="#191970" points="572.5882,-451.5001 563.0266,-456.0636 573.3471,-458.4589 572.5882,-451.5001"/>
</g>
-<!-- Node161 -->
+<!-- Node162 -->
<g id="node21" class="node">
-<title>Node161</title>
+<title>Node162</title>
<g id="a_node21"><a xlink:href="tensor__type_8h.html" target="_top" xlink:title="Polymorphic tensor types. ">
<polygon fill="#ffffff" stroke="#000000" points="732,-324.5 732,-354.5 850,-354.5 850,-324.5 732,-324.5"/>
<text text-anchor="start" x="740" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/tensor</text>
@@ -318,9 +318,9 @@
</a>
</g>
</g>
-<!-- Node9->Node161 -->
+<!-- Node9->Node162 -->
<g id="edge54" class="edge">
-<title>Node9->Node161</title>
+<title>Node9->Node162</title>
<path fill="none" stroke="#191970" d="M535.9726,-448.7974C590.1769,-425.567 699.6873,-378.634 755.9248,-354.5322"/>
<polygon fill="#191970" stroke="#191970" points="534.3331,-445.6921 526.5204,-452.8484 537.0906,-452.1262 534.3331,-445.6921"/>
</g>
@@ -539,15 +539,15 @@
<path fill="none" stroke="#191970" d="M999.0603,-395.8072C913.1686,-388.6478 771.7606,-374.3961 723,-355 701.1086,-346.292 701.8131,-332.9022 680,-324 673.5484,-321.367 495.5756,-295.1693 399.7616,-281.1886"/>
<polygon fill="#191970" stroke="#191970" points="998.9965,-399.3136 1009.2497,-396.645 999.5702,-392.3372 998.9965,-399.3136"/>
</g>
-<!-- Node140->Node161 -->
+<!-- Node140->Node162 -->
<g id="edge31" class="edge">
-<title>Node140->Node161</title>
+<title>Node140->Node162</title>
<path fill="none" stroke="#191970" d="M1015.0091,-389.2349C967.8496,-378.7644 898.9607,-363.4696 850.0486,-352.6101"/>
<polygon fill="#191970" stroke="#191970" points="1014.5018,-392.7074 1025.0227,-391.4581 1016.019,-385.8738 1014.5018,-392.7074"/>
</g>
-<!-- Node140->Node165 -->
+<!-- Node140->Node166 -->
<g id="edge35" class="edge">
-<title>Node140->Node165</title>
+<title>Node140->Node166</title>
<path fill="none" stroke="#191970" d="M1087.0936,-384.7108C1110.976,-364.6238 1150.8152,-332.1337 1168,-324 1219.4969,-299.6262 1283.302,-286.378 1329.2746,-279.4512"/>
<polygon fill="#191970" stroke="#191970" points="1084.7505,-382.1086 1079.3749,-391.2384 1089.2707,-387.4535 1084.7505,-382.1086"/>
</g>
@@ -624,66 +624,66 @@
<path fill="none" stroke="#191970" d="M335.4914,-253.011C328.741,-223.8319 316.2613,-169.8876 311.2092,-148.0496"/>
<polygon fill="#191970" stroke="#191970" points="332.1066,-253.9093 337.7706,-262.8631 338.9265,-252.3315 332.1066,-253.9093"/>
</g>
-<!-- Node161->Node127 -->
+<!-- Node162->Node127 -->
<g id="edge32" class="edge">
-<title>Node161->Node127</title>
+<title>Node162->Node127</title>
<path fill="none" stroke="#191970" d="M784.8135,-314.6303C773.7186,-270.0287 750.8986,-178.2926 743.4269,-148.2562"/>
<polygon fill="#191970" stroke="#191970" points="781.4313,-315.533 787.2419,-324.3923 788.2243,-313.8431 781.4313,-315.533"/>
</g>
-<!-- Node165->Node105 -->
+<!-- Node166->Node105 -->
<g id="edge46" class="edge">
-<title>Node165->Node105</title>
+<title>Node166->Node105</title>
<path fill="none" stroke="#191970" d="M1395.6419,-253.0483C1398.1298,-243.4855 1400.7658,-231.735 1402,-221 1407.373,-174.2639 1401.3986,-148.7077 1362,-123 1311.5161,-90.0591 1138.6444,-80.7014 1048.1905,-78.0475"/>
<polygon fill="#191970" stroke="#191970" points="1392.2523,-252.1748 1392.9785,-262.7447 1399.0023,-254.0289 1392.2523,-252.1748"/>
</g>
-<!-- Node165->Node106 -->
+<!-- Node166->Node106 -->
<g id="edge38" class="edge">
-<title>Node165->Node106</title>
+<title>Node166->Node106</title>
<path fill="none" stroke="#191970" d="M1400.6225,-254.1669C1405.9631,-244.4986 1412.2845,-232.323 1417,-221 1445.8435,-151.7404 1469.3778,-64.9143 1478.2265,-30.5223"/>
<polygon fill="#191970" stroke="#191970" points="1397.5745,-252.4465 1395.7045,-262.8751 1403.6697,-255.8889 1397.5745,-252.4465"/>
</g>
-<!-- Node165->Node127 -->
+<!-- Node166->Node127 -->
<g id="edge42" class="edge">
-<title>Node165->Node127</title>
+<title>Node166->Node127</title>
<path fill="none" stroke="#191970" d="M1386.0735,-252.9201C1381.0552,-233.465 1370.3289,-204.4994 1349,-190 1305.1963,-160.2223 958.7421,-145.5391 809.2645,-140.5402"/>
<polygon fill="#191970" stroke="#191970" points="1382.7039,-253.8849 1388.3555,-262.8464 1389.5259,-252.3164 1382.7039,-253.8849"/>
</g>
-<!-- Node165->Node44 -->
+<!-- Node166->Node44 -->
<g id="edge43" class="edge">
-<title>Node165->Node44</title>
+<title>Node166->Node44</title>
<path fill="none" stroke="#191970" d="M1460.4655,-262.3744C1472.9682,-260.5814 1485.8671,-258.7339 1498,-257 1610.4348,-240.9325 1639.4946,-242.5941 1751,-221 1754.0963,-220.4004 1757.2689,-219.7465 1760.4642,-219.0574"/>
<polygon fill="#191970" stroke="#191970" points="1459.9474,-258.9128 1450.5458,-263.7975 1460.9415,-265.8419 1459.9474,-258.9128"/>
</g>
-<!-- Node165->Node142 -->
+<!-- Node166->Node142 -->
<g id="edge36" class="edge">
-<title>Node165->Node142</title>
+<title>Node166->Node142</title>
<path fill="none" stroke="#191970" d="M1393.0945,-252.712C1394.8111,-234.7938 1394.5156,-208.4542 1382,-190 1367.5315,-168.6663 1341.3967,-155.6407 1319.4733,-148.0275"/>
<polygon fill="#191970" stroke="#191970" points="1389.5924,-252.511 1391.8207,-262.8689 1396.538,-253.3822 1389.5924,-252.511"/>
</g>
-<!-- Node165->Node143 -->
+<!-- Node166->Node143 -->
<g id="edge37" class="edge">
-<title>Node165->Node143</title>
+<title>Node166->Node143</title>
<path fill="none" stroke="#191970" d="M1365.8706,-257.5308C1348.0926,-246.5019 1324.0122,-231.5631 1306.2142,-220.5218"/>
<polygon fill="#191970" stroke="#191970" points="1364.1835,-260.603 1374.5262,-262.9005 1367.8737,-254.6546 1364.1835,-260.603"/>
</g>
-<!-- Node165->Node144 -->
+<!-- Node166->Node144 -->
<g id="edge39" class="edge">
-<title>Node165->Node144</title>
+<title>Node166->Node144</title>
<path fill="none" stroke="#191970" d="M1343.3955,-260.3973C1293.3762,-247.4078 1214.9338,-227.0371 1168.9823,-215.1039"/>
<polygon fill="#191970" stroke="#191970" points="1342.7502,-263.8457 1353.3089,-262.9717 1344.5097,-257.0705 1342.7502,-263.8457"/>
</g>
-<!-- Node166 -->
+<!-- Node167 -->
<g id="node23" class="node">
-<title>Node166</title>
+<title>Node167</title>
<g id="a_node23"><a xlink:href="relay_2base_8h.html" target="_top" xlink:title="Base classes for the Relay IR. ">
<polygon fill="#ffffff" stroke="#ff0000" points="826.5,-196 826.5,-215 963.5,-215 963.5,-196 826.5,-196"/>
<text text-anchor="middle" x="895" y="-203" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/base.h</text>
</a>
</g>
</g>
-<!-- Node165->Node166 -->
+<!-- Node166->Node167 -->
<g id="edge40" class="edge">
-<title>Node165->Node166</title>
+<title>Node166->Node167</title>
<path fill="none" stroke="#191970" d="M1319.2401,-262.9224C1224.4834,-250.0967 1057.4664,-227.4904 963.7463,-214.8051"/>
<polygon fill="#191970" stroke="#191970" points="1318.9242,-266.4115 1329.3033,-264.2845 1319.8631,-259.4747 1318.9242,-266.4115"/>
</g>
@@ -696,15 +696,15 @@
</a>
</g>
</g>
-<!-- Node165->Node147 -->
+<!-- Node166->Node147 -->
<g id="edge44" class="edge">
-<title>Node165->Node147</title>
+<title>Node166->Node147</title>
<path fill="none" stroke="#191970" d="M1439.9162,-260.6405C1494.5367,-247.6633 1581.0369,-227.1118 1631.5775,-215.1039"/>
<polygon fill="#191970" stroke="#191970" points="1439.0243,-257.2549 1430.1042,-262.9717 1440.6425,-264.0653 1439.0243,-257.2549"/>
</g>
-<!-- Node166->Node127 -->
+<!-- Node167->Node127 -->
<g id="edge41" class="edge">
-<title>Node166->Node127</title>
+<title>Node167->Node127</title>
<path fill="none" stroke="#191970" d="M863.378,-191.7424C833.4515,-178.7224 789.3071,-159.5167 763.0235,-148.0817"/>
<polygon fill="#191970" stroke="#191970" points="862.3695,-195.1205 872.9356,-195.9005 865.1621,-188.7016 862.3695,-195.1205"/>
</g>
@@ -714,9 +714,9 @@
<path fill="none" stroke="#191970" d="M1672,-185.7758C1672,-175.4641 1672,-163.0437 1672,-153.5218"/>
<polygon fill="#191970" stroke="#191970" points="1668.5001,-185.9005 1672,-195.9005 1675.5001,-185.9006 1668.5001,-185.9005"/>
</g>
-<!-- Node173->Node165 -->
+<!-- Node173->Node166 -->
<g id="edge50" class="edge">
-<title>Node173->Node165</title>
+<title>Node173->Node166</title>
<path fill="none" stroke="#191970" d="M1485.7018,-325.0575C1462.0286,-312.0567 1427.9071,-293.3179 1407.4472,-282.0817"/>
<polygon fill="#191970" stroke="#191970" points="1484.0704,-328.1546 1494.5204,-329.9005 1487.44,-322.019 1484.0704,-328.1546"/>
</g>
@@ -812,15 +812,15 @@
<path fill="none" stroke="#191970" d="M1029.8132,-510.0446C1039.5754,-482.1679 1057.18,-431.8972 1064.5784,-410.7705"/>
<polygon fill="#191970" stroke="#191970" points="1026.3842,-509.2468 1026.3823,-519.8416 1032.9908,-511.5604 1026.3842,-509.2468"/>
</g>
-<!-- Node178->Node165 -->
+<!-- Node178->Node166 -->
<g id="edge71" class="edge">
-<title>Node178->Node165</title>
+<title>Node178->Node166</title>
<path fill="none" stroke="#191970" d="M1033.3478,-510.898C1044.0301,-492.918 1062.0553,-465.7224 1083,-447 1107.7392,-424.8857 1310.205,-315.3709 1372.0957,-282.1018"/>
<polygon fill="#191970" stroke="#191970" points="1030.1624,-509.4155 1028.1968,-519.8264 1036.2257,-512.9136 1030.1624,-509.4155"/>
</g>
-<!-- Node178->Node166 -->
+<!-- Node178->Node167 -->
<g id="edge70" class="edge">
-<title>Node178->Node166</title>
+<title>Node178->Node167</title>
<path fill="none" stroke="#191970" d="M1015.3759,-510.2016C991.1624,-448.911 916.5601,-260.0739 898.8158,-215.1589"/>
<polygon fill="#191970" stroke="#191970" points="1012.2397,-511.7888 1019.1692,-519.8033 1018.7501,-509.2168 1012.2397,-511.7888"/>
</g>
diff --git a/docs/reference/api/doxygen/functor_8h.html b/docs/reference/api/doxygen/functor_8h.html
index 71699a055..a60b21f03 100644
--- a/docs/reference/api/doxygen/functor_8h.html
+++ b/docs/reference/api/doxygen/functor_8h.html
@@ -87,7 +87,7 @@ Include dependency graph for functor.h:</div>
</div><div class="textblock"><div class="dynheader">
This graph shows which files directly or indirectly include this file:</div>
<div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="functor_8h__dep__incl.svg" width="3135" height="1334"><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="functor_8h__dep__incl.svg" width="3136" height="1334"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
</div>
</div>
</div>
diff --git a/docs/reference/api/doxygen/functor_8h__dep__incl.svg b/docs/reference/api/doxygen/functor_8h__dep__incl.svg
index 727220b83..5668abaa6 100644
--- a/docs/reference/api/doxygen/functor_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/functor_8h__dep__incl.svg
@@ -4,573 +4,573 @@
<!-- Generated by graphviz version 2.40.1 (20161225.0304)
-->
<!-- Title: include/tvm/node/functor.h Pages: 1 -->
-<svg width="2351pt" height="1000pt"
- viewBox="0.00 0.00 2351.00 1000.00" xmlns="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink">
+<svg width="2352pt" height="1000pt"
+ viewBox="0.00 0.00 2352.00 1000.00" 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(4 996)">
<title>include/tvm/node/functor.h</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-996 2347,-996 2347,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-996 2348,-996 2348,4 -4,4"/>
<!-- Node13 -->
<g id="node1" class="node">
<title>Node13</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="1378.5,-972.5 1378.5,-991.5 1525.5,-991.5 1525.5,-972.5 1378.5,-972.5"/>
-<text text-anchor="middle" x="1452" y="-979.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/functor.h</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="1387.5,-972.5 1387.5,-991.5 1534.5,-991.5 1534.5,-972.5 1387.5,-972.5"/>
+<text text-anchor="middle" x="1461" y="-979.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/functor.h</text>
</g>
<!-- Node14 -->
<g id="node2" class="node">
<title>Node14</title>
<g id="a_node2"><a xlink:href="type__functor_8h.html" target="_top" xlink:title="A way to defined arbitrary function signature with dispatch on types. ">
-<polygon fill="#ffffff" stroke="#000000" points="458.5,-905.5 458.5,-935.5 567.5,-935.5 567.5,-905.5 458.5,-905.5"/>
-<text text-anchor="start" x="466.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/type</text>
-<text text-anchor="middle" x="513" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="847.5,-905.5 847.5,-935.5 956.5,-935.5 956.5,-905.5 847.5,-905.5"/>
+<text text-anchor="start" x="855.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/type</text>
+<text text-anchor="middle" x="902" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
</a>
</g>
</g>
<!-- Node13->Node14 -->
<g id="edge1" class="edge">
<title>Node13->Node14</title>
-<path fill="none" stroke="#191970" d="M1368.145,-979.6169C1179.654,-973.9315 727.1679,-958.3602 576,-936 573.2144,-935.588 570.372,-935.1086 567.5106,-934.5789"/>
-<polygon fill="#191970" stroke="#191970" points="1368.177,-983.1193 1378.2774,-979.9206 1368.3868,-976.1224 1368.177,-983.1193"/>
+<path fill="none" stroke="#191970" d="M1377.0327,-977.3803C1279.1199,-971.24 1112.6139,-958.4594 971,-936 966.3921,-935.2692 961.6318,-934.4138 956.8659,-933.487"/>
+<polygon fill="#191970" stroke="#191970" points="1376.969,-980.8829 1387.1661,-978.0073 1377.4014,-973.8963 1376.969,-980.8829"/>
</g>
<!-- Node15 -->
<g id="node3" class="node">
<title>Node15</title>
<g id="a_node3"><a xlink:href="repr__printer_8h.html" target="_top" xlink:title="Printer class to print repr string of each AST/IR nodes. ">
-<polygon fill="#ffffff" stroke="#000000" points="1150,-838.5 1150,-868.5 1274,-868.5 1274,-838.5 1150,-838.5"/>
-<text text-anchor="start" x="1158" y="-856.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/repr</text>
-<text text-anchor="middle" x="1212" y="-845.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_printer.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="816,-838.5 816,-868.5 940,-868.5 940,-838.5 816,-838.5"/>
+<text text-anchor="start" x="824" y="-856.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/repr</text>
+<text text-anchor="middle" x="878" y="-845.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_printer.h</text>
</a>
</g>
</g>
<!-- Node13->Node15 -->
<g id="edge2" class="edge">
<title>Node13->Node15</title>
-<path fill="none" stroke="#191970" d="M1424.9374,-967.5103C1379.1929,-943.0179 1287.0192,-893.6665 1240.3921,-868.7016"/>
-<polygon fill="#191970" stroke="#191970" points="1423.493,-970.707 1433.961,-972.3416 1426.7972,-964.5358 1423.493,-970.707"/>
+<path fill="none" stroke="#191970" d="M1377.1997,-979.203C1264.2145,-974.4716 1069.6154,-962.6624 1004,-936 983.4868,-927.6646 983.2365,-917.5579 965,-905 945.7733,-891.7602 923.042,-878.3769 905.637,-868.5728"/>
+<polygon fill="#191970" stroke="#191970" points="1377.0662,-982.7004 1387.201,-979.6123 1377.3525,-975.7062 1377.0662,-982.7004"/>
</g>
-<!-- Node171 -->
-<g id="node33" class="node">
-<title>Node171</title>
-<g id="a_node33"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="965.5,-464 965.5,-483 1086.5,-483 1086.5,-464 965.5,-464"/>
-<text text-anchor="middle" x="1026" y="-471" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
+<!-- Node156 -->
+<g id="node27" class="node">
+<title>Node156</title>
+<g id="a_node27"><a xlink:href="stmt__functor_8h.html" target="_top" xlink:title="Functors for tir stmts utility functions to call common functors. ">
+<polygon fill="#ffffff" stroke="#000000" points="1738.5,-268.5 1738.5,-298.5 1851.5,-298.5 1851.5,-268.5 1738.5,-268.5"/>
+<text text-anchor="start" x="1746.5" y="-286.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/stmt</text>
+<text text-anchor="middle" x="1795" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
</a>
</g>
</g>
-<!-- Node13->Node171 -->
-<g id="edge112" class="edge">
-<title>Node13->Node171</title>
-<path fill="none" stroke="#191970" d="M1452,-962.2188C1452,-936.9855 1452,-891.9988 1452,-853.5 1452,-853.5 1452,-853.5 1452,-602 1452,-563.8373 1445.2243,-546.9416 1414,-525 1355.812,-484.1107 1166.1484,-500.7038 1096,-489 1086.0553,-487.3408 1075.3958,-485.1798 1065.4695,-483.0007"/>
-<polygon fill="#191970" stroke="#191970" points="1448.5001,-962.3281 1452,-972.3281 1455.5001,-962.3282 1448.5001,-962.3281"/>
+<!-- Node13->Node156 -->
+<g id="edge115" class="edge">
+<title>Node13->Node156</title>
+<path fill="none" stroke="#191970" d="M1544.7194,-976.031C1615.1885,-969.7753 1718.2288,-957.7322 1806,-936 1908.9833,-910.5013 2030,-959.5931 2030,-853.5 2030,-853.5 2030,-853.5 2030,-406.5 2030,-326.2593 1921.1237,-298.2662 1851.5874,-288.5703"/>
+<polygon fill="#191970" stroke="#191970" points="1544.2243,-972.5607 1534.5644,-976.9124 1544.8296,-979.5345 1544.2243,-972.5607"/>
</g>
-<!-- Node173 -->
-<g id="node38" class="node">
-<title>Node173</title>
-<g id="a_node38"><a xlink:href="tir_2expr__functor_8h.html" target="_top" xlink:title="Functors for tir expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="1887.5,-391.5 1887.5,-421.5 2000.5,-421.5 2000.5,-391.5 1887.5,-391.5"/>
-<text text-anchor="start" x="1895.5" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr</text>
-<text text-anchor="middle" x="1944" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
+<!-- Node172 -->
+<g id="node34" class="node">
+<title>Node172</title>
+<g id="a_node34"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
+<polygon fill="#ffffff" stroke="#000000" points="946.5,-464 946.5,-483 1067.5,-483 1067.5,-464 946.5,-464"/>
+<text text-anchor="middle" x="1007" y="-471" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
</a>
</g>
</g>
-<!-- Node13->Node173 -->
+<!-- Node13->Node172 -->
<g id="edge113" class="edge">
-<title>Node13->Node173</title>
-<path fill="none" stroke="#191970" d="M1523.7426,-970.7459C1626.675,-952.5414 1804,-912.603 1804,-853.5 1804,-853.5 1804,-853.5 1804,-540.5 1804,-482.631 1867.1982,-442.0338 1908.6721,-421.6223"/>
-<polygon fill="#191970" stroke="#191970" points="1522.9676,-967.3281 1513.7165,-972.4921 1524.1688,-974.2243 1522.9676,-967.3281"/>
+<title>Node13->Node172</title>
+<path fill="none" stroke="#191970" d="M1461,-962.2188C1461,-936.9855 1461,-891.9988 1461,-853.5 1461,-853.5 1461,-853.5 1461,-602 1461,-561.6729 1447.021,-546.6528 1413,-525 1356.727,-489.1848 1181.1288,-497.7386 1115,-489 1099.6013,-486.9651 1082.9866,-484.6436 1067.5148,-482.4277"/>
+<polygon fill="#191970" stroke="#191970" points="1457.5001,-962.3281 1461,-972.3281 1464.5001,-962.3282 1457.5001,-962.3281"/>
</g>
<!-- Node174 -->
<g id="node39" class="node">
<title>Node174</title>
-<g id="a_node39"><a xlink:href="stmt__functor_8h.html" target="_top" xlink:title="Functors for tir stmts utility functions to call common functors. ">
-<polygon fill="#ffffff" stroke="#000000" points="1724.5,-324.5 1724.5,-354.5 1837.5,-354.5 1837.5,-324.5 1724.5,-324.5"/>
-<text text-anchor="start" x="1732.5" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/stmt</text>
-<text text-anchor="middle" x="1781" y="-331.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
+<g id="a_node39"><a xlink:href="tir_2expr__functor_8h.html" target="_top" xlink:title="Functors for tir expressions. ">
+<polygon fill="#ffffff" stroke="#000000" points="1888.5,-391.5 1888.5,-421.5 2001.5,-421.5 2001.5,-391.5 1888.5,-391.5"/>
+<text text-anchor="start" x="1896.5" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr</text>
+<text text-anchor="middle" x="1945" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
</a>
</g>
</g>
<!-- Node13->Node174 -->
<g id="edge114" class="edge">
<title>Node13->Node174</title>
-<path fill="none" stroke="#191970" d="M1535.9502,-971.9488C1695.9989,-951.6426 2029,-903.3976 2029,-853.5 2029,-853.5 2029,-853.5 2029,-473.5 2029,-435.8735 2036.6743,-417.5374 2010,-391 1986.0947,-367.2174 1897.3895,-352.6609 1837.6596,-345.3413"/>
-<polygon fill="#191970" stroke="#191970" points="1535.1472,-968.5222 1525.6635,-973.2456 1536.0228,-975.4672 1535.1472,-968.5222"/>
+<path fill="none" stroke="#191970" d="M1544.7498,-973.5401C1667.2786,-958.8055 1881,-923.1252 1881,-853.5 1881,-853.5 1881,-853.5 1881,-540.5 1881,-493.5784 1913.3477,-445.6455 1932.0486,-421.855"/>
+<polygon fill="#191970" stroke="#191970" points="1544.1598,-970.0854 1534.639,-974.7335 1544.9804,-977.0371 1544.1598,-970.0854"/>
</g>
<!-- Node184 -->
<g id="node46" class="node">
<title>Node184</title>
<g id="a_node46"><a xlink:href="structural__equal_8h.html" target="_top" xlink:title="Structural equality comparison. ">
-<polygon fill="#ffffff" stroke="#000000" points="585.5,-905.5 585.5,-935.5 736.5,-935.5 736.5,-905.5 585.5,-905.5"/>
-<text text-anchor="start" x="593.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/structural</text>
-<text text-anchor="middle" x="661" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_equal.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1013.5,-905.5 1013.5,-935.5 1164.5,-935.5 1164.5,-905.5 1013.5,-905.5"/>
+<text text-anchor="start" x="1021.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/structural</text>
+<text text-anchor="middle" x="1089" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_equal.h</text>
</a>
</g>
</g>
<!-- Node13->Node184 -->
-<g id="edge101" class="edge">
+<g id="edge102" class="edge">
<title>Node13->Node184</title>
-<path fill="none" stroke="#191970" d="M1368.362,-979.7386C1235.7416,-975.4383 969.2455,-963.8865 745,-936 742.3511,-935.6706 739.6594,-935.3109 736.944,-934.9267"/>
-<polygon fill="#191970" stroke="#191970" points="1368.345,-983.2397 1378.4514,-980.0601 1368.568,-976.2433 1368.345,-983.2397"/>
+<path fill="none" stroke="#191970" d="M1388.8962,-970.8817C1330.8116,-961.8196 1247.079,-948.5181 1174,-936 1170.8952,-935.4682 1167.7277,-934.9193 1164.5289,-934.3597"/>
+<polygon fill="#191970" stroke="#191970" points="1388.6538,-974.3861 1399.0734,-972.4671 1389.7313,-967.4695 1388.6538,-974.3861"/>
</g>
<!-- Node190 -->
<g id="node48" class="node">
<title>Node190</title>
<g id="a_node48"><a xlink:href="structural__hash_8h.html" target="_top" xlink:title="include/tvm/node/structural\l_hash.h">
-<polygon fill="#ffffff" stroke="#000000" points="754.5,-905.5 754.5,-935.5 905.5,-935.5 905.5,-905.5 754.5,-905.5"/>
-<text text-anchor="start" x="762.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/structural</text>
-<text text-anchor="middle" x="830" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_hash.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1182.5,-905.5 1182.5,-935.5 1333.5,-935.5 1333.5,-905.5 1182.5,-905.5"/>
+<text text-anchor="start" x="1190.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/structural</text>
+<text text-anchor="middle" x="1258" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_hash.h</text>
</a>
</g>
</g>
<!-- Node13->Node190 -->
-<g id="edge106" class="edge">
+<g id="edge107" class="edge">
<title>Node13->Node190</title>
-<path fill="none" stroke="#191970" d="M1368.1872,-973.713C1246.4677,-961.6781 1022.5833,-939.5416 905.6379,-927.9787"/>
-<polygon fill="#191970" stroke="#191970" points="1367.91,-977.2026 1378.2059,-974.7036 1368.5988,-970.2366 1367.91,-977.2026"/>
+<path fill="none" stroke="#191970" d="M1419.6599,-969.4758C1387.2489,-959.6567 1342.0327,-945.9582 1307.6594,-935.5446"/>
+<polygon fill="#191970" stroke="#191970" points="1418.696,-972.8408 1429.2813,-972.3906 1420.7256,-966.1415 1418.696,-972.8408"/>
</g>
<!-- Node80 -->
<g id="node49" class="node">
<title>Node80</title>
<g id="a_node49"><a xlink:href="relay_2expr__functor_8h.html" target="_top" xlink:title="A more powerful visitor which enables defining arbitrary function signatures with type based dispatch...">
-<polygon fill="#ffffff" stroke="#000000" points="2057.5,-905.5 2057.5,-935.5 2184.5,-935.5 2184.5,-905.5 2057.5,-905.5"/>
-<text text-anchor="start" x="2065.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/expr</text>
-<text text-anchor="middle" x="2121" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2058.5,-905.5 2058.5,-935.5 2185.5,-935.5 2185.5,-905.5 2058.5,-905.5"/>
+<text text-anchor="start" x="2066.5" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/expr</text>
+<text text-anchor="middle" x="2122" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
</a>
</g>
</g>
<!-- Node13->Node80 -->
-<g id="edge110" class="edge">
+<g id="edge111" class="edge">
<title>Node13->Node80</title>
-<path fill="none" stroke="#191970" d="M1535.6721,-978.613C1650.4887,-973.2542 1863.1321,-960.7826 2043,-936 2047.6503,-935.3593 2052.4412,-934.6153 2057.2533,-933.8069"/>
-<polygon fill="#191970" stroke="#191970" points="1535.3756,-975.1228 1525.5471,-979.0789 1535.6975,-982.1154 1535.3756,-975.1228"/>
+<path fill="none" stroke="#191970" d="M1544.946,-978.4759C1658.5314,-973.0015 1867.2973,-960.4421 2044,-936 2048.6499,-935.3568 2053.4406,-934.6111 2058.2525,-933.8013"/>
+<polygon fill="#191970" stroke="#191970" points="1544.7477,-974.9813 1534.9252,-978.9524 1545.0802,-981.9734 1544.7477,-974.9813"/>
</g>
<!-- Node81 -->
<g id="node50" class="node">
<title>Node81</title>
<g id="a_node50"><a xlink:href="pattern__functor_8h.html" target="_top" xlink:title="A more powerful visitor on ADT patterns that enables defining arbitrary function signatures with type...">
-<polygon fill="#ffffff" stroke="#000000" points="2203,-905.5 2203,-935.5 2343,-935.5 2343,-905.5 2203,-905.5"/>
-<text text-anchor="start" x="2211" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/pattern</text>
-<text text-anchor="middle" x="2273" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2204,-905.5 2204,-935.5 2344,-935.5 2344,-905.5 2204,-905.5"/>
+<text text-anchor="start" x="2212" y="-923.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/pattern</text>
+<text text-anchor="middle" x="2274" y="-912.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_functor.h</text>
</a>
</g>
</g>
<!-- Node13->Node81 -->
-<g id="edge111" class="edge">
+<g id="edge112" class="edge">
<title>Node13->Node81</title>
-<path fill="none" stroke="#191970" d="M1535.8347,-980.3988C1673.6659,-976.9614 1956.556,-966.5185 2194,-936 2196.9112,-935.6258 2199.8763,-935.2087 2202.867,-934.7583"/>
-<polygon fill="#191970" stroke="#191970" points="1535.6744,-976.9016 1525.7625,-980.6439 1535.8448,-983.8995 1535.6744,-976.9016"/>
+<path fill="none" stroke="#191970" d="M1545.1501,-980.3107C1681.9388,-976.7614 1960.8143,-966.1798 2195,-936 2197.911,-935.6248 2200.8761,-935.2069 2203.8667,-934.7558"/>
+<polygon fill="#191970" stroke="#191970" points="1544.66,-976.8219 1534.7518,-980.574 1544.8373,-983.8196 1544.66,-976.8219"/>
</g>
<!-- Node16 -->
<g id="node4" class="node">
<title>Node16</title>
<g id="a_node4"><a xlink:href="node_8h.html" target="_top" xlink:title="Definitions and helper macros for IR/AST nodes. ">
-<polygon fill="#ffffff" stroke="#000000" points="707.5,-782.5 707.5,-801.5 844.5,-801.5 844.5,-782.5 707.5,-782.5"/>
-<text text-anchor="middle" x="776" y="-789.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/node.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="811.5,-782.5 811.5,-801.5 948.5,-801.5 948.5,-782.5 811.5,-782.5"/>
+<text text-anchor="middle" x="880" y="-789.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/node.h</text>
</a>
</g>
</g>
<!-- Node15->Node16 -->
<g id="edge3" class="edge">
<title>Node15->Node16</title>
-<path fill="none" stroke="#191970" d="M1139.9688,-843.3396C1057.6377,-831.7264 923.9198,-812.8648 843.5226,-801.5244"/>
-<polygon fill="#191970" stroke="#191970" points="1139.604,-846.8227 1149.9949,-844.7539 1140.5818,-839.8914 1139.604,-846.8227"/>
+<path fill="none" stroke="#191970" d="M878.82,-828.2849C879.1224,-818.9858 879.4475,-808.9883 879.6798,-801.8469"/>
+<polygon fill="#191970" stroke="#191970" points="875.3214,-828.1892 878.4944,-838.2977 882.3177,-828.4168 875.3214,-828.1892"/>
</g>
<!-- Node17 -->
<g id="node5" class="node">
<title>Node17</title>
<g id="a_node5"><a xlink:href="auto__scheduler_2cost__model_8h.html" target="_top" xlink:title="Cost models that estimate the performance of programs. ">
-<polygon fill="#ffffff" stroke="#000000" points="294,-715.5 294,-745.5 446,-745.5 446,-715.5 294,-715.5"/>
-<text text-anchor="start" x="302" y="-733.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
-<text text-anchor="middle" x="370" y="-722.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/cost_model.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="142,-715.5 142,-745.5 294,-745.5 294,-715.5 142,-715.5"/>
+<text text-anchor="start" x="150" y="-733.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
+<text text-anchor="middle" x="218" y="-722.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/cost_model.h</text>
</a>
</g>
</g>
<!-- Node16->Node17 -->
<g id="edge4" class="edge">
<title>Node16->Node17</title>
-<path fill="none" stroke="#191970" d="M697.4818,-781.1482C633.4907,-772.1651 540.8169,-758.8392 460,-746 455.459,-745.2786 450.7919,-744.5217 446.0855,-743.7466"/>
-<polygon fill="#191970" stroke="#191970" points="697.1105,-784.6303 707.4995,-782.5519 698.0819,-777.698 697.1105,-784.6303"/>
+<path fill="none" stroke="#191970" d="M801.0485,-787.9113C690.4929,-781.6721 483.4083,-768.1119 308,-746 303.4381,-745.4249 298.7549,-744.7813 294.0363,-744.0919"/>
+<polygon fill="#191970" stroke="#191970" points="800.9367,-791.4104 811.1165,-788.4743 801.3276,-784.4214 800.9367,-791.4104"/>
</g>
<!-- Node18 -->
<g id="node6" class="node">
<title>Node18</title>
<g id="a_node6"><a xlink:href="search__policy_8h.html" target="_top" xlink:title="The base class of search policies, including the abstract definition of search policy and other suppo...">
-<polygon fill="#ffffff" stroke="#ff0000" points="502,-715.5 502,-745.5 654,-745.5 654,-715.5 502,-715.5"/>
-<text text-anchor="start" x="510" y="-733.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
-<text text-anchor="middle" x="578" y="-722.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/search_policy.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="523,-715.5 523,-745.5 675,-745.5 675,-715.5 523,-715.5"/>
+<text text-anchor="start" x="531" y="-733.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
+<text text-anchor="middle" x="599" y="-722.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/search_policy.h</text>
</a>
</g>
</g>
<!-- Node16->Node18 -->
<g id="edge5" class="edge">
<title>Node16->Node18</title>
-<path fill="none" stroke="#191970" d="M735.2687,-779.3486C703.6854,-769.5386 659.8184,-755.9133 626.4362,-745.5446"/>
-<polygon fill="#191970" stroke="#191970" points="734.4743,-782.7668 745.0625,-782.3906 736.5508,-776.0818 734.4743,-782.7668"/>
+<path fill="none" stroke="#191970" d="M826.5212,-780.2956C781.4109,-770.4227 716.5584,-756.229 667.6248,-745.5193"/>
+<polygon fill="#191970" stroke="#191970" points="825.885,-783.7391 836.4021,-782.4581 827.3816,-776.9009 825.885,-783.7391"/>
</g>
<!-- Node20 -->
<g id="node7" class="node">
<title>Node20</title>
<g id="a_node7"><a xlink:href="transform__step_8h.html" target="_top" xlink:title="Transformation steps. These steps are used to manipulate LoopState. They are similar to the schedule ...">
-<polygon fill="#ffffff" stroke="#ff0000" points="334,-134.5 334,-164.5 486,-164.5 486,-134.5 334,-134.5"/>
-<text text-anchor="start" x="342" y="-152.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
-<text text-anchor="middle" x="410" y="-141.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/transform_step.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="592,-134.5 592,-164.5 744,-164.5 744,-134.5 592,-134.5"/>
+<text text-anchor="start" x="600" y="-152.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/auto_scheduler</text>
+<text text-anchor="middle" x="668" y="-141.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/transform_step.h</text>
</a>
</g>
</g>
<!-- Node16->Node20 -->
<g id="edge6" class="edge">
<title>Node16->Node20</title>
-<path fill="none" stroke="#191970" d="M696.9166,-788.1798C555.2478,-780.9236 268.9928,-764.2344 227,-746 102.1711,-691.7958 0,-676.5895 0,-540.5 0,-540.5 0,-540.5 0,-473.5 0,-294.5481 243.4097,-199.0108 355.131,-164.5855"/>
-<polygon fill="#191970" stroke="#191970" points="697.1098,-791.6941 707.2747,-788.7067 697.4654,-784.7031 697.1098,-791.6941"/>
+<path fill="none" stroke="#191970" d="M801.1836,-789.8125C618.2436,-784.3732 175.5199,-768.9959 112,-746 53.8677,-724.9545 0,-725.3245 0,-663.5 0,-663.5 0,-663.5 0,-473.5 0,-432.7508 10.9851,-419.6117 40,-391 196.4713,-236.7031 462.1077,-178.7646 591.9681,-158.8881"/>
+<polygon fill="#191970" stroke="#191970" points="801.1775,-793.3138 811.2765,-790.1104 801.3841,-786.3168 801.1775,-793.3138"/>
</g>
<!-- Node27 -->
<g id="node8" class="node">
<title>Node27</title>
<g id="a_node8"><a xlink:href="ir_2adt_8h.html" target="_top" xlink:title="Algebraic data type definitions. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="360,-531 360,-550 472,-550 472,-531 360,-531"/>
-<text text-anchor="middle" x="416" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/adt.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="104,-531 104,-550 216,-550 216,-531 104,-531"/>
+<text text-anchor="middle" x="160" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/adt.h</text>
</a>
</g>
</g>
<!-- Node16->Node27 -->
<g id="edge7" class="edge">
<title>Node16->Node27</title>
-<path fill="none" stroke="#191970" d="M697.4423,-790.319C565.7696,-786.6811 312.1788,-775.8441 285,-746 224.1783,-679.2138 355.9197,-580.8935 401.3383,-550.0888"/>
-<polygon fill="#191970" stroke="#191970" points="697.3884,-793.8187 707.4789,-790.5889 697.5766,-786.8213 697.3884,-793.8187"/>
+<path fill="none" stroke="#191970" d="M801.1695,-790.0189C613.7891,-784.9263 155.6285,-770.0648 133,-746 80.0026,-689.6386 135.0684,-582.9999 154.066,-550.2878"/>
+<polygon fill="#191970" stroke="#191970" points="801.1421,-793.5193 811.2328,-790.29 801.3307,-786.5219 801.1421,-793.5193"/>
</g>
<!-- Node145 -->
<g id="node9" class="node">
<title>Node145</title>
<g id="a_node9"><a xlink:href="ir_2expr_8h.html" target="_top" xlink:title="Base expr nodes in TVM. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="660.5,-592.5 660.5,-611.5 777.5,-611.5 777.5,-592.5 660.5,-592.5"/>
-<text text-anchor="middle" x="719" y="-599.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/expr.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="694.5,-592.5 694.5,-611.5 811.5,-611.5 811.5,-592.5 694.5,-592.5"/>
+<text text-anchor="middle" x="753" y="-599.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/expr.h</text>
</a>
</g>
</g>
<!-- Node16->Node145 -->
<g id="edge8" class="edge">
<title>Node16->Node145</title>
-<path fill="none" stroke="#191970" d="M788.1247,-773.7224C792.8179,-765.5929 797.6257,-755.6836 800,-746 803.281,-732.6186 804.2156,-728.117 800,-715 785.8938,-671.1083 747.7476,-629.8092 729.2858,-611.6665"/>
-<polygon fill="#191970" stroke="#191970" points="785.0881,-771.9786 782.8208,-782.328 791.0472,-775.6514 785.0881,-771.9786"/>
+<path fill="none" stroke="#191970" d="M867.926,-773.9365C841.6321,-734.5992 779.9244,-642.2806 759.6118,-611.8917"/>
+<polygon fill="#191970" stroke="#191970" points="865.0605,-775.9479 873.5275,-782.3167 870.8801,-772.0579 865.0605,-775.9479"/>
</g>
<!-- Node154 -->
<g id="node25" class="node">
<title>Node154</title>
<g id="a_node25"><a xlink:href="arg__info_8h.html" target="_top" xlink:title="include/tvm/meta_schedule\l/arg_info.h">
-<polygon fill="#ffffff" stroke="#ff0000" points="1546,-257.5 1546,-287.5 1698,-287.5 1698,-257.5 1546,-257.5"/>
-<text text-anchor="start" x="1554" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/meta_schedule</text>
-<text text-anchor="middle" x="1622" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/arg_info.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1372,-268.5 1372,-298.5 1524,-298.5 1524,-268.5 1372,-268.5"/>
+<text text-anchor="start" x="1380" y="-286.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/meta_schedule</text>
+<text text-anchor="middle" x="1448" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/arg_info.h</text>
</a>
</g>
</g>
<!-- Node16->Node154 -->
-<g id="edge93" class="edge">
+<g id="edge94" class="edge">
<title>Node16->Node154</title>
-<path fill="none" stroke="#191970" d="M854.5644,-785.0421C980.1176,-771.7267 1212,-737.1634 1212,-663.5 1212,-663.5 1212,-663.5 1212,-602 1212,-565.7124 1258.8699,-478.223 1289,-458 1353.9206,-414.4259 1586.6174,-480.0469 1639,-422 1672.9513,-384.3774 1645.6824,-317.9217 1630.4826,-287.939"/>
-<polygon fill="#191970" stroke="#191970" points="854.1891,-781.5622 844.6046,-786.0775 854.913,-788.5246 854.1891,-781.5622"/>
+<path fill="none" stroke="#191970" d="M947.9746,-780.1749C1009.9812,-765.0243 1092,-732.0314 1092,-663.5 1092,-663.5 1092,-663.5 1092,-602 1092,-537.1888 1067.5782,-502.1777 1115,-458 1181.9217,-395.6565 1463.22,-489.4423 1525,-422 1534.3065,-411.8405 1530.6053,-403.586 1525,-391 1515.2694,-369.1513 1499.6219,-373.9275 1485,-355 1471.3286,-337.3028 1460.2933,-313.7974 1453.9394,-298.6571"/>
+<polygon fill="#191970" stroke="#191970" points="947.1519,-776.7727 938.2103,-782.456 948.7444,-783.5892 947.1519,-776.7727"/>
</g>
<!-- Node37 -->
-<g id="node30" class="node">
+<g id="node31" class="node">
<title>Node37</title>
-<g id="a_node30"><a xlink:href="target_8h.html" target="_top" xlink:title="Compilation target object. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1297.5,-525.5 1297.5,-555.5 1404.5,-555.5 1404.5,-525.5 1297.5,-525.5"/>
-<text text-anchor="start" x="1305.5" y="-543.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
-<text text-anchor="middle" x="1351" y="-532.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/target.h</text>
+<g id="a_node31"><a xlink:href="target_8h.html" target="_top" xlink:title="Compilation target object. ">
+<polygon fill="#ffffff" stroke="#ff0000" points="1296.5,-525.5 1296.5,-555.5 1403.5,-555.5 1403.5,-525.5 1296.5,-525.5"/>
+<text text-anchor="start" x="1304.5" y="-543.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
+<text text-anchor="middle" x="1350" y="-532.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/target.h</text>
</a>
</g>
</g>
<!-- Node16->Node37 -->
-<g id="edge96" class="edge">
+<g id="edge97" class="edge">
<title>Node16->Node37</title>
-<path fill="none" stroke="#191970" d="M854.8247,-789.7953C979.4535,-785.5089 1215.9747,-773.8967 1295,-746 1345.4469,-728.1917 1369.811,-726.1968 1395,-679 1417.3326,-637.1553 1383.1347,-582.0761 1363.3914,-555.7676"/>
-<polygon fill="#191970" stroke="#191970" points="854.4618,-786.3054 844.5852,-790.1396 854.6972,-793.3015 854.4618,-786.3054"/>
+<path fill="none" stroke="#191970" d="M958.7606,-787.523C1092.1427,-778.0315 1353.0089,-750.5568 1404,-679 1432.3068,-639.2765 1389.3837,-582.7165 1364.9771,-555.8319"/>
+<polygon fill="#191970" stroke="#191970" points="958.3349,-784.044 948.6015,-788.2286 958.82,-791.0272 958.3349,-784.044"/>
</g>
<!-- Node72 -->
-<g id="node32" class="node">
+<g id="node33" class="node">
<title>Node72</title>
-<g id="a_node32"><a xlink:href="tag_8h.html" target="_top" xlink:title="Target tag registry. ">
-<polygon fill="#ffffff" stroke="#000000" points="1297.5,-458.5 1297.5,-488.5 1404.5,-488.5 1404.5,-458.5 1297.5,-458.5"/>
-<text text-anchor="start" x="1305.5" y="-476.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
-<text text-anchor="middle" x="1351" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/tag.h</text>
+<g id="a_node33"><a xlink:href="tag_8h.html" target="_top" xlink:title="Target tag registry. ">
+<polygon fill="#ffffff" stroke="#000000" points="1287.5,-458.5 1287.5,-488.5 1394.5,-488.5 1394.5,-458.5 1287.5,-458.5"/>
+<text text-anchor="start" x="1295.5" y="-476.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
+<text text-anchor="middle" x="1341" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/tag.h</text>
</a>
</g>
</g>
<!-- Node16->Node72 -->
-<g id="edge95" class="edge">
+<g id="edge96" class="edge">
<title>Node16->Node72</title>
-<path fill="none" stroke="#191970" d="M854.5876,-788.6953C975.9887,-782.8922 1198.1279,-769.2289 1226,-746 1255.3347,-721.5521 1250,-701.6867 1250,-663.5 1250,-663.5 1250,-663.5 1250,-602 1250,-552.0594 1297.7383,-510.0508 1327.5747,-488.6724"/>
-<polygon fill="#191970" stroke="#191970" points="854.4185,-785.1993 844.5944,-789.1661 854.748,-792.1916 854.4185,-785.1993"/>
+<path fill="none" stroke="#191970" d="M958.778,-781.6635C1013.7328,-773.393 1081.4482,-760.7616 1106,-746 1215.0495,-680.4345 1201.7952,-618.5914 1288,-525 1299.6474,-512.3546 1313.6227,-498.837 1324.2977,-488.825"/>
+<polygon fill="#191970" stroke="#191970" points="958.0592,-778.2317 948.6786,-783.1566 959.0829,-785.1564 958.0592,-778.2317"/>
</g>
-<!-- Node16->Node171 -->
-<g id="edge99" class="edge">
-<title>Node16->Node171</title>
-<path fill="none" stroke="#191970" d="M829.4931,-780.0778C914.0251,-759.6709 1069,-715.2073 1069,-663.5 1069,-663.5 1069,-663.5 1069,-602 1069,-555.4206 1043.2772,-504.0243 1031.6215,-483.1263"/>
-<polygon fill="#191970" stroke="#191970" points="828.4658,-776.7245 819.5487,-782.446 830.0875,-783.534 828.4658,-776.7245"/>
+<!-- Node16->Node172 -->
+<g id="edge100" class="edge">
+<title>Node16->Node172</title>
+<path fill="none" stroke="#191970" d="M910.0551,-777.8654C947.325,-758.0095 1006,-718.0824 1006,-663.5 1006,-663.5 1006,-663.5 1006,-602 1006,-557.6037 1006.6043,-504.5819 1006.8733,-483.1717"/>
+<polygon fill="#191970" stroke="#191970" points="908.3807,-774.7907 901.0896,-782.4777 911.583,-781.0153 908.3807,-774.7907"/>
</g>
-<!-- Node172 -->
-<g id="node34" class="node">
-<title>Node172</title>
-<g id="a_node34"><a xlink:href="relay_2base_8h.html" target="_top" xlink:title="Base classes for the Relay IR. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="28.5,-397 28.5,-416 165.5,-416 165.5,-397 28.5,-397"/>
-<text text-anchor="middle" x="97" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/base.h</text>
+<!-- Node173 -->
+<g id="node35" class="node">
+<title>Node173</title>
+<g id="a_node35"><a xlink:href="relay_2base_8h.html" target="_top" xlink:title="Base classes for the Relay IR. ">
+<polygon fill="#ffffff" stroke="#ff0000" points="104.5,-397 104.5,-416 241.5,-416 241.5,-397 104.5,-397"/>
+<text text-anchor="middle" x="173" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/base.h</text>
</a>
</g>
</g>
-<!-- Node16->Node172 -->
-<g id="edge94" class="edge">
-<title>Node16->Node172</title>
-<path fill="none" stroke="#191970" d="M697.2391,-788.9948C563.2328,-783.2785 301.914,-769.2871 267,-746 143.3391,-663.5199 105.6287,-462.924 98.3983,-416.2965"/>
-<polygon fill="#191970" stroke="#191970" points="697.3086,-792.5008 707.4471,-789.4248 697.6032,-785.507 697.3086,-792.5008"/>
+<!-- Node16->Node173 -->
+<g id="edge95" class="edge">
+<title>Node16->Node173</title>
+<path fill="none" stroke="#191970" d="M800.8678,-789.9776C704.3396,-786.2541 536.6341,-775.7184 396,-746 353.3722,-736.992 344.6997,-727.6257 303,-715 244.7999,-697.3784 220.9225,-713.7207 171,-679 108.3393,-635.42 73.0964,-598.1151 95,-525 108.6415,-479.4641 146.0474,-435.2354 163.632,-416.2244"/>
+<polygon fill="#191970" stroke="#191970" points="801.0084,-793.4851 811.1312,-790.358 801.2677,-786.4899 801.0084,-793.4851"/>
</g>
<!-- Node179 -->
<g id="node42" class="node">
<title>Node179</title>
<g id="a_node42"><a xlink:href="var_8h.html" target="_top" xlink:title="Variables in the TIR. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="925.5,-531 925.5,-550 1040.5,-550 1040.5,-531 925.5,-531"/>
-<text text-anchor="middle" x="983" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/var.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="857.5,-531 857.5,-550 972.5,-550 972.5,-531 857.5,-531"/>
+<text text-anchor="middle" x="915" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/var.h</text>
</a>
</g>
</g>
<!-- Node16->Node179 -->
-<g id="edge100" class="edge">
+<g id="edge101" class="edge">
<title>Node16->Node179</title>
-<path fill="none" stroke="#191970" d="M791.2915,-774.6909C798.8035,-766.118 807.9489,-755.5766 816,-746 878.7174,-671.3995 952.0862,-579.426 975.454,-550.0139"/>
-<polygon fill="#191970" stroke="#191970" points="788.6537,-772.3904 784.6788,-782.2113 793.9106,-777.0127 788.6537,-772.3904"/>
+<path fill="none" stroke="#191970" d="M882.7208,-772.4493C889.7839,-721.6954 908.5128,-587.1153 913.6694,-550.0615"/>
+<polygon fill="#191970" stroke="#191970" points="879.2412,-772.061 881.3293,-782.448 886.1743,-773.0259 879.2412,-772.061"/>
</g>
<!-- Node181 -->
<g id="node43" class="node">
<title>Node181</title>
<g id="a_node43"><a xlink:href="ir_2span_8h.html" target="_top" xlink:title="Span information for debugging purposes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="672.5,-721 672.5,-740 791.5,-740 791.5,-721 672.5,-721"/>
-<text text-anchor="middle" x="732" y="-728" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/span.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="693.5,-721 693.5,-740 812.5,-740 812.5,-721 693.5,-721"/>
+<text text-anchor="middle" x="753" y="-728" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/span.h</text>
</a>
</g>
</g>
<!-- Node16->Node181 -->
-<g id="edge84" class="edge">
+<g id="edge85" class="edge">
<title>Node16->Node181</title>
-<path fill="none" stroke="#191970" d="M763.2983,-774.2465C755.379,-763.1775 745.4899,-749.3552 738.9719,-740.2449"/>
-<polygon fill="#191970" stroke="#191970" points="760.4598,-776.2943 769.125,-782.3906 766.1528,-772.2212 760.4598,-776.2943"/>
+<path fill="none" stroke="#191970" d="M850.8443,-777.8813C826.8937,-766.2832 793.6195,-750.1701 772.7044,-740.0419"/>
+<polygon fill="#191970" stroke="#191970" points="849.6305,-781.1823 860.1563,-782.3906 852.6814,-774.8821 849.6305,-781.1823"/>
</g>
<!-- Node182 -->
<g id="node44" class="node">
<title>Node182</title>
<g id="a_node44"><a xlink:href="ir_2type_8h.html" target="_top" xlink:title="IR/AST nodes for the unified type system in TVM. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="438,-654 438,-673 556,-673 556,-654 438,-654"/>
-<text text-anchor="middle" x="497" y="-661" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/type.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="180,-654 180,-673 298,-673 298,-654 180,-654"/>
+<text text-anchor="middle" x="239" y="-661" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/type.h</text>
</a>
</g>
</g>
<!-- Node16->Node182 -->
-<g id="edge92" class="edge">
+<g id="edge93" class="edge">
<title>Node16->Node182</title>
-<path fill="none" stroke="#191970" d="M697.3776,-784.4314C618.3,-776.007 506.2023,-761.5215 493,-746 475.4404,-725.3557 485.9889,-690.2266 492.7131,-673.2528"/>
-<polygon fill="#191970" stroke="#191970" points="697.0211,-787.9131 707.3325,-785.4786 697.7536,-780.9515 697.0211,-787.9131"/>
+<path fill="none" stroke="#191970" d="M801.0564,-785.9012C726.5021,-779.2428 611.8438,-766.6749 514,-746 421.0963,-726.3689 314.1795,-690.2704 265.55,-673.0833"/>
+<polygon fill="#191970" stroke="#191970" points="801.0362,-789.4128 811.3042,-786.8017 801.6491,-782.4397 801.0362,-789.4128"/>
</g>
<!-- Node36 -->
<g id="node45" class="node">
<title>Node36</title>
<g id="a_node45"><a xlink:href="target__kind_8h.html" target="_top" xlink:title="Target kind registry. ">
-<polygon fill="#ffffff" stroke="#000000" points="1278.5,-648.5 1278.5,-678.5 1385.5,-678.5 1385.5,-648.5 1278.5,-648.5"/>
-<text text-anchor="start" x="1286.5" y="-666.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
-<text text-anchor="middle" x="1332" y="-655.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/target_kind.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1287.5,-648.5 1287.5,-678.5 1394.5,-678.5 1394.5,-648.5 1287.5,-648.5"/>
+<text text-anchor="start" x="1295.5" y="-666.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
+<text text-anchor="middle" x="1341" y="-655.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/target_kind.h</text>
</a>
</g>
</g>
<!-- Node16->Node36 -->
-<g id="edge97" class="edge">
+<g id="edge98" class="edge">
<title>Node16->Node36</title>
-<path fill="none" stroke="#191970" d="M854.7082,-787.8883C983.5563,-780.6669 1228.7755,-764.6851 1264,-746 1292.5309,-730.8656 1313.6919,-698.1916 1324.3209,-678.8126"/>
-<polygon fill="#191970" stroke="#191970" points="854.293,-784.4059 844.5027,-788.4556 854.6815,-791.3951 854.293,-784.4059"/>
+<path fill="none" stroke="#191970" d="M958.8061,-784.4788C1015.1528,-777.7748 1092.3195,-765.8691 1158,-746 1215.5412,-728.5932 1278.6268,-697.1861 1313.6585,-678.5498"/>
+<polygon fill="#191970" stroke="#191970" points="958.0645,-781.0413 948.5346,-785.6708 958.8715,-787.9946 958.0645,-781.0413"/>
</g>
<!-- Node145->Node27 -->
<g id="edge25" class="edge">
<title>Node145->Node27</title>
-<path fill="none" stroke="#191970" d="M661.9744,-590.4255C604.4872,-578.7573 516.7971,-560.9588 463.1216,-550.0643"/>
-<polygon fill="#191970" stroke="#191970" points="661.4923,-593.8989 671.9887,-592.4581 662.8848,-587.0388 661.4923,-593.8989"/>
+<path fill="none" stroke="#191970" d="M684.2034,-594.8651C565.7292,-582.5782 325.4237,-557.6561 216.2436,-546.333"/>
+<polygon fill="#191970" stroke="#191970" points="683.938,-598.3563 694.2457,-595.9066 684.6601,-591.3936 683.938,-598.3563"/>
</g>
<!-- Node147 -->
<g id="node10" class="node">
<title>Node147</title>
<g id="a_node10"><a xlink:href="bound_8h.html" target="_top" xlink:title="Bound deducers. ">
-<polygon fill="#ffffff" stroke="#000000" points="694,-330 694,-349 836,-349 836,-330 694,-330"/>
-<text text-anchor="middle" x="765" y="-337" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/bound.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="768,-335.5 768,-354.5 910,-354.5 910,-335.5 768,-335.5"/>
+<text text-anchor="middle" x="839" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/bound.h</text>
</a>
</g>
</g>
<!-- Node145->Node147 -->
<g id="edge9" class="edge">
<title>Node145->Node147</title>
-<path fill="none" stroke="#191970" d="M688.8811,-588.1362C669.6452,-579.1788 644.2321,-567.1532 622,-556 538.8326,-514.2775 519.7283,-500.5865 437,-458 405.1902,-441.6251 383.8091,-452.4338 365,-422 357.7566,-410.2799 355.9009,-401.3457 365,-391 375.7366,-378.7925 581.4825,-357.1867 693.5932,-346.2481"/>
-<polygon fill="#191970" stroke="#191970" points="687.5125,-591.3596 698.0567,-592.3946 690.4594,-585.0101 687.5125,-591.3596"/>
+<path fill="none" stroke="#191970" d="M747.6432,-582.3336C738.0785,-542.8355 722.3033,-452.401 759,-391 769.6027,-373.2596 789.8336,-361.7178 807.297,-354.6322"/>
+<polygon fill="#191970" stroke="#191970" points="744.3136,-583.4406 750.1827,-592.2613 751.0953,-581.7058 744.3136,-583.4406"/>
</g>
<!-- Node148 -->
<g id="node19" class="node">
<title>Node148</title>
<g id="a_node19"><a xlink:href="int__set_8h.html" target="_top" xlink:title="Integer set. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="654,-391.5 654,-421.5 770,-421.5 770,-391.5 654,-391.5"/>
-<text text-anchor="start" x="662" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/int</text>
-<text text-anchor="middle" x="712" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_set.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="768,-391.5 768,-421.5 884,-421.5 884,-391.5 768,-391.5"/>
+<text text-anchor="start" x="776" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/int</text>
+<text text-anchor="middle" x="826" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_set.h</text>
</a>
</g>
</g>
<!-- Node145->Node148 -->
<g id="edge21" class="edge">
<title>Node145->Node148</title>
-<path fill="none" stroke="#191970" d="M718.2909,-582.1969C716.89,-543.0716 713.7949,-456.6286 712.5422,-421.6418"/>
-<polygon fill="#191970" stroke="#191970" points="714.7972,-582.4368 718.6529,-592.3051 721.7927,-582.1863 714.7972,-582.4368"/>
+<path fill="none" stroke="#191970" d="M759.4972,-582.9116C764.8608,-567.2917 772.7429,-544.6509 780,-525 793.7616,-487.7364 810.9126,-444.2902 819.8832,-421.7782"/>
+<polygon fill="#191970" stroke="#191970" points="756.149,-581.886 756.2249,-592.4806 762.7724,-584.151 756.149,-581.886"/>
</g>
<!-- Node115 -->
<g id="node20" class="node">
<title>Node115</title>
<g id="a_node20"><a xlink:href="int__solver_8h.html" target="_top" xlink:title="integer constraints data structures and solvers ">
-<polygon fill="#ffffff" stroke="#000000" points="580,-257.5 580,-287.5 696,-287.5 696,-257.5 580,-257.5"/>
-<text text-anchor="start" x="588" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/int</text>
-<text text-anchor="middle" x="638" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_solver.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="312,-268.5 312,-298.5 428,-298.5 428,-268.5 312,-268.5"/>
+<text text-anchor="start" x="320" y="-286.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/int</text>
+<text text-anchor="middle" x="370" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_solver.h</text>
</a>
</g>
</g>
<!-- Node145->Node115 -->
<g id="edge23" class="edge">
<title>Node145->Node115</title>
-<path fill="none" stroke="#191970" d="M666.4036,-590.0195C634.6975,-582.0307 593.8408,-570.3656 559,-556 508.8566,-535.3249 378.2429,-471.0061 355,-422 310.1147,-327.3625 484.7667,-291.0377 579.9889,-278.4616"/>
-<polygon fill="#191970" stroke="#191970" points="665.809,-593.4779 676.357,-592.4829 667.4907,-586.6829 665.809,-593.4779"/>
+<path fill="none" stroke="#191970" d="M684.3135,-597.4457C601.905,-591.1078 469.2974,-577.8734 425,-556 386.9575,-537.2152 305.3814,-461.1376 289,-422 269.1032,-374.4636 320.2643,-323.303 350.1732,-298.5813"/>
+<polygon fill="#191970" stroke="#191970" points="684.1181,-600.9408 694.353,-598.2029 684.6446,-593.9607 684.1181,-600.9408"/>
</g>
<!-- Node149 -->
<g id="node21" class="node">
<title>Node149</title>
<g id="a_node21"><a xlink:href="pattern_8h.html" target="_top" xlink:title="Expression pattern detectors. ">
-<polygon fill="#ffffff" stroke="#000000" points="374.5,-397 374.5,-416 521.5,-416 521.5,-397 374.5,-397"/>
-<text text-anchor="middle" x="448" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/pattern.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="298.5,-397 298.5,-416 445.5,-416 445.5,-397 298.5,-397"/>
+<text text-anchor="middle" x="372" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/arith/pattern.h</text>
</a>
</g>
</g>
<!-- Node145->Node149 -->
<g id="edge24" class="edge">
<title>Node145->Node149</title>
-<path fill="none" stroke="#191970" d="M699.6084,-585.7948C666.3814,-558.3025 596.0416,-501.2931 533,-458 510.1005,-442.2741 482.3848,-425.959 464.9681,-416.0157"/>
-<polygon fill="#191970" stroke="#191970" points="697.666,-588.7312 707.5953,-592.4271 702.138,-583.3459 697.666,-588.7312"/>
+<path fill="none" stroke="#191970" d="M684.1394,-597.9437C609.2701,-592.3729 495.5306,-580.1648 460,-556 407.9513,-520.6011 382.2347,-443.3573 374.5462,-416.1876"/>
+<polygon fill="#191970" stroke="#191970" points="684.2077,-601.4576 694.4337,-598.6863 684.7114,-594.4758 684.2077,-601.4576"/>
</g>
<!-- Node151 -->
<g id="node22" class="node">
<title>Node151</title>
<g id="a_node22"><a xlink:href="ir_2attrs_8h.html" target="_top" xlink:title="Helpers for attribute objects. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="222.5,-531 222.5,-550 341.5,-550 341.5,-531 222.5,-531"/>
-<text text-anchor="middle" x="282" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/attrs.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1120.5,-531 1120.5,-550 1239.5,-550 1239.5,-531 1120.5,-531"/>
+<text text-anchor="middle" x="1180" y="-538" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/attrs.h</text>
</a>
</g>
</g>
<!-- Node145->Node151 -->
<g id="edge26" class="edge">
<title>Node145->Node151</title>
-<path fill="none" stroke="#191970" d="M650.2809,-595.4149C576.1364,-587.8432 454.7422,-574.0925 351,-556 341.3239,-554.3125 330.9571,-552.166 321.2762,-550.0122"/>
-<polygon fill="#191970" stroke="#191970" points="650.1208,-598.9165 660.4226,-596.4421 650.8262,-591.9521 650.1208,-598.9165"/>
+<path fill="none" stroke="#191970" d="M821.981,-592.0648C904.6948,-580.1517 1041.9533,-560.3826 1120.3071,-549.0975"/>
+<polygon fill="#191970" stroke="#191970" points="821.3231,-588.6233 811.9242,-593.5133 822.3211,-595.5519 821.3231,-588.6233"/>
</g>
<!-- Node152 -->
<g id="node23" class="node">
<title>Node152</title>
<g id="a_node23"><a xlink:href="ir_2function_8h.html" target="_top" xlink:title="Function nodes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1105,-464 1105,-483 1241,-483 1241,-464 1105,-464"/>
-<text text-anchor="middle" x="1173" y="-471" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/function.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1124,-464 1124,-483 1260,-483 1260,-464 1124,-464"/>
+<text text-anchor="middle" x="1192" y="-471" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/function.h</text>
</a>
</g>
</g>
<!-- Node145->Node152 -->
-<g id="edge35" class="edge">
+<g id="edge37" class="edge">
<title>Node145->Node152</title>
-<path fill="none" stroke="#191970" d="M787.8353,-596.0395C870.1398,-588.2812 1002.6008,-573.6126 1049,-556 1095.4143,-538.3817 1141.9572,-500.7846 1162.3277,-483.0872"/>
-<polygon fill="#191970" stroke="#191970" points="787.4405,-592.561 777.8084,-596.9739 788.0901,-599.5308 787.4405,-592.561"/>
+<path fill="none" stroke="#191970" d="M774.745,-586.3654C786.562,-577.6803 801.3115,-566.53 814,-556 829.7362,-542.9408 829.533,-533.783 848,-525 860.3384,-519.1318 1027.1928,-495.8171 1123.9,-482.6598"/>
+<polygon fill="#191970" stroke="#191970" points="772.3822,-583.7566 766.3629,-592.4754 776.5056,-589.4133 772.3822,-583.7566"/>
</g>
<!-- Node145->Node37 -->
-<g id="edge36" class="edge">
+<g id="edge38" class="edge">
<title>Node145->Node37</title>
-<path fill="none" stroke="#191970" d="M787.5882,-595.3257C914.3575,-582.9897 1182.8272,-556.8649 1297.3807,-545.7177"/>
-<polygon fill="#191970" stroke="#191970" points="787.2205,-591.8448 777.6066,-596.297 787.8986,-598.8119 787.2205,-591.8448"/>
+<path fill="none" stroke="#191970" d="M821.6752,-596.9364C917.5375,-589.5711 1096.8251,-574.7163 1249,-556 1264.4064,-554.1051 1281.0804,-551.6948 1296.3557,-549.3427"/>
+<polygon fill="#191970" stroke="#191970" points="821.2381,-593.4595 811.5339,-597.7115 821.7716,-600.4392 821.2381,-593.4595"/>
</g>
-<!-- Node145->Node171 -->
-<g id="edge39" class="edge">
-<title>Node145->Node171</title>
-<path fill="none" stroke="#191970" d="M775.1861,-590.0258C805.134,-582.4671 842.0737,-571.2243 873,-556 894.1374,-545.5946 895.6874,-536.9357 916,-525 944.5431,-508.2281 979.4488,-492.6581 1002.1776,-483.1239"/>
-<polygon fill="#191970" stroke="#191970" points="774.194,-586.6653 765.3167,-592.4483 775.8628,-593.4635 774.194,-586.6653"/>
+<!-- Node145->Node172 -->
+<g id="edge41" class="edge">
+<title>Node145->Node172</title>
+<path fill="none" stroke="#191970" d="M763.1862,-583.5044C773.8432,-565.8698 792.19,-539.8623 815,-525 854.4694,-499.2829 906.435,-486.349 946.3495,-479.8776"/>
+<polygon fill="#191970" stroke="#191970" points="760.0653,-581.9094 758.0773,-592.3161 766.121,-585.4205 760.0653,-581.9094"/>
</g>
<!-- Node114 -->
<g id="node40" class="node">
<title>Node114</title>
<g id="a_node40"><a xlink:href="tir_2op_8h.html" target="_top" xlink:title="Common operators defined for Expr. ">
-<polygon fill="#ffffff" stroke="#000000" points="930,-330 930,-349 1042,-349 1042,-330 930,-330"/>
-<text text-anchor="middle" x="986" y="-337" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/op.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="562,-335.5 562,-354.5 674,-354.5 674,-335.5 562,-335.5"/>
+<text text-anchor="middle" x="618" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/op.h</text>
</a>
</g>
</g>
<!-- Node145->Node114 -->
-<g id="edge81" class="edge">
+<g id="edge82" class="edge">
<title>Node145->Node114</title>
-<path fill="none" stroke="#191970" d="M736.1056,-585.1827C786.9651,-535.1803 936.1871,-388.4733 976.0914,-349.2416"/>
-<polygon fill="#191970" stroke="#191970" points="733.4965,-582.8396 728.8193,-592.3462 738.404,-587.8312 733.4965,-582.8396"/>
+<path fill="none" stroke="#191970" d="M740.8358,-583.4032C720.1524,-551.3188 677.5078,-483.1177 648,-422 636.6473,-398.4859 626.2738,-369.4876 621.2252,-354.6878"/>
+<polygon fill="#191970" stroke="#191970" points="738.1328,-585.6678 746.5108,-592.1531 744.0057,-581.8587 738.1328,-585.6678"/>
</g>
<!-- Node145->Node179 -->
-<g id="edge82" class="edge">
+<g id="edge83" class="edge">
<title>Node145->Node179</title>
-<path fill="none" stroke="#191970" d="M769.9355,-590.1343C820.037,-578.463 895.5641,-560.8686 941.9436,-550.0643"/>
-<polygon fill="#191970" stroke="#191970" points="768.9055,-586.7805 759.9603,-592.4581 770.4937,-593.598 768.9055,-586.7805"/>
+<path fill="none" stroke="#191970" d="M787.7725,-588.7993C818.4669,-577.1468 862.4747,-560.4402 889.8652,-550.0419"/>
+<polygon fill="#191970" stroke="#191970" points="786.4193,-585.5693 778.3125,-592.3906 788.9037,-592.1136 786.4193,-585.5693"/>
</g>
<!-- Node116 -->
<g id="node11" class="node">
<title>Node116</title>
<g id="a_node11"><a xlink:href="tensor_8h.html" target="_top" xlink:title="Dataflow tensor object. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="714.5,-263 714.5,-282 843.5,-282 843.5,-263 714.5,-263"/>
-<text text-anchor="middle" x="779" y="-270" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/tensor.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="771.5,-274 771.5,-293 900.5,-293 900.5,-274 771.5,-274"/>
+<text text-anchor="middle" x="836" y="-281" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/tensor.h</text>
</a>
</g>
</g>
<!-- Node147->Node116 -->
<g id="edge10" class="edge">
<title>Node147->Node116</title>
-<path fill="none" stroke="#191970" d="M769.0775,-319.9863C771.6806,-307.5286 774.9409,-291.9258 776.9979,-282.0817"/>
-<polygon fill="#191970" stroke="#191970" points="765.6253,-319.396 767.0059,-329.9005 772.4773,-320.8279 765.6253,-319.396"/>
+<path fill="none" stroke="#191970" d="M838.0364,-325.2462C837.5147,-314.5519 836.8934,-301.814 836.4754,-293.2449"/>
+<polygon fill="#191970" stroke="#191970" points="834.5481,-325.5731 838.5313,-335.3906 841.5398,-325.2319 834.5481,-325.5731"/>
</g>
<!-- Node117 -->
<g id="node12" class="node">
<title>Node117</title>
<g id="a_node12"><a xlink:href="autodiff_8h.html" target="_top" xlink:title="Automatic differentiation of tensor expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="543.5,-201.5 543.5,-220.5 680.5,-220.5 680.5,-201.5 543.5,-201.5"/>
-<text text-anchor="middle" x="612" y="-208.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/autodiff.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="860.5,-207 860.5,-226 997.5,-226 997.5,-207 860.5,-207"/>
+<text text-anchor="middle" x="929" y="-214" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/autodiff.h</text>
</a>
</g>
</g>
<!-- Node116->Node117 -->
<g id="edge11" class="edge">
<title>Node116->Node117</title>
-<path fill="none" stroke="#191970" d="M743.1543,-259.2993C711.5125,-247.6468 666.1465,-230.9402 637.9105,-220.5419"/>
-<polygon fill="#191970" stroke="#191970" points="742.3128,-262.7192 752.9063,-262.8906 744.7319,-256.1504 742.3128,-262.7192"/>
+<path fill="none" stroke="#191970" d="M857.5008,-268.0102C875.4667,-255.067 900.5176,-237.0195 915.7001,-226.0817"/>
+<polygon fill="#191970" stroke="#191970" points="855.3925,-265.2153 849.3246,-273.9005 859.4842,-270.8949 855.3925,-265.2153"/>
</g>
<!-- Node118 -->
<g id="node13" class="node">
<title>Node118</title>
<g id="a_node13"><a xlink:href="operation_8h.html" target="_top" xlink:title="Operation node can generate one or multiple Tensors. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="859,-140 859,-159 1005,-159 1005,-140 859,-140"/>
-<text text-anchor="middle" x="932" y="-147" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/operation.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="795,-140 795,-159 941,-159 941,-140 795,-140"/>
+<text text-anchor="middle" x="868" y="-147" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/operation.h</text>
</a>
</g>
</g>
<!-- Node116->Node118 -->
<g id="edge12" class="edge">
<title>Node116->Node118</title>
-<path fill="none" stroke="#191970" d="M755.4029,-256.8098C736.2106,-242.0998 714.4973,-219.6414 729,-201 745.1028,-180.3019 807.5122,-166.578 858.9301,-158.5405"/>
-<polygon fill="#191970" stroke="#191970" points="753.4859,-259.7433 763.6321,-262.7937 757.6027,-254.0818 753.4859,-259.7433"/>
+<path fill="none" stroke="#191970" d="M838.8397,-263.4958C841.4159,-246.6838 845.6294,-222.0908 851,-201 854.764,-186.2186 860.5719,-169.5216 864.3295,-159.2528"/>
+<polygon fill="#191970" stroke="#191970" points="835.3234,-263.3488 837.3236,-273.7531 842.2482,-264.3724 835.3234,-263.3488"/>
</g>
<!-- Node128 -->
<g id="node18" class="node">
<title>Node128</title>
<g id="a_node18"><a xlink:href="te_2schedule_8h.html" target="_top" xlink:title="Define a schedule. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="738.5,-201.5 738.5,-220.5 881.5,-220.5 881.5,-201.5 738.5,-201.5"/>
-<text text-anchor="middle" x="810" y="-208.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/schedule.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="660.5,-207 660.5,-226 803.5,-226 803.5,-207 660.5,-207"/>
+<text text-anchor="middle" x="732" y="-214" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/te/schedule.h</text>
</a>
</g>
</g>
<!-- Node116->Node128 -->
<g id="edge18" class="edge">
<title>Node116->Node128</title>
-<path fill="none" stroke="#191970" d="M788.4497,-253.753C793.9402,-242.8606 800.6321,-229.5846 805.088,-220.7449"/>
-<polygon fill="#191970" stroke="#191970" points="785.2196,-252.3855 783.8438,-262.8906 791.4704,-255.5363 785.2196,-252.3855"/>
+<path fill="none" stroke="#191970" d="M812.502,-268.3619C792.3755,-255.3957 764.0056,-237.119 746.873,-226.0817"/>
+<polygon fill="#191970" stroke="#191970" points="810.7973,-271.427 821.0993,-273.9005 814.5883,-265.5425 810.7973,-271.427"/>
</g>
<!-- Node92 -->
<g id="node14" class="node">
@@ -585,583 +585,589 @@
<!-- Node118->Node92 -->
<g id="edge13" class="edge">
<title>Node118->Node92</title>
-<path fill="none" stroke="#191970" d="M979.0616,-137.4652C1022.7627,-126.2897 1087.9941,-109.6084 1135.1268,-97.5553"/>
-<polygon fill="#191970" stroke="#191970" points="978.081,-134.1032 969.26,-139.9717 979.8154,-140.885 978.081,-134.1032"/>
+<path fill="none" stroke="#191970" d="M924.346,-137.9197C981.7367,-126.1247 1070.5699,-107.8675 1130.4382,-95.5633"/>
+<polygon fill="#191970" stroke="#191970" points="923.4523,-134.5301 914.3616,-139.9717 924.8615,-141.3868 923.4523,-134.5301"/>
</g>
<!-- Node98 -->
<g id="node16" class="node">
<title>Node98</title>
<g id="a_node16"><a xlink:href="topi_2nn_8h.html" target="_top" xlink:title="NN op constructions. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="923,-6 923,-25 1043,-25 1043,-6 923,-6"/>
-<text text-anchor="middle" x="983" y="-13" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/nn.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1001,-6 1001,-25 1121,-25 1121,-6 1001,-6"/>
+<text text-anchor="middle" x="1061" y="-13" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/nn.h</text>
</a>
</g>
</g>
<!-- Node118->Node98 -->
<g id="edge17" class="edge">
<title>Node118->Node98</title>
-<path fill="none" stroke="#191970" d="M939.2712,-130.3952C950.3418,-101.3078 971.0174,-46.9838 979.3654,-25.0496"/>
-<polygon fill="#191970" stroke="#191970" points="935.9538,-129.2721 935.6678,-139.8631 942.496,-131.7621 935.9538,-129.2721"/>
+<path fill="none" stroke="#191970" d="M890.221,-134.072C930.458,-106.1354 1014.2684,-47.9458 1047.2457,-25.0496"/>
+<polygon fill="#191970" stroke="#191970" points="888.0982,-131.2849 881.8801,-139.8631 892.0905,-137.0348 888.0982,-131.2849"/>
</g>
<!-- Node121 -->
<g id="node17" class="node">
<title>Node121</title>
<g id="a_node17"><a xlink:href="pad__utils_8h.html" target="_top" xlink:title="Padding helpers. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="767.5,-67.5 767.5,-97.5 894.5,-97.5 894.5,-67.5 767.5,-67.5"/>
-<text text-anchor="start" x="775.5" y="-85.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/detail</text>
-<text text-anchor="middle" x="831" y="-74.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/pad_utils.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="804.5,-67.5 804.5,-97.5 931.5,-97.5 931.5,-67.5 804.5,-67.5"/>
+<text text-anchor="start" x="812.5" y="-85.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/detail</text>
+<text text-anchor="middle" x="868" y="-74.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/pad_utils.h</text>
</a>
</g>
</g>
<!-- Node118->Node121 -->
<g id="edge16" class="edge">
<title>Node118->Node121</title>
-<path fill="none" stroke="#191970" d="M908.9755,-134.2263C892.3961,-123.2281 870.1365,-108.4618 853.6447,-97.5218"/>
-<polygon fill="#191970" stroke="#191970" points="907.2612,-137.2892 917.5292,-139.9005 911.1308,-131.4559 907.2612,-137.2892"/>
+<path fill="none" stroke="#191970" d="M868,-129.7758C868,-119.4641 868,-107.0437 868,-97.5218"/>
+<polygon fill="#191970" stroke="#191970" points="864.5001,-129.9005 868,-139.9005 871.5001,-129.9006 864.5001,-129.9005"/>
</g>
<!-- Node103 -->
<g id="node15" class="node">
<title>Node103</title>
<g id="a_node15"><a xlink:href="strided__slice_8h.html" target="_top" xlink:title="Utility functions for strided_slice op. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1466.5,-.5 1466.5,-30.5 1593.5,-30.5 1593.5,-.5 1466.5,-.5"/>
-<text text-anchor="start" x="1474.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/detail</text>
-<text text-anchor="middle" x="1530" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/strided_slice.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1171.5,-.5 1171.5,-30.5 1298.5,-30.5 1298.5,-.5 1171.5,-.5"/>
+<text text-anchor="start" x="1179.5" y="-18.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/detail</text>
+<text text-anchor="middle" x="1235" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/strided_slice.h</text>
</a>
</g>
</g>
<!-- Node92->Node103 -->
<g id="edge14" class="edge">
<title>Node92->Node103</title>
-<path fill="none" stroke="#191970" d="M1267.6638,-67.8111C1327.0876,-55.9617 1409.8105,-39.4664 1466.3809,-28.186"/>
-<polygon fill="#191970" stroke="#191970" points="1266.7165,-64.431 1257.594,-69.819 1268.0854,-71.2958 1266.7165,-64.431"/>
+<path fill="none" stroke="#191970" d="M1208.6965,-58.4837C1214.3778,-49.1996 1220.7072,-38.8565 1225.7333,-30.6432"/>
+<polygon fill="#191970" stroke="#191970" points="1205.5378,-56.9401 1203.3035,-67.2967 1211.5086,-60.5939 1205.5378,-56.9401"/>
</g>
<!-- Node92->Node98 -->
<g id="edge15" class="edge">
<title>Node92->Node98</title>
-<path fill="none" stroke="#191970" d="M1137.0124,-64.4044C1097.0079,-51.7016 1045.1067,-35.2211 1012.9346,-25.0053"/>
-<polygon fill="#191970" stroke="#191970" points="1136.0572,-67.7732 1146.6475,-67.4639 1138.1757,-61.1015 1136.0572,-67.7732"/>
+<path fill="none" stroke="#191970" d="M1155.0333,-62.8702C1130.3609,-50.4412 1099.5475,-34.9187 1080.0645,-25.1039"/>
+<polygon fill="#191970" stroke="#191970" points="1153.6467,-66.0906 1164.1522,-67.4639 1156.7961,-59.8391 1153.6467,-66.0906"/>
</g>
<!-- Node128->Node20 -->
<g id="edge19" class="edge">
<title>Node128->Node20</title>
-<path fill="none" stroke="#191970" d="M737.7322,-199.8888C666.6759,-188.9639 558.6318,-172.3521 486.2361,-161.2213"/>
-<polygon fill="#191970" stroke="#191970" points="737.5231,-203.3977 747.9389,-201.4581 738.5869,-196.479 737.5231,-203.3977"/>
+<path fill="none" stroke="#191970" d="M715.6272,-199.3597C705.3485,-188.5992 692.2188,-174.8541 682.3491,-164.5218"/>
+<polygon fill="#191970" stroke="#191970" points="713.3921,-202.087 722.8304,-206.9005 718.4539,-197.2518 713.3921,-202.087"/>
</g>
<!-- Node128->Node118 -->
<g id="edge20" class="edge">
<title>Node128->Node118</title>
-<path fill="none" stroke="#191970" d="M838.0078,-196.8813C861.0155,-185.2832 892.9797,-169.1701 913.0714,-159.0419"/>
-<polygon fill="#191970" stroke="#191970" points="836.4166,-193.7638 829.0625,-201.3906 839.5676,-200.0145 836.4166,-193.7638"/>
+<path fill="none" stroke="#191970" d="M760.6182,-202.4013C787.0314,-189.389 825.541,-170.4173 848.5507,-159.0817"/>
+<polygon fill="#191970" stroke="#191970" points="758.9092,-199.3415 751.4855,-206.9005 762.0028,-205.6209 758.9092,-199.3415"/>
</g>
<!-- Node148->Node147 -->
<g id="edge22" class="edge">
<title>Node148->Node147</title>
-<path fill="none" stroke="#191970" d="M730.4436,-383.1845C739.7103,-371.47 750.4294,-357.9195 757.4416,-349.055"/>
-<polygon fill="#191970" stroke="#191970" points="727.4856,-381.2824 724.0265,-391.2967 732.9756,-385.6252 727.4856,-381.2824"/>
+<path fill="none" stroke="#191970" d="M831.33,-381.2849C833.2957,-371.9858 835.409,-361.9883 836.9185,-354.8469"/>
+<polygon fill="#191970" stroke="#191970" points="827.8574,-380.7901 829.2135,-391.2977 834.706,-382.2378 827.8574,-380.7901"/>
</g>
<!-- Node151->Node152 -->
<g id="edge27" class="edge">
<title>Node151->Node152</title>
-<path fill="none" stroke="#191970" d="M328.8952,-528.8276C336.2553,-527.3258 343.8065,-525.9746 351,-525 679.0565,-480.5551 766.4174,-529.3725 1095,-489 1106.7849,-487.552 1119.4545,-485.3276 1131.0843,-483.0119"/>
-<polygon fill="#191970" stroke="#191970" points="328.0703,-525.4249 319.0289,-530.9478 329.541,-532.2687 328.0703,-525.4249"/>
+<path fill="none" stroke="#191970" d="M1183.495,-520.9863C1185.7262,-508.5286 1188.5208,-492.9258 1190.2839,-483.0817"/>
+<polygon fill="#191970" stroke="#191970" points="1180.0372,-520.4401 1181.7193,-530.9005 1186.9276,-521.6742 1180.0372,-520.4401"/>
</g>
<!-- Node133 -->
-<g id="node29" class="node">
+<g id="node30" class="node">
<title>Node133</title>
-<g id="a_node29"><a xlink:href="relay_2type_8h.html" target="_top" xlink:title="Relay typed AST nodes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="145,-330 145,-349 281,-349 281,-330 145,-330"/>
-<text text-anchor="middle" x="213" y="-337" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/type.h</text>
+<g id="a_node30"><a xlink:href="relay_2type_8h.html" target="_top" xlink:title="Relay typed AST nodes. ">
+<polygon fill="#ffffff" stroke="#ff0000" points="334,-335.5 334,-354.5 470,-354.5 470,-335.5 334,-335.5"/>
+<text text-anchor="middle" x="402" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/type.h</text>
</a>
</g>
</g>
<!-- Node151->Node133 -->
-<g id="edge34" class="edge">
+<g id="edge36" class="edge">
<title>Node151->Node133</title>
-<path fill="none" stroke="#191970" d="M275.3667,-521.1769C260.9084,-479.0593 226.9942,-380.2656 216.2946,-349.0974"/>
-<polygon fill="#191970" stroke="#191970" points="272.1147,-522.4835 278.672,-530.8054 278.7354,-520.2107 272.1147,-522.4835"/>
+<path fill="none" stroke="#191970" d="M1109.9221,-536.428C987.6466,-527.4578 728.996,-500.1447 526,-422 481.3036,-404.7938 435.0169,-371.0241 413.8502,-354.5263"/>
+<polygon fill="#191970" stroke="#191970" points="1110.0372,-539.945 1120.2612,-537.1664 1110.536,-532.9628 1110.0372,-539.945"/>
</g>
<!-- Node153 -->
<g id="node24" class="node">
<title>Node153</title>
<g id="a_node24"><a xlink:href="tir_2function_8h.html" target="_top" xlink:title="TIR Function. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1316,-330 1316,-349 1456,-349 1456,-330 1316,-330"/>
-<text text-anchor="middle" x="1386" y="-337" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/function.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1494,-335.5 1494,-354.5 1634,-354.5 1634,-335.5 1494,-335.5"/>
+<text text-anchor="middle" x="1564" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/function.h</text>
</a>
</g>
</g>
<!-- Node152->Node153 -->
<g id="edge28" class="edge">
<title>Node152->Node153</title>
-<path fill="none" stroke="#191970" d="M1229.923,-461.9355C1298.561,-447.86 1405.1653,-425.5232 1408,-422 1425.5626,-400.1718 1405.2016,-365.6069 1393.3409,-349.0095"/>
-<polygon fill="#191970" stroke="#191970" points="1229.1831,-458.5143 1220.0884,-463.949 1230.5873,-465.372 1229.1831,-458.5143"/>
+<path fill="none" stroke="#191970" d="M1250.0801,-462.2201C1259.7124,-460.6178 1269.6172,-459.1327 1279,-458 1341.1214,-450.5006 1798.1922,-467.6377 1841,-422 1850.4258,-411.9511 1849.9087,-401.5101 1841,-391 1827.7218,-375.3349 1711.918,-360.3628 1634.3065,-351.958"/>
+<polygon fill="#191970" stroke="#191970" points="1249.4369,-458.7792 1240.1742,-463.9224 1250.6226,-465.6781 1249.4369,-458.7792"/>
</g>
<!-- Node153->Node154 -->
<g id="edge29" class="edge">
<title>Node153->Node154</title>
-<path fill="none" stroke="#191970" d="M1429.3899,-327.1817C1468.7442,-316.0091 1526.8712,-299.5069 1568.9692,-287.5553"/>
-<polygon fill="#191970" stroke="#191970" points="1428.2264,-323.8736 1419.5624,-329.9717 1430.1382,-330.6075 1428.2264,-323.8736"/>
+<path fill="none" stroke="#191970" d="M1536.9473,-330.6574C1518.7607,-321.0154 1494.8334,-308.3298 1476.3768,-298.5446"/>
+<polygon fill="#191970" stroke="#191970" points="1535.4004,-333.7987 1545.875,-335.3906 1538.6794,-327.6142 1535.4004,-333.7987"/>
</g>
<!-- Node91 -->
<g id="node26" class="node">
<title>Node91</title>
<g id="a_node26"><a xlink:href="tir_2analysis_8h.html" target="_top" xlink:title="Analysis utilitie and passes for TIR. ">
-<polygon fill="#ffffff" stroke="#000000" points="1388,-263 1388,-282 1528,-282 1528,-263 1388,-263"/>
-<text text-anchor="middle" x="1458" y="-270" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/analysis.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1580,-274 1580,-293 1720,-293 1720,-274 1580,-274"/>
+<text text-anchor="middle" x="1650" y="-281" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/analysis.h</text>
</a>
</g>
</g>
<!-- Node153->Node91 -->
<g id="edge30" class="edge">
<title>Node153->Node91</title>
-<path fill="none" stroke="#191970" d="M1403.7958,-322.94C1417.607,-310.0879 1436.2697,-292.7212 1447.7033,-282.0817"/>
-<polygon fill="#191970" stroke="#191970" points="1401.2522,-320.5259 1396.3158,-329.9005 1406.0209,-325.6504 1401.2522,-320.5259"/>
-</g>
-<!-- Node107 -->
-<g id="node27" class="node">
-<title>Node107</title>
-<g id="a_node27"><a xlink:href="tir_2transform_8h.html" target="_top" xlink:title="TIR specific transformation passes. ">
-<polygon fill="#ffffff" stroke="#000000" points="1222,-263 1222,-282 1370,-282 1370,-263 1222,-263"/>
-<text text-anchor="middle" x="1296" y="-270" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/transform.h</text>
-</a>
+<path fill="none" stroke="#191970" d="M1585.7188,-329.4685C1601.7509,-318.0037 1623.0604,-302.765 1636.6569,-293.0419"/>
+<polygon fill="#191970" stroke="#191970" points="1583.5357,-326.7268 1577.4375,-335.3906 1587.6076,-332.4207 1583.5357,-326.7268"/>
</g>
-</g>
-<!-- Node153->Node107 -->
+<!-- Node153->Node156 -->
<g id="edge32" class="edge">
-<title>Node153->Node107</title>
-<path fill="none" stroke="#191970" d="M1364.7169,-323.6559C1347.3665,-310.7395 1323.43,-292.9201 1308.8709,-282.0817"/>
-<polygon fill="#191970" stroke="#191970" points="1362.9938,-326.7365 1373.1052,-329.9005 1367.1739,-321.1216 1362.9938,-326.7365"/>
+<title>Node153->Node156</title>
+<path fill="none" stroke="#191970" d="M1609.8286,-332.7989C1646.749,-322.9694 1698.834,-309.1026 1738.364,-298.5784"/>
+<polygon fill="#191970" stroke="#191970" points="1608.6032,-329.5031 1599.8403,-335.4581 1610.4042,-336.2675 1608.6032,-329.5031"/>
</g>
<!-- Node40 -->
<g id="node28" class="node">
<title>Node40</title>
<g id="a_node28"><a xlink:href="greedy_8h.html" target="_top" xlink:title="This header file contains helper methods used in greedy algorithms for planning memory for USMP...">
-<polygon fill="#ffffff" stroke="#000000" points="1716,-257.5 1716,-287.5 1834,-287.5 1834,-257.5 1716,-257.5"/>
-<text text-anchor="start" x="1724" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/usmp</text>
-<text text-anchor="middle" x="1775" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/algo/greedy.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1493,-201.5 1493,-231.5 1611,-231.5 1611,-201.5 1493,-201.5"/>
+<text text-anchor="start" x="1501" y="-219.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/usmp</text>
+<text text-anchor="middle" x="1552" y="-208.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/algo/greedy.h</text>
</a>
</g>
</g>
<!-- Node153->Node40 -->
-<g id="edge33" class="edge">
+<g id="edge35" class="edge">
<title>Node153->Node40</title>
-<path fill="none" stroke="#191970" d="M1443.2599,-328.0959C1450.8965,-326.6659 1458.6317,-325.2635 1466,-324 1572.7415,-305.6958 1600.5724,-308.0487 1707,-288 1709.8558,-287.462 1712.7738,-286.8887 1715.7172,-286.2914"/>
-<polygon fill="#191970" stroke="#191970" points="1442.4891,-324.6796 1433.3168,-329.9824 1443.794,-331.5569 1442.4891,-324.6796"/>
+<path fill="none" stroke="#191970" d="M1562.1577,-325.2718C1559.7528,-299.5195 1555.5949,-254.9952 1553.4196,-231.7016"/>
+<polygon fill="#191970" stroke="#191970" points="1558.6833,-325.7104 1563.098,-335.3416 1565.653,-325.0595 1558.6833,-325.7104"/>
+</g>
+<!-- Node107 -->
+<g id="node29" class="node">
+<title>Node107</title>
+<g id="a_node29"><a xlink:href="tir_2transform_8h.html" target="_top" xlink:title="TIR specific transformation passes. ">
+<polygon fill="#ffffff" stroke="#000000" points="1206,-274 1206,-293 1354,-293 1354,-274 1206,-274"/>
+<text text-anchor="middle" x="1280" y="-281" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/transform.h</text>
+</a>
+</g>
+</g>
+<!-- Node153->Node107 -->
+<g id="edge34" class="edge">
+<title>Node153->Node107</title>
+<path fill="none" stroke="#191970" d="M1509.8795,-333.2802C1455.9877,-321.61 1374.2682,-303.9137 1324.1668,-293.0643"/>
+<polygon fill="#191970" stroke="#191970" points="1509.4224,-336.7623 1519.9366,-335.4581 1510.9039,-329.9209 1509.4224,-336.7623"/>
</g>
<!-- Node91->Node92 -->
<g id="edge31" class="edge">
<title>Node91->Node92</title>
-<path fill="none" stroke="#191970" d="M1436.3963,-256.9519C1386.631,-221.1359 1264.4152,-133.1776 1215.1215,-97.7011"/>
-<polygon fill="#191970" stroke="#191970" points="1434.3842,-259.816 1444.5453,-262.8167 1438.4732,-254.1344 1434.3842,-259.816"/>
+<path fill="none" stroke="#191970" d="M1647.9517,-263.7072C1644.9669,-244.9778 1637.6833,-217.3639 1620,-201 1567.1783,-152.1192 1363.2079,-111.2728 1257.8338,-92.9133"/>
+<polygon fill="#191970" stroke="#191970" points="1644.5052,-264.3427 1649.284,-273.7985 1651.4449,-263.4264 1644.5052,-264.3427"/>
+</g>
+<!-- Node156->Node40 -->
+<g id="edge33" class="edge">
+<title>Node156->Node40</title>
+<path fill="none" stroke="#191970" d="M1730.3846,-265.6842C1691.8402,-255.0568 1643.4316,-241.7095 1606.7718,-231.6017"/>
+<polygon fill="#191970" stroke="#191970" points="1729.8954,-269.1799 1740.4661,-268.4639 1731.7561,-262.4317 1729.8954,-269.1799"/>
</g>
<!-- Node54 -->
-<g id="node31" class="node">
+<g id="node32" class="node">
<title>Node54</title>
-<g id="a_node31"><a xlink:href="codegen_8h.html" target="_top" xlink:title="Translates IRModule to runtime::Module. ">
-<polygon fill="#ffffff" stroke="#000000" points="1762.5,-391.5 1762.5,-421.5 1869.5,-421.5 1869.5,-391.5 1762.5,-391.5"/>
-<text text-anchor="start" x="1770.5" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
-<text text-anchor="middle" x="1816" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/codegen.h</text>
+<g id="a_node32"><a xlink:href="codegen_8h.html" target="_top" xlink:title="Translates IRModule to runtime::Module. ">
+<polygon fill="#ffffff" stroke="#000000" points="1724.5,-391.5 1724.5,-421.5 1831.5,-421.5 1831.5,-391.5 1724.5,-391.5"/>
+<text text-anchor="start" x="1732.5" y="-409.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
+<text text-anchor="middle" x="1778" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/codegen.h</text>
</a>
</g>
</g>
<!-- Node37->Node54 -->
-<g id="edge37" class="edge">
+<g id="edge39" class="edge">
<title>Node37->Node54</title>
-<path fill="none" stroke="#191970" d="M1413.2141,-522.5716C1505.62,-495.9428 1677.4844,-446.4163 1763.7541,-421.5558"/>
-<polygon fill="#191970" stroke="#191970" points="1412.0773,-519.2567 1403.4375,-525.389 1414.0157,-525.983 1412.0773,-519.2567"/>
+<path fill="none" stroke="#191970" d="M1408.0405,-522.3284C1493.2769,-495.6423 1650.7472,-446.3408 1729.9113,-421.5558"/>
+<polygon fill="#191970" stroke="#191970" points="1406.7625,-519.061 1398.2651,-525.389 1408.854,-525.7412 1406.7625,-519.061"/>
</g>
<!-- Node37->Node72 -->
-<g id="edge38" class="edge">
-<title>Node37->Node72</title>
-<path fill="none" stroke="#191970" d="M1351,-515.0249C1351,-506.128 1351,-496.4287 1351,-488.6432"/>
-<polygon fill="#191970" stroke="#191970" points="1347.5001,-515.2966 1351,-525.2967 1354.5001,-515.2967 1347.5001,-515.2966"/>
-</g>
-<!-- Node171->Node147 -->
<g id="edge40" class="edge">
-<title>Node171->Node147</title>
-<path fill="none" stroke="#191970" d="M988.3777,-460.5206C962.4193,-451.0482 927.4208,-437.2414 898,-422 875.4506,-410.3183 871.5289,-404.4702 850,-391 826.1578,-376.0825 798.2173,-359.2828 781.0632,-349.0456"/>
-<polygon fill="#191970" stroke="#191970" points="987.259,-463.8377 997.8534,-463.9328 989.6306,-457.2517 987.259,-463.8377"/>
+<title>Node37->Node72</title>
+<path fill="none" stroke="#191970" d="M1346.6173,-515.3179C1345.4115,-506.3414 1344.0915,-496.5143 1343.0342,-488.6432"/>
+<polygon fill="#191970" stroke="#191970" points="1343.1575,-515.8517 1347.9578,-525.2967 1350.0952,-514.9197 1343.1575,-515.8517"/>
</g>
-<!-- Node171->Node116 -->
-<g id="edge49" class="edge">
-<title>Node171->Node116</title>
-<path fill="none" stroke="#191970" d="M996.2042,-459.3556C977.9539,-450.0055 954.6089,-436.7757 936,-422 889.4514,-385.0397 888.7919,-364.1884 845,-324 827.867,-308.2768 806.369,-292.0904 792.6548,-282.1617"/>
-<polygon fill="#191970" stroke="#191970" points="994.7231,-462.5282 1005.2315,-463.8791 997.8591,-456.27 994.7231,-462.5282"/>
+<!-- Node172->Node147 -->
+<g id="edge42" class="edge">
+<title>Node172->Node147</title>
+<path fill="none" stroke="#191970" d="M985.3788,-458.1865C971.2193,-448.0676 952.3728,-434.4273 936,-422 904.8298,-398.3412 868.9528,-369.404 850.8159,-354.6515"/>
+<polygon fill="#191970" stroke="#191970" points="983.3498,-461.0384 993.5254,-463.9897 987.4112,-455.337 983.3498,-461.0384"/>
</g>
-<!-- Node171->Node117 -->
-<g id="edge48" class="edge">
-<title>Node171->Node117</title>
-<path fill="none" stroke="#191970" d="M955.1179,-471.4172C859.0945,-467.3489 694.023,-455.5598 645,-422 588.861,-383.5688 587.982,-353.8799 571,-288 567.5609,-274.6583 566.095,-269.8751 571,-257 576.6367,-242.2045 589.4593,-229.1418 599.2752,-220.7534"/>
-<polygon fill="#191970" stroke="#191970" points="955.1966,-474.9232 965.3304,-471.8319 955.4807,-467.929 955.1966,-474.9232"/>
+<!-- Node172->Node116 -->
+<g id="edge51" class="edge">
+<title>Node172->Node116</title>
+<path fill="none" stroke="#191970" d="M995.1765,-455.1082C988.8752,-445.2912 980.9999,-432.9979 974,-422 949.4374,-383.4083 951.4825,-367.2107 919,-335 900.6968,-316.85 874.8679,-302.0974 856.9027,-293.1063"/>
+<polygon fill="#191970" stroke="#191970" points="992.3065,-457.1163 1000.6557,-463.6386 998.1961,-453.3332 992.3065,-457.1163"/>
</g>
-<!-- Node171->Node118 -->
+<!-- Node172->Node117 -->
<g id="edge50" class="edge">
-<title>Node171->Node118</title>
-<path fill="none" stroke="#191970" d="M1039.7349,-456.0681C1074.1012,-411.4812 1159.4901,-294.1883 1137,-257 1102.6862,-200.2607 1027.4438,-172.0042 978.287,-159.085"/>
-<polygon fill="#191970" stroke="#191970" points="1036.9652,-453.9283 1033.589,-463.9708 1042.4909,-458.2256 1036.9652,-453.9283"/>
-</g>
-<!-- Node171->Node92 -->
-<g id="edge76" class="edge">
-<title>Node171->Node92</title>
-<path fill="none" stroke="#191970" d="M1048.4178,-458.0048C1061.1975,-448.5125 1077.0047,-435.6018 1089,-422 1112.2457,-395.641 1116.2905,-386.9185 1131,-355 1173.3735,-263.053 1188.5146,-140.1205 1192.6745,-97.6666"/>
-<polygon fill="#191970" stroke="#191970" points="1046.1872,-455.2986 1040.138,-463.9968 1050.2912,-460.9694 1046.1872,-455.2986"/>
+<title>Node172->Node117</title>
+<path fill="none" stroke="#191970" d="M1001.0354,-453.8475C985.3713,-402.2362 943.4814,-264.2143 931.9654,-226.2706"/>
+<polygon fill="#191970" stroke="#191970" points="997.7841,-455.1866 1004.0376,-463.7391 1004.4824,-453.1536 997.7841,-455.1866"/>
</g>
-<!-- Node171->Node103 -->
-<g id="edge78" class="edge">
-<title>Node171->Node103</title>
-<path fill="none" stroke="#191970" d="M1073.8528,-461.8611C1081.2347,-460.3685 1088.7968,-459.0113 1096,-458 1355.7193,-421.538 1430.3539,-484.7611 1685,-422 1717.6427,-413.9548 1722.9796,-403.9616 1754,-391 1794.8956,-373.9121 1817.9782,-388.499 1847,-355 1871.6376,-326.5616 1866,-310.1265 1866,-272.5 1866,-272.5 1866,-272.5 1866,-149.5 1866,-91.4874 1691.004,-47.4178 1593.594,-27.3806"/>
-<polygon fill="#191970" stroke="#191970" points="1073.0057,-458.4628 1063.9489,-463.9605 1074.4573,-465.3106 1073.0057,-458.4628"/>
-</g>
-<!-- Node171->Node98 -->
-<g id="edge80" class="edge">
-<title>Node171->Node98</title>
-<path fill="none" stroke="#191970" d="M955.3479,-470.23C845.5715,-464.3562 639.6293,-449.9399 572,-422 567.6417,-420.1994 306,-215.7156 306,-211 306,-211 306,-211 306,-149.5 306,-86.8967 757.2335,-37.1993 922.9208,-21.0612"/>
-<polygon fill="#191970" stroke="#191970" points="955.1708,-473.7254 965.3408,-470.7553 955.5383,-466.735 955.1708,-473.7254"/>
+<!-- Node172->Node118 -->
+<g id="edge52" class="edge">
+<title>Node172->Node118</title>
+<path fill="none" stroke="#191970" d="M1011.7386,-453.8521C1017.3661,-428.7584 1026,-383.9173 1026,-345 1026,-345 1026,-345 1026,-283.5 1026,-245.8735 1031.8433,-229.2589 1007,-201 987.5444,-178.8695 957.6859,-166.233 930.6224,-159.025"/>
+<polygon fill="#191970" stroke="#191970" points="1008.2581,-453.3675 1009.4064,-463.8999 1015.0768,-454.9503 1008.2581,-453.3675"/>
</g>
-<!-- Node171->Node121 -->
+<!-- Node172->Node92 -->
<g id="edge77" class="edge">
-<title>Node171->Node121</title>
-<path fill="none" stroke="#191970" d="M955.0091,-472.1262C850.6861,-468.8974 662.2431,-458.0958 606,-422 573.1705,-400.9307 514,-311.5089 514,-272.5 514,-272.5 514,-272.5 514,-211 514,-156.8859 674.6625,-114.6944 767.2316,-94.876"/>
-<polygon fill="#191970" stroke="#191970" points="955.0444,-475.6286 965.143,-472.4242 955.2502,-468.6317 955.0444,-475.6286"/>
+<title>Node172->Node92</title>
+<path fill="none" stroke="#191970" d="M1054.7355,-460.6183C1076.1112,-452.6546 1100.0406,-440.352 1116,-422 1139.5223,-394.9512 1140,-380.846 1140,-345 1140,-345 1140,-345 1140,-216.5 1140,-170.6002 1167.5941,-121.8486 1183.3313,-97.7772"/>
+<polygon fill="#191970" stroke="#191970" points="1053.415,-457.3722 1045.1318,-463.9781 1055.7266,-463.9795 1053.415,-457.3722"/>
</g>
-<!-- Node171->Node128 -->
-<g id="edge51" class="edge">
-<title>Node171->Node128</title>
-<path fill="none" stroke="#191970" d="M1009.9208,-456.5786C988.7613,-434.0409 950.7687,-392.6315 921,-355 887.9517,-313.2227 885.3587,-298.5298 852,-257 841.4353,-243.8475 828.1501,-229.6431 819.3832,-220.5528"/>
-<polygon fill="#191970" stroke="#191970" points="1007.5005,-459.1135 1016.9084,-463.9862 1012.5925,-454.3102 1007.5005,-459.1135"/>
+<!-- Node172->Node103 -->
+<g id="edge79" class="edge">
+<title>Node172->Node103</title>
+<path fill="none" stroke="#191970" d="M1077.9717,-462.5489C1090.2976,-460.8661 1103.0062,-459.2693 1115,-458 1348.0233,-433.3386 1416.3973,-488.7957 1641,-422 1667.8436,-414.0169 1844.1927,-321.4014 1861,-299 1883.5812,-268.9028 1880,-254.1265 1880,-216.5 1880,-216.5 1880,-216.5 1880,-149.5 1880,-90.5279 1461.334,-39.6005 1298.6103,-22.0337"/>
+<polygon fill="#191970" stroke="#191970" points="1077.1525,-459.1292 1067.732,-463.9774 1078.1197,-466.0621 1077.1525,-459.1292"/>
</g>
-<!-- Node171->Node148 -->
-<g id="edge41" class="edge">
-<title>Node171->Node148</title>
-<path fill="none" stroke="#191970" d="M971.5579,-461.8834C915.2724,-449.8734 827.7202,-431.1919 770.1176,-418.9009"/>
-<polygon fill="#191970" stroke="#191970" points="970.8347,-465.3078 981.3449,-463.9717 972.2955,-458.4619 970.8347,-465.3078"/>
+<!-- Node172->Node98 -->
+<g id="edge81" class="edge">
+<title>Node172->Node98</title>
+<path fill="none" stroke="#191970" d="M1029.5588,-457.594C1057.654,-435.7269 1102,-393.5255 1102,-345 1102,-345 1102,-345 1102,-149.5 1102,-101.0863 1077.2252,-46.9695 1066.1927,-25.2707"/>
+<polygon fill="#191970" stroke="#191970" points="1027.3144,-454.9027 1021.4285,-463.7121 1031.5234,-460.496 1027.3144,-454.9027"/>
</g>
-<!-- Node171->Node115 -->
-<g id="edge42" class="edge">
-<title>Node171->Node115</title>
-<path fill="none" stroke="#191970" d="M978.4505,-461.3449C942.7322,-451.7919 892.8102,-437.5382 850,-422 817.6337,-410.2525 810.8636,-404.0496 779,-391 737.6006,-374.045 718.9456,-384.139 685,-355 663.6115,-336.64 649.8261,-305.9978 643.0298,-287.6735"/>
-<polygon fill="#191970" stroke="#191970" points="977.5674,-464.7317 988.1305,-463.9119 979.3617,-457.9655 977.5674,-464.7317"/>
+<!-- Node172->Node121 -->
+<g id="edge78" class="edge">
+<title>Node172->Node121</title>
+<path fill="none" stroke="#191970" d="M1020.5997,-455.7453C1037.4601,-431.9147 1064,-387.6321 1064,-345 1064,-345 1064,-345 1064,-216.5 1064,-151.1661 988.8663,-115.367 931.6356,-97.5752"/>
+<polygon fill="#191970" stroke="#191970" points="1017.7645,-453.693 1014.6873,-463.8311 1023.4151,-457.8248 1017.7645,-453.693"/>
+</g>
+<!-- Node172->Node128 -->
+<g id="edge53" class="edge">
+<title>Node172->Node128</title>
+<path fill="none" stroke="#191970" d="M936.2426,-467.8349C871.6164,-461.1163 782.6444,-447.465 759,-422 706.359,-365.3056 723.0943,-258.8932 729.8051,-226.2622"/>
+<polygon fill="#191970" stroke="#191970" points="936.0922,-471.3373 946.3919,-468.8544 936.7919,-464.3723 936.0922,-471.3373"/>
</g>
-<!-- Node171->Node149 -->
+<!-- Node172->Node148 -->
<g id="edge43" class="edge">
-<title>Node171->Node149</title>
-<path fill="none" stroke="#191970" d="M955.3813,-467.6334C859.9835,-459.3534 684.7462,-442.8907 536,-422 523.8639,-420.2955 510.8378,-418.167 498.6267,-416.0384"/>
-<polygon fill="#191970" stroke="#191970" points="955.2222,-471.1326 965.4862,-468.5057 955.8243,-464.1585 955.2222,-471.1326"/>
+<title>Node172->Node148</title>
+<path fill="none" stroke="#191970" d="M971.3816,-460.3153C941.2411,-449.1583 898.0513,-433.1709 866.5812,-421.5218"/>
+<polygon fill="#191970" stroke="#191970" points="970.474,-463.7114 981.0671,-463.9005 972.904,-457.1467 970.474,-463.7114"/>
</g>
-<!-- Node171->Node153 -->
-<g id="edge60" class="edge">
-<title>Node171->Node153</title>
-<path fill="none" stroke="#191970" d="M1076.4246,-461.7844C1082.9926,-460.4203 1089.6447,-459.1189 1096,-458 1156.4819,-447.3515 1325.2726,-464.0822 1370,-422 1390.2956,-402.9047 1389.2128,-366.3791 1387.3546,-349.0619"/>
-<polygon fill="#191970" stroke="#191970" points="1075.2757,-458.4508 1066.2274,-463.9624 1076.7379,-465.2964 1075.2757,-458.4508"/>
+<!-- Node172->Node115 -->
+<g id="edge44" class="edge">
+<title>Node172->Node115</title>
+<path fill="none" stroke="#191970" d="M936.1607,-470.2809C862.6096,-465.3454 744.9612,-452.9468 648,-422 567.5202,-396.3135 554.758,-372.3833 479,-335 453.4745,-322.4042 424.3071,-308.6568 402.5714,-298.5362"/>
+<polygon fill="#191970" stroke="#191970" points="936.0357,-473.78 946.2397,-470.929 936.4849,-466.7945 936.0357,-473.78"/>
</g>
-<!-- Node171->Node91 -->
-<g id="edge52" class="edge">
-<title>Node171->Node91</title>
-<path fill="none" stroke="#191970" d="M1076.9427,-461.7865C1083.3451,-460.4422 1089.8154,-459.1452 1096,-458 1149.2446,-448.1407 1301.6517,-463.0207 1337,-422 1345.994,-411.5628 1341.6903,-403.9549 1337,-391 1329.9098,-371.4167 1314.0902,-374.5833 1307,-355 1302.3097,-342.0451 1298.9828,-335.205 1307,-324 1319.4909,-306.5425 1372.2116,-291.3454 1411.8672,-282.0713"/>
-<polygon fill="#191970" stroke="#191970" points="1076.0314,-458.4023 1066.9885,-463.9228 1077.5004,-465.2464 1076.0314,-458.4023"/>
+<!-- Node172->Node149 -->
+<g id="edge45" class="edge">
+<title>Node172->Node149</title>
+<path fill="none" stroke="#191970" d="M936.1215,-466.0215C814.8468,-453.2256 568.7333,-427.2577 445.6112,-414.2668"/>
+<polygon fill="#191970" stroke="#191970" points="936.0891,-469.5374 946.4012,-467.1061 936.8237,-462.576 936.0891,-469.5374"/>
</g>
-<!-- Node171->Node107 -->
+<!-- Node172->Node153 -->
+<g id="edge61" class="edge">
+<title>Node172->Node153</title>
+<path fill="none" stroke="#191970" d="M1077.9933,-462.7442C1090.3167,-461.0394 1103.0189,-459.3847 1115,-458 1205.8569,-447.4994 1449.2446,-473.2479 1525,-422 1548.5822,-406.0468 1558.6424,-371.3593 1562.2623,-354.5339"/>
+<polygon fill="#191970" stroke="#191970" points="1077.1703,-459.3254 1067.7548,-464.1832 1078.1446,-466.2573 1077.1703,-459.3254"/>
+</g>
+<!-- Node172->Node91 -->
+<g id="edge54" class="edge">
+<title>Node172->Node91</title>
+<path fill="none" stroke="#191970" d="M1077.9873,-462.6915C1090.3114,-460.9926 1103.0154,-459.3535 1115,-458 1308.2089,-436.179 1368.7484,-489.7508 1551,-422 1598.4127,-404.3747 1617.8896,-398.91 1643,-355 1654.2688,-335.2945 1653.0528,-307.4694 1651.4082,-293.0295"/>
+<polygon fill="#191970" stroke="#191970" points="1077.1653,-459.2724 1067.7484,-464.1276 1078.1377,-466.2046 1077.1653,-459.2724"/>
+</g>
+<!-- Node172->Node156 -->
<g id="edge75" class="edge">
-<title>Node171->Node107</title>
-<path fill="none" stroke="#191970" d="M1070.3692,-460.8773C1096.1308,-452.3111 1128.4384,-439.3687 1154,-422 1215.2895,-380.3548 1270.3054,-308.311 1289.1207,-282.2575"/>
-<polygon fill="#191970" stroke="#191970" points="1069.2889,-457.5481 1060.8441,-463.9462 1071.4356,-464.2108 1069.2889,-457.5481"/>
+<title>Node172->Node156</title>
+<path fill="none" stroke="#191970" d="M1077.9786,-462.6136C1090.3037,-460.9235 1103.0103,-459.3075 1115,-458 1221.5561,-446.3796 1494.997,-457.8835 1596,-422 1673.3442,-394.5218 1748.3222,-328.5451 1779.6625,-298.6632"/>
+<polygon fill="#191970" stroke="#191970" points="1077.1582,-459.1942 1067.7393,-464.0455 1078.1277,-466.1267 1077.1582,-459.1942"/>
</g>
-<!-- Node171->Node133 -->
-<g id="edge46" class="edge">
-<title>Node171->Node133</title>
-<path fill="none" stroke="#191970" d="M955.049,-471.7032C807.8558,-467.3683 473.9382,-454.2137 365,-422 309.4298,-405.5676 251.2561,-367.0213 225.9892,-349.0444"/>
-<polygon fill="#191970" stroke="#191970" points="955.2215,-475.2096 965.3186,-472.0002 955.424,-468.2125 955.2215,-475.2096"/>
+<!-- Node172->Node107 -->
+<g id="edge76" class="edge">
+<title>Node172->Node107</title>
+<path fill="none" stroke="#191970" d="M1061.112,-461.9325C1068.1279,-460.5457 1075.2266,-459.1986 1082,-458 1189.9415,-438.8993 1233.2957,-480.4988 1326,-422 1362.8158,-398.7683 1392.9284,-372.6066 1371,-335 1358.7469,-313.9863 1334.3884,-300.861 1313.9546,-293.1234"/>
+<polygon fill="#191970" stroke="#191970" points="1060.0525,-458.5754 1050.9397,-463.9798 1061.4337,-465.4378 1060.0525,-458.5754"/>
</g>
-<!-- Node171->Node54 -->
-<g id="edge47" class="edge">
-<title>Node171->Node54</title>
-<path fill="none" stroke="#191970" d="M1073.587,-461.8758C1081.0521,-460.3675 1088.7091,-459.0017 1096,-458 1383.5182,-418.4977 1460.6056,-462.3933 1748,-422 1752.6763,-421.3427 1757.506,-420.526 1762.3338,-419.6137"/>
-<polygon fill="#191970" stroke="#191970" points="1072.6341,-458.5 1063.5787,-464 1074.0875,-465.3474 1072.6341,-458.5"/>
+<!-- Node172->Node133 -->
+<g id="edge48" class="edge">
+<title>Node172->Node133</title>
+<path fill="none" stroke="#191970" d="M935.9399,-468.9556C830.7061,-461.5659 638.8749,-445.381 574,-422 550.1086,-413.3895 548.7067,-402.3729 526,-391 495.6153,-375.7815 458.9186,-362.7849 433.3636,-354.5276"/>
+<polygon fill="#191970" stroke="#191970" points="935.9384,-472.4638 946.1567,-469.6645 936.423,-465.4806 935.9384,-472.4638"/>
</g>
-<!-- Node171->Node172 -->
-<g id="edge44" class="edge">
-<title>Node171->Node172</title>
-<path fill="none" stroke="#191970" d="M954.7895,-469.4625C818.8551,-461.5945 513.0242,-443.1183 256,-422 226.4232,-419.5698 193.7938,-416.4633 165.8306,-413.6704"/>
-<polygon fill="#191970" stroke="#191970" points="954.8787,-472.9734 965.0639,-470.0557 955.2823,-465.9851 954.8787,-472.9734"/>
+<!-- Node172->Node54 -->
+<g id="edge49" class="edge">
+<title>Node172->Node54</title>
+<path fill="none" stroke="#191970" d="M1077.1358,-462.5812C1089.7205,-460.8655 1102.7326,-459.2493 1115,-458 1378.5649,-431.1592 1447.7487,-459.5646 1710,-422 1714.6746,-421.3304 1719.503,-420.5051 1724.33,-419.5871"/>
+<polygon fill="#191970" stroke="#191970" points="1076.5775,-459.1251 1067.1571,-463.9734 1077.5449,-466.058 1076.5775,-459.1251"/>
+</g>
+<!-- Node172->Node173 -->
+<g id="edge46" class="edge">
+<title>Node172->Node173</title>
+<path fill="none" stroke="#191970" d="M936.2404,-469.8651C807.3218,-462.9222 525.6385,-446.2736 289,-422 272.7295,-420.331 255.2214,-418.179 238.863,-416.0128"/>
+<polygon fill="#191970" stroke="#191970" points="936.2013,-473.3679 946.3742,-470.4078 936.5757,-466.378 936.2013,-473.3679"/>
</g>
<!-- Node110 -->
-<g id="node35" class="node">
+<g id="node36" class="node">
<title>Node110</title>
-<g id="a_node35"><a xlink:href="builtin_8h.html" target="_top" xlink:title="TIR builtin intrinsics. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1498.5,-397 1498.5,-416 1629.5,-416 1629.5,-397 1498.5,-397"/>
-<text text-anchor="middle" x="1564" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/builtin.h</text>
+<g id="a_node36"><a xlink:href="builtin_8h.html" target="_top" xlink:title="TIR builtin intrinsics. ">
+<polygon fill="#ffffff" stroke="#ff0000" points="1168.5,-397 1168.5,-416 1299.5,-416 1299.5,-397 1168.5,-397"/>
+<text text-anchor="middle" x="1234" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/builtin.h</text>
</a>
</g>
</g>
-<!-- Node171->Node110 -->
-<g id="edge53" class="edge">
-<title>Node171->Node110</title>
-<path fill="none" stroke="#191970" d="M1075.1509,-461.8035C1082.1238,-460.3819 1089.2248,-459.0601 1096,-458 1267.1031,-431.2268 1312.5567,-446.5009 1484,-422 1495.344,-420.3788 1507.5183,-418.2318 1518.862,-416.054"/>
-<polygon fill="#191970" stroke="#191970" points="1074.1359,-458.4408 1065.0793,-463.9389 1075.5879,-465.2886 1074.1359,-458.4408"/>
+<!-- Node172->Node110 -->
+<g id="edge55" class="edge">
+<title>Node172->Node110</title>
+<path fill="none" stroke="#191970" d="M1049.3648,-460.9958C1093.4525,-447.9832 1161.5307,-427.8896 1201.5368,-416.0817"/>
+<polygon fill="#191970" stroke="#191970" points="1048.1237,-457.7128 1039.5236,-463.9005 1050.1053,-464.4265 1048.1237,-457.7128"/>
</g>
<!-- Node94 -->
-<g id="node36" class="node">
+<g id="node37" class="node">
<title>Node94</title>
-<g id="a_node36"><a xlink:href="elemwise_8h.html" target="_top" xlink:title="Elementwise op constructions. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1474.5,-330 1474.5,-349 1629.5,-349 1629.5,-330 1474.5,-330"/>
-<text text-anchor="middle" x="1552" y="-337" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/elemwise.h</text>
+<g id="a_node37"><a xlink:href="elemwise_8h.html" target="_top" xlink:title="Elementwise op constructions. ">
+<polygon fill="#ffffff" stroke="#ff0000" points="1206.5,-335.5 1206.5,-354.5 1361.5,-354.5 1361.5,-335.5 1206.5,-335.5"/>
+<text text-anchor="middle" x="1284" y="-342.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/topi/elemwise.h</text>
</a>
</g>
</g>
-<!-- Node171->Node94 -->
-<g id="edge79" class="edge">
-<title>Node171->Node94</title>
-<path fill="none" stroke="#191970" d="M1075.6906,-461.8784C1082.4947,-460.4684 1089.4057,-459.1297 1096,-458 1226.1871,-435.6963 1264.7178,-460.7159 1391,-422 1448.6536,-404.3244 1510.3893,-366.756 1537.6538,-349.0765"/>
-<polygon fill="#191970" stroke="#191970" points="1074.8968,-458.4689 1065.8477,-463.9792 1076.358,-465.3147 1074.8968,-458.4689"/>
+<!-- Node172->Node94 -->
+<g id="edge80" class="edge">
+<title>Node172->Node94</title>
+<path fill="none" stroke="#191970" d="M1061.3979,-461.9478C1068.3226,-460.5685 1075.3211,-459.2191 1082,-458 1132.2447,-448.829 1275.8334,-460.841 1309,-422 1325.9761,-402.1195 1305.0865,-370.4801 1292.3932,-354.6637"/>
+<polygon fill="#191970" stroke="#191970" points="1060.4593,-458.5666 1051.3508,-463.9783 1061.846,-465.4279 1060.4593,-458.5666"/>
</g>
<!-- Node131 -->
-<g id="node37" class="node">
+<g id="node38" class="node">
<title>Node131</title>
-<g id="a_node37"><a xlink:href="data__layout_8h.html" target="_top" xlink:title="Layout expression to describe the data organization of a tensor. And BijectiveLayout to mapping two d...">
-<polygon fill="#ffffff" stroke="#ff0000" points="1014.5,-257.5 1014.5,-287.5 1127.5,-287.5 1127.5,-257.5 1014.5,-257.5"/>
-<text text-anchor="start" x="1022.5" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/data</text>
-<text text-anchor="middle" x="1071" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_layout.h</text>
+<g id="a_node38"><a xlink:href="data__layout_8h.html" target="_top" xlink:title="Layout expression to describe the data organization of a tensor. And BijectiveLayout to mapping two d...">
+<polygon fill="#ffffff" stroke="#ff0000" points="598.5,-268.5 598.5,-298.5 711.5,-298.5 711.5,-268.5 598.5,-268.5"/>
+<text text-anchor="start" x="606.5" y="-286.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/data</text>
+<text text-anchor="middle" x="655" y="-275.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_layout.h</text>
</a>
</g>
</g>
-<!-- Node171->Node131 -->
-<g id="edge56" class="edge">
-<title>Node171->Node131</title>
-<path fill="none" stroke="#191970" d="M1030.3785,-453.9427C1039.3387,-413.9205 1059.5974,-323.4316 1067.6286,-287.5588"/>
-<polygon fill="#191970" stroke="#191970" points="1026.9398,-453.2822 1028.1704,-463.8054 1033.7707,-454.8116 1026.9398,-453.2822"/>
-</g>
-<!-- Node171->Node173 -->
-<g id="edge57" class="edge">
-<title>Node171->Node173</title>
-<path fill="none" stroke="#191970" d="M1073.579,-461.8161C1081.0451,-460.3158 1088.7046,-458.9681 1096,-458 1441.3403,-412.1736 1534.0416,-470.6182 1879,-422 1881.757,-421.6114 1884.5684,-421.1595 1887.3997,-420.6595"/>
-<polygon fill="#191970" stroke="#191970" points="1072.6281,-458.4398 1063.57,-463.9353 1074.0781,-465.288 1072.6281,-458.4398"/>
+<!-- Node172->Node131 -->
+<g id="edge58" class="edge">
+<title>Node172->Node131</title>
+<path fill="none" stroke="#191970" d="M936.0757,-467.7493C863.9574,-460.651 758.3556,-446.4191 726,-422 691.5726,-396.0173 702.8604,-373.2872 683,-335 676.6269,-322.7139 669.1229,-308.9471 663.4882,-298.7403"/>
+<polygon fill="#191970" stroke="#191970" points="936.172,-471.2745 946.4604,-468.7452 936.8403,-464.3064 936.172,-471.2745"/>
</g>
-<!-- Node171->Node174 -->
-<g id="edge74" class="edge">
-<title>Node171->Node174</title>
-<path fill="none" stroke="#191970" d="M1073.8575,-461.8943C1081.2387,-460.3972 1088.7995,-459.0298 1096,-458 1335.4268,-423.7571 1407.6737,-492.6111 1639,-422 1686.7214,-407.4333 1735.5554,-374.2297 1761.6806,-354.6764"/>
-<polygon fill="#191970" stroke="#191970" points="1073.0092,-458.4963 1063.954,-463.9967 1074.4629,-465.3437 1073.0092,-458.4963"/>
+<!-- Node172->Node174 -->
+<g id="edge59" class="edge">
+<title>Node172->Node174</title>
+<path fill="none" stroke="#191970" d="M1076.4387,-462.5629C1089.2371,-460.8236 1102.5028,-459.204 1115,-458 1436.5742,-427.0189 1519.5107,-453.85 1841,-422 1856.4411,-420.4702 1873.0939,-418.2475 1888.4513,-415.9607"/>
+<polygon fill="#191970" stroke="#191970" points="1075.7188,-459.1293 1066.2984,-463.9777 1076.6861,-466.0622 1075.7188,-459.1293"/>
</g>
-<!-- Node171->Node114 -->
-<g id="edge61" class="edge">
-<title>Node171->Node114</title>
-<path fill="none" stroke="#191970" d="M1020.4677,-454.1136C1015.6658,-437.3764 1008.4848,-412.5608 1002,-391 997.6315,-376.4754 992.3846,-359.7103 989.1205,-349.3566"/>
-<polygon fill="#191970" stroke="#191970" points="1017.1321,-455.1794 1023.2485,-463.8304 1023.862,-453.2533 1017.1321,-455.1794"/>
+<!-- Node172->Node114 -->
+<g id="edge62" class="edge">
+<title>Node172->Node114</title>
+<path fill="none" stroke="#191970" d="M936.3278,-467.5588C855.7392,-459.8394 729.7438,-444.6165 688,-422 657.9777,-405.7341 634.1311,-371.5163 623.7053,-354.7195"/>
+<polygon fill="#191970" stroke="#191970" points="936.1246,-471.0551 946.4087,-468.5081 936.7809,-464.0859 936.1246,-471.0551"/>
</g>
<!-- Node175 -->
<g id="node41" class="node">
<title>Node175</title>
<g id="a_node41"><a xlink:href="stmt_8h.html" target="_top" xlink:title="TIR statements. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1206,-397 1206,-416 1328,-416 1328,-397 1206,-397"/>
-<text text-anchor="middle" x="1267" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/stmt.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1394,-397 1394,-416 1516,-416 1516,-397 1394,-397"/>
+<text text-anchor="middle" x="1455" y="-404" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/stmt.h</text>
</a>
</g>
</g>
-<!-- Node171->Node175 -->
-<g id="edge68" class="edge">
-<title>Node171->Node175</title>
-<path fill="none" stroke="#191970" d="M1070.1203,-461.2342C1116.8684,-448.2378 1189.7101,-427.9872 1232.4546,-416.1039"/>
-<polygon fill="#191970" stroke="#191970" points="1068.9706,-457.921 1060.2735,-463.9717 1070.8456,-464.6652 1068.9706,-457.921"/>
+<!-- Node172->Node175 -->
+<g id="edge69" class="edge">
+<title>Node172->Node175</title>
+<path fill="none" stroke="#191970" d="M1060.0342,-461.966C1067.3946,-460.5331 1074.8732,-459.1648 1082,-458 1213.6604,-436.4811 1248.3396,-443.5189 1380,-422 1390.3663,-420.3057 1401.4769,-418.181 1411.8852,-416.053"/>
+<polygon fill="#191970" stroke="#191970" points="1059.2356,-458.5563 1050.1148,-463.947 1060.6066,-465.4207 1059.2356,-458.5563"/>
</g>
-<!-- Node172->Node133 -->
-<g id="edge45" class="edge">
-<title>Node172->Node133</title>
-<path fill="none" stroke="#191970" d="M122.3043,-391.8846C144.8006,-378.891 177.0432,-360.2681 196.4109,-349.0817"/>
-<polygon fill="#191970" stroke="#191970" points="120.5288,-388.8682 113.62,-396.9005 124.0299,-394.9297 120.5288,-388.8682"/>
+<!-- Node173->Node133 -->
+<g id="edge47" class="edge">
+<title>Node173->Node133</title>
+<path fill="none" stroke="#191970" d="M218.5366,-394.2707C262.0233,-382.592 326.6618,-365.2328 366.47,-354.5419"/>
+<polygon fill="#191970" stroke="#191970" points="217.5312,-390.9166 208.7813,-396.8906 219.3469,-397.6771 217.5312,-390.9166"/>
</g>
<!-- Node110->Node40 -->
-<g id="edge54" class="edge">
+<g id="edge56" class="edge">
<title>Node110->Node40</title>
-<path fill="none" stroke="#191970" d="M1587.6188,-391.5004C1628.16,-365.7538 1710.5443,-313.434 1751.2058,-287.611"/>
-<polygon fill="#191970" stroke="#191970" points="1585.7398,-388.5475 1579.1746,-396.8631 1589.4925,-394.4566 1585.7398,-388.5475"/>
+<path fill="none" stroke="#191970" d="M1217.92,-389.1064C1210.3377,-379.7401 1202.0754,-367.5512 1198,-355 1186.0577,-318.2211 1171.6955,-297.2403 1197,-268 1216.052,-245.9846 1396.6928,-228.6527 1492.6853,-220.9091"/>
+<polygon fill="#191970" stroke="#191970" points="1215.387,-391.5294 1224.5525,-396.8441 1220.7018,-386.9738 1215.387,-391.5294"/>
</g>
<!-- Node110->Node94 -->
-<g id="edge55" class="edge">
+<g id="edge57" class="edge">
<title>Node110->Node94</title>
-<path fill="none" stroke="#191970" d="M1560.505,-386.9863C1558.2738,-374.5286 1555.4792,-358.9258 1553.7161,-349.0817"/>
-<polygon fill="#191970" stroke="#191970" points="1557.0724,-387.6742 1562.2807,-396.9005 1563.9628,-386.4401 1557.0724,-387.6742"/>
-</g>
-<!-- Node173->Node174 -->
-<g id="edge58" class="edge">
-<title>Node173->Node174</title>
-<path fill="none" stroke="#191970" d="M1897.9942,-387.5897C1872.6006,-377.1518 1841.4935,-364.3654 1817.7399,-354.6017"/>
-<polygon fill="#191970" stroke="#191970" points="1896.8398,-390.8992 1907.4196,-391.4639 1899.5011,-384.4248 1896.8398,-390.8992"/>
+<path fill="none" stroke="#191970" d="M1248.1671,-389.0744C1257.21,-377.9517 1268.5972,-363.9454 1276.0774,-354.7449"/>
+<polygon fill="#191970" stroke="#191970" points="1245.4051,-386.9235 1241.8125,-396.8906 1250.8366,-391.3393 1245.4051,-386.9235"/>
</g>
-<!-- Node174->Node40 -->
-<g id="edge59" class="edge">
-<title>Node174->Node40</title>
-<path fill="none" stroke="#191970" d="M1778.7449,-314.3179C1777.941,-305.3414 1777.061,-295.5143 1776.3561,-287.6432"/>
-<polygon fill="#191970" stroke="#191970" points="1775.2604,-314.6487 1779.6385,-324.2967 1782.2325,-314.0243 1775.2604,-314.6487"/>
+<!-- Node174->Node156 -->
+<g id="edge60" class="edge">
+<title>Node174->Node156</title>
+<path fill="none" stroke="#191970" d="M1918.6292,-384.8759C1888.4931,-360.1643 1839.9428,-320.3531 1813.5434,-298.7056"/>
+<polygon fill="#191970" stroke="#191970" points="1916.6551,-387.7834 1926.6071,-391.4178 1921.0936,-382.3705 1916.6551,-387.7834"/>
</g>
<!-- Node114->Node116 -->
-<g id="edge63" class="edge">
+<g id="edge64" class="edge">
<title>Node114->Node116</title>
-<path fill="none" stroke="#191970" d="M946.6034,-326.7484C906.3814,-313.7297 844.8539,-293.815 808.603,-282.0817"/>
-<polygon fill="#191970" stroke="#191970" points="945.7501,-330.151 956.3419,-329.9005 947.9057,-323.4911 945.7501,-330.151"/>
+<path fill="none" stroke="#191970" d="M661.8733,-332.6229C703.2648,-320.9459 764.4402,-303.6877 802.1767,-293.0419"/>
+<polygon fill="#191970" stroke="#191970" points="660.7365,-329.3069 652.0625,-335.3906 662.6372,-336.044 660.7365,-329.3069"/>
</g>
<!-- Node114->Node118 -->
-<g id="edge64" class="edge">
+<g id="edge65" class="edge">
<title>Node114->Node118</title>
-<path fill="none" stroke="#191970" d="M980.5074,-320.1741C969.1345,-280.1584 943.3571,-189.46 934.8113,-159.3917"/>
-<polygon fill="#191970" stroke="#191970" points="977.1473,-321.1545 983.2479,-329.8167 983.8807,-319.2407 977.1473,-321.1545"/>
+<path fill="none" stroke="#191970" d="M603.3413,-327.0113C597.6878,-318.9515 591.8938,-309.0164 589,-299 585.1759,-285.7636 583.975,-280.8287 589,-268 603.797,-230.2234 615.9657,-221.4603 651,-201 677.9158,-185.281 756.8051,-168.9641 811.6491,-159.0186"/>
+<polygon fill="#191970" stroke="#191970" points="600.6997,-329.3224 609.5023,-335.2186 606.2979,-325.12 600.6997,-329.3224"/>
</g>
<!-- Node114->Node98 -->
-<g id="edge67" class="edge">
+<g id="edge68" class="edge">
<title>Node114->Node98</title>
-<path fill="none" stroke="#191970" d="M990.2076,-319.8227C996.8525,-287.6353 1009.4851,-221.6348 1014,-165 1015.0949,-151.2658 1015.7579,-147.6652 1014,-134 1008.6822,-92.6608 993.3357,-45.1597 986.4341,-25.1644"/>
-<polygon fill="#191970" stroke="#191970" points="986.7576,-319.2212 988.13,-329.7268 993.6085,-320.6584 986.7576,-319.2212"/>
+<path fill="none" stroke="#191970" d="M575.6645,-331.984C527.6653,-314.2848 456,-277.6244 456,-216.5 456,-216.5 456,-216.5 456,-149.5 456,-94.0962 847.7508,-41.0899 1000.7784,-22.4979"/>
+<polygon fill="#191970" stroke="#191970" points="574.6221,-335.3278 585.2169,-335.3738 576.9633,-328.7309 574.6221,-335.3278"/>
</g>
<!-- Node114->Node121 -->
-<g id="edge66" class="edge">
+<g id="edge67" class="edge">
<title>Node114->Node121</title>
-<path fill="none" stroke="#191970" d="M975.0646,-320.9427C958.4216,-293.4019 924.9593,-240.6953 890,-201 874.1924,-183.0509 862.1358,-185.61 850,-165 837.471,-143.7224 833.2037,-114.8418 831.7504,-97.5007"/>
-<polygon fill="#191970" stroke="#191970" points="972.2718,-323.0923 980.4069,-329.88 978.2803,-319.5008 972.2718,-323.0923"/>
+<path fill="none" stroke="#191970" d="M600.2519,-328.0096C593.1213,-319.9751 585.7187,-309.7856 582,-299 568.1147,-258.728 524.4143,-203.5917 583,-134 610.7673,-101.0163 729.2374,-89.1126 804.4209,-84.848"/>
+<polygon fill="#191970" stroke="#191970" points="597.7209,-330.4271 607.1452,-335.2679 602.7967,-325.6066 597.7209,-330.4271"/>
</g>
<!-- Node114->Node115 -->
-<g id="edge62" class="edge">
+<g id="edge63" class="edge">
<title>Node114->Node115</title>
-<path fill="none" stroke="#191970" d="M919.9014,-328.2718C862.8513,-318.3693 778.304,-303.1881 705,-288 702.0773,-287.3944 699.0865,-286.7593 696.0687,-286.1062"/>
-<polygon fill="#191970" stroke="#191970" points="919.4213,-331.7407 929.8717,-329.9973 920.6151,-324.8432 919.4213,-331.7407"/>
+<path fill="none" stroke="#191970" d="M569.6386,-333.0072C529.216,-322.983 471.4394,-308.6553 428.4334,-297.9905"/>
+<polygon fill="#191970" stroke="#191970" points="568.9736,-336.4482 579.5221,-335.4581 570.6585,-329.654 568.9736,-336.4482"/>
</g>
<!-- Node114->Node131 -->
-<g id="edge65" class="edge">
+<g id="edge66" class="edge">
<title>Node114->Node131</title>
-<path fill="none" stroke="#191970" d="M1006.1571,-323.6115C1020.0222,-312.6825 1038.3204,-298.2592 1051.9425,-287.5218"/>
-<polygon fill="#191970" stroke="#191970" points="1003.8653,-320.9613 998.1784,-329.9005 1008.1987,-326.4588 1003.8653,-320.9613"/>
+<path fill="none" stroke="#191970" d="M628.9471,-326.8041C634.3615,-317.8045 640.8179,-307.073 645.9488,-298.5446"/>
+<polygon fill="#191970" stroke="#191970" points="625.9374,-325.0175 623.7813,-335.3906 631.9356,-328.6262 625.9374,-325.0175"/>
</g>
<!-- Node175->Node147 -->
-<g id="edge69" class="edge">
+<g id="edge70" class="edge">
<title>Node175->Node147</title>
-<path fill="none" stroke="#191970" d="M1195.8029,-396.9976C1100.2753,-384.2479 931.6084,-361.7366 836.1201,-348.9921"/>
-<polygon fill="#191970" stroke="#191970" points="1195.5717,-400.4977 1205.9468,-398.3515 1196.4978,-393.5592 1195.5717,-400.4977"/>
+<path fill="none" stroke="#191970" d="M1383.9087,-398.8344C1360.0996,-396.2986 1333.4071,-393.4899 1309,-391 1167.6908,-376.5841 1002.3853,-360.6123 910.2826,-351.793"/>
+<polygon fill="#191970" stroke="#191970" points="1383.6061,-402.3219 1393.921,-399.9027 1384.3488,-395.3614 1383.6061,-402.3219"/>
</g>
<!-- Node175->Node153 -->
-<g id="edge71" class="edge">
+<g id="edge72" class="edge">
<title>Node175->Node153</title>
-<path fill="none" stroke="#191970" d="M1292.9587,-391.8846C1316.0368,-378.891 1349.1133,-360.2681 1368.9818,-349.0817"/>
-<polygon fill="#191970" stroke="#191970" points="1291.0465,-388.9446 1284.0498,-396.9005 1294.4808,-395.0442 1291.0465,-388.9446"/>
+<path fill="none" stroke="#191970" d="M1480.8498,-391.915C1501.3391,-380.3546 1529.3769,-364.535 1547.0883,-354.5419"/>
+<polygon fill="#191970" stroke="#191970" points="1479.0207,-388.9283 1472.0313,-396.8906 1482.4605,-395.0249 1479.0207,-388.9283"/>
</g>
<!-- Node175->Node91 -->
-<g id="edge70" class="edge">
+<g id="edge71" class="edge">
<title>Node175->Node91</title>
-<path fill="none" stroke="#191970" d="M1262.8888,-386.9185C1260.2416,-368.5649 1259.5895,-341.4152 1274,-324 1283.0671,-313.0424 1362.564,-293.6533 1414.0566,-282.0473"/>
-<polygon fill="#191970" stroke="#191970" points="1259.4485,-387.5639 1264.6284,-396.8061 1266.3426,-386.3509 1259.4485,-387.5639"/>
+<path fill="none" stroke="#191970" d="M1458.5261,-386.7031C1462.3813,-370.6506 1470.1116,-348.5408 1485,-335 1502.7047,-318.8978 1562.3986,-302.8688 1604.8865,-293.0724"/>
+<polygon fill="#191970" stroke="#191970" points="1455.0632,-386.1634 1456.4359,-396.6689 1461.9141,-387.6003 1455.0632,-386.1634"/>
</g>
-<!-- Node175->Node174 -->
-<g id="edge73" class="edge">
-<title>Node175->Node174</title>
-<path fill="none" stroke="#191970" d="M1338.249,-397.2127C1441.4096,-383.7657 1630.299,-359.1439 1724.4854,-346.8667"/>
-<polygon fill="#191970" stroke="#191970" points="1337.6989,-393.7547 1328.2352,-398.518 1338.6038,-400.696 1337.6989,-393.7547"/>
+<!-- Node175->Node156 -->
+<g id="edge74" class="edge">
+<title>Node175->Node156</title>
+<path fill="none" stroke="#191970" d="M1507.7028,-394.6425C1545.8178,-385.447 1598.2638,-371.5092 1643,-355 1687.4586,-338.5932 1736.7206,-314.2061 1766.7743,-298.5806"/>
+<polygon fill="#191970" stroke="#191970" points="1506.7597,-391.2691 1497.8438,-396.9925 1508.3828,-398.0784 1506.7597,-391.2691"/>
</g>
<!-- Node175->Node114 -->
-<g id="edge72" class="edge">
+<g id="edge73" class="edge">
<title>Node175->Node114</title>
-<path fill="none" stroke="#191970" d="M1217.2608,-394.6405C1162.834,-381.6633 1076.6406,-361.1118 1026.2791,-349.1039"/>
-<polygon fill="#191970" stroke="#191970" points="1216.4989,-398.0568 1227.038,-396.9717 1218.1224,-391.2477 1216.4989,-398.0568"/>
+<path fill="none" stroke="#191970" d="M1383.9518,-398.375C1360.1456,-395.8078 1333.444,-393.0967 1309,-391 1064.9287,-370.065 1003.4,-371.6651 759,-355 731.0885,-353.0968 700.0749,-350.9056 674.257,-349.0603"/>
+<polygon fill="#191970" stroke="#191970" points="1383.6423,-401.8618 1393.9622,-399.4641 1384.3995,-394.9029 1383.6423,-401.8618"/>
</g>
-<!-- Node179->Node171 -->
-<g id="edge83" class="edge">
-<title>Node179->Node171</title>
-<path fill="none" stroke="#191970" d="M994.5649,-522.4803C1002.7037,-509.7989 1013.2775,-493.3235 1019.8506,-483.0817"/>
-<polygon fill="#191970" stroke="#191970" points="991.6166,-520.5942 989.1609,-530.9005 997.5077,-524.3751 991.6166,-520.5942"/>
+<!-- Node179->Node172 -->
+<g id="edge84" class="edge">
+<title>Node179->Node172</title>
+<path fill="none" stroke="#191970" d="M936.2696,-525.0102C954.0423,-512.067 978.8239,-494.0195 993.8431,-483.0817"/>
+<polygon fill="#191970" stroke="#191970" points="934.2045,-522.1843 928.1814,-530.9005 938.3254,-527.8428 934.2045,-522.1843"/>
</g>
<!-- Node181->Node145 -->
-<g id="edge85" class="edge">
+<g id="edge86" class="edge">
<title>Node181->Node145</title>
-<path fill="none" stroke="#191970" d="M729.9942,-710.6733C727.1666,-682.7236 722.1163,-632.8037 719.9885,-611.7705"/>
-<polygon fill="#191970" stroke="#191970" points="726.534,-711.2447 731.0229,-720.8416 733.4985,-710.5401 726.534,-711.2447"/>
+<path fill="none" stroke="#191970" d="M753,-710.6733C753,-682.7236 753,-632.8037 753,-611.7705"/>
+<polygon fill="#191970" stroke="#191970" points="749.5001,-710.8416 753,-720.8416 756.5001,-710.8416 749.5001,-710.8416"/>
</g>
-<!-- Node181->Node172 -->
-<g id="edge91" class="edge">
-<title>Node181->Node172</title>
-<path fill="none" stroke="#191970" d="M681.7849,-718.8001C675.474,-717.4535 669.0961,-716.1523 663,-715 482.1939,-680.8252 420.2742,-705.8512 262,-612 228.3238,-592.0312 219.9034,-585.3573 194,-556 159.8999,-517.3531 160.2667,-501.0974 132,-458 122.3953,-443.356 110.9307,-426.6226 103.8064,-416.3083"/>
-<polygon fill="#191970" stroke="#191970" points="681.0809,-722.2287 691.5968,-720.9386 682.5715,-715.3893 681.0809,-722.2287"/>
+<!-- Node181->Node173 -->
+<g id="edge92" class="edge">
+<title>Node181->Node173</title>
+<path fill="none" stroke="#191970" d="M718.3269,-717.0227C650.7876,-690.2006 497.7262,-626.6033 378,-556 299.8374,-509.907 214.7257,-441.1904 184.4606,-416.1063"/>
+<polygon fill="#191970" stroke="#191970" points="717.4495,-720.4393 728.036,-720.8586 720.0217,-713.929 717.4495,-720.4393"/>
</g>
<!-- Node181->Node182 -->
-<g id="edge86" class="edge">
+<g id="edge87" class="edge">
<title>Node181->Node182</title>
-<path fill="none" stroke="#191970" d="M688.4301,-718.0779C642.7976,-705.0678 572.1104,-684.9145 530.6073,-673.0817"/>
-<polygon fill="#191970" stroke="#191970" points="687.7538,-721.5245 698.3302,-720.9005 689.6731,-714.7928 687.7538,-721.5245"/>
+<path fill="none" stroke="#191970" d="M704.5529,-718.8201C697.6793,-717.396 690.6792,-716.0691 684,-715 519.9822,-688.746 476.1692,-704.2901 312,-679 301.4998,-677.3825 290.2365,-675.2108 279.7763,-673.0061"/>
+<polygon fill="#191970" stroke="#191970" points="703.968,-722.2743 714.4807,-720.9575 705.4414,-715.4311 703.968,-722.2743"/>
</g>
<!-- Node182->Node27 -->
-<g id="edge87" class="edge">
+<g id="edge88" class="edge">
<title>Node182->Node27</title>
-<path fill="none" stroke="#191970" d="M485.142,-645.4934C467.7285,-619.0507 435.8445,-570.6342 422.3264,-550.1068"/>
-<polygon fill="#191970" stroke="#191970" points="482.221,-647.4217 490.6441,-653.8484 488.0672,-643.5717 482.221,-647.4217"/>
+<path fill="none" stroke="#191970" d="M227.2115,-645.1458C210.18,-618.6284 179.2969,-570.5445 166.1702,-550.1068"/>
+<polygon fill="#191970" stroke="#191970" points="224.4519,-647.3258 232.801,-653.8484 230.3417,-643.5429 224.4519,-647.3258"/>
</g>
<!-- Node182->Node145 -->
-<g id="edge88" class="edge">
+<g id="edge89" class="edge">
<title>Node182->Node145</title>
-<path fill="none" stroke="#191970" d="M541.4112,-651.1969C583.5658,-639.5189 646.0459,-622.2103 684.5561,-611.5419"/>
-<polygon fill="#191970" stroke="#191970" points="540.3901,-647.8479 531.6875,-653.8906 542.259,-654.5938 540.3901,-647.8479"/>
+<path fill="none" stroke="#191970" d="M308.4077,-655.1954C410.3982,-642.9922 598.9614,-620.4307 694.3978,-609.0117"/>
+<polygon fill="#191970" stroke="#191970" points="307.7265,-651.7518 298.2132,-656.4152 308.5582,-658.7023 307.7265,-651.7518"/>
</g>
<!-- Node182->Node133 -->
-<g id="edge89" class="edge">
+<g id="edge90" class="edge">
<title>Node182->Node133</title>
-<path fill="none" stroke="#191970" d="M445.7082,-651.4772C371.5299,-632.9331 240.1852,-595.3577 213,-556 167.0726,-489.508 198.2889,-381.9962 209.4659,-349.2644"/>
-<polygon fill="#191970" stroke="#191970" points="445.2313,-654.9645 455.7789,-653.9657 446.9106,-648.1689 445.2313,-654.9645"/>
+<path fill="none" stroke="#191970" d="M209.3357,-649.3044C176.3071,-631.9446 123.7937,-599.4263 95,-556 53.6727,-493.6707 12.7911,-447.3149 62,-391 79.5884,-370.8718 239.6701,-356.3547 333.7733,-349.4759"/>
+<polygon fill="#191970" stroke="#191970" points="208.0091,-652.5572 218.5055,-653.9981 211.1986,-646.326 208.0091,-652.5572"/>
</g>
<!-- Node182->Node114 -->
-<g id="edge90" class="edge">
+<g id="edge91" class="edge">
<title>Node182->Node114</title>
-<path fill="none" stroke="#191970" d="M519.2101,-647.9614C579.7852,-605.704 753.219,-485.5295 901,-391 924.5467,-375.9381 952.3305,-359.3157 969.5595,-349.1393"/>
-<polygon fill="#191970" stroke="#191970" points="516.9875,-645.2445 510.7922,-653.8392 520.995,-650.9838 516.9875,-645.2445"/>
+<path fill="none" stroke="#191970" d="M241.6256,-643.4712C244.3473,-615.2003 245.701,-562.8253 225,-525 185.1641,-452.211 38.9559,-452.1898 95,-391 98.9151,-386.7255 427.0899,-360.1793 561.5282,-349.471"/>
+<polygon fill="#191970" stroke="#191970" points="238.1168,-643.3542 240.4769,-653.6828 245.0729,-644.1368 238.1168,-643.3542"/>
</g>
<!-- Node36->Node37 -->
-<g id="edge98" class="edge">
+<g id="edge99" class="edge">
<title>Node36->Node37</title>
-<path fill="none" stroke="#191970" d="M1335.8936,-638.2943C1339.7295,-613.4615 1345.4644,-576.3355 1348.6512,-555.7056"/>
-<polygon fill="#191970" stroke="#191970" points="1332.3975,-638.0007 1334.3298,-648.4178 1339.3154,-639.0693 1332.3975,-638.0007"/>
+<path fill="none" stroke="#191970" d="M1342.8443,-638.2943C1344.6614,-613.4615 1347.3779,-576.3355 1348.8874,-555.7056"/>
+<polygon fill="#191970" stroke="#191970" points="1339.3428,-638.189 1342.1036,-648.4178 1346.3241,-638.6999 1339.3428,-638.189"/>
</g>
<!-- Node184->Node16 -->
-<g id="edge103" class="edge">
+<g id="edge104" class="edge">
<title>Node184->Node16</title>
-<path fill="none" stroke="#191970" d="M642.4682,-896.6785C631.2438,-879.1498 621.3196,-855.4369 634,-838 645.2022,-822.5957 690.3607,-809.7149 726.785,-801.5386"/>
-<polygon fill="#191970" stroke="#191970" points="639.7908,-898.9619 648.3435,-905.2152 645.5572,-894.9933 639.7908,-898.9619"/>
+<path fill="none" stroke="#191970" d="M1045.3497,-901.0364C1026.8456,-892.1361 1005.3603,-880.9225 987,-869 968.72,-857.1297 966.5324,-850.949 949,-838 931.0496,-824.7423 909.5411,-810.6756 895.3637,-801.6397"/>
+<polygon fill="#191970" stroke="#191970" points="1043.9079,-904.2262 1054.4438,-905.3417 1046.9031,-897.8994 1043.9079,-904.2262"/>
</g>
<!-- Node184->Node151 -->
-<g id="edge102" class="edge">
+<g id="edge103" class="edge">
<title>Node184->Node151</title>
-<path fill="none" stroke="#191970" d="M575.4573,-904.8722C397.5344,-872.2233 4,-799.196 4,-792 4,-792 4,-792 4,-730.5 4,-663.0582 14.0707,-635.0319 66,-592 89.5616,-572.4754 166.9631,-557.3618 222.3699,-548.6966"/>
-<polygon fill="#191970" stroke="#191970" points="574.8457,-908.3184 585.3131,-906.68 576.1086,-901.4332 574.8457,-908.3184"/>
+<path fill="none" stroke="#191970" d="M1129.9767,-900.1402C1142.579,-892.0409 1155.385,-881.6099 1164,-869 1183.8257,-839.981 1182,-827.1449 1182,-792 1182,-792 1182,-792 1182,-663.5 1182,-621.2132 1180.8036,-570.7595 1180.2615,-550.0683"/>
+<polygon fill="#191970" stroke="#191970" points="1128.0782,-897.1981 1121.3386,-905.373 1131.7051,-903.1853 1128.0782,-897.1981"/>
</g>
<!-- Node185 -->
<g id="node47" class="node">
<title>Node185</title>
<g id="a_node47"><a xlink:href="reflection_8h.html" target="_top" xlink:title="Reflection and serialization of compiler IR/AST nodes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="643.5,-844 643.5,-863 802.5,-863 802.5,-844 643.5,-844"/>
-<text text-anchor="middle" x="723" y="-851" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/reflection.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="995.5,-844 995.5,-863 1154.5,-863 1154.5,-844 995.5,-844"/>
+<text text-anchor="middle" x="1075" y="-851" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/node/reflection.h</text>
</a>
</g>
</g>
<!-- Node184->Node185 -->
-<g id="edge104" class="edge">
+<g id="edge105" class="edge">
<title>Node184->Node185</title>
-<path fill="none" stroke="#191970" d="M681.946,-897.8648C692.9272,-885.998 705.7959,-872.0916 714.1581,-863.055"/>
-<polygon fill="#191970" stroke="#191970" points="679.2918,-895.5798 675.0688,-905.2967 684.4296,-900.3342 679.2918,-895.5798"/>
+<path fill="none" stroke="#191970" d="M1083.7691,-895.4663C1081.4077,-884.1655 1078.7589,-871.4891 1076.9966,-863.055"/>
+<polygon fill="#191970" stroke="#191970" points="1080.3517,-896.224 1085.8232,-905.2967 1087.2037,-894.7922 1080.3517,-896.224"/>
</g>
<!-- Node185->Node16 -->
-<g id="edge105" class="edge">
+<g id="edge106" class="edge">
<title>Node185->Node16</title>
-<path fill="none" stroke="#191970" d="M738.0172,-836.0744C747.6026,-824.9517 759.6731,-810.9454 767.602,-801.7449"/>
-<polygon fill="#191970" stroke="#191970" points="735.1582,-834.0306 731.2813,-843.8906 740.4608,-838.6003 735.1582,-834.0306"/>
+<path fill="none" stroke="#191970" d="M1034.5755,-840.7507C997.5765,-829.0818 943.6531,-812.0752 910.2548,-801.5419"/>
+<polygon fill="#191970" stroke="#191970" points="1033.9416,-844.2207 1044.5313,-843.8906 1036.0471,-837.5448 1033.9416,-844.2207"/>
</g>
<!-- Node190->Node16 -->
-<g id="edge108" class="edge">
+<g id="edge109" class="edge">
<title>Node190->Node16</title>
-<path fill="none" stroke="#191970" d="M826.9927,-895.1062C824.3292,-878.3074 819.4957,-856.1217 811,-838 804.5253,-824.1892 793.4005,-810.562 785.4301,-801.7691"/>
-<polygon fill="#191970" stroke="#191970" points="823.5658,-895.8669 828.4463,-905.2707 830.4953,-894.8759 823.5658,-895.8669"/>
+<path fill="none" stroke="#191970" d="M1238.9312,-897.7771C1221.4859,-878.6046 1193.9034,-852.0946 1164,-838 1126.8387,-820.4845 1020.17,-806.4854 948.6659,-798.7159"/>
+<polygon fill="#191970" stroke="#191970" points="1236.5123,-900.3247 1245.7644,-905.4871 1241.751,-895.6818 1236.5123,-900.3247"/>
</g>
<!-- Node190->Node151 -->
-<g id="edge107" class="edge">
+<g id="edge108" class="edge">
<title>Node190->Node151</title>
-<path fill="none" stroke="#191970" d="M744.3531,-905.0859C576.3733,-881.489 80,-938.8221 80,-792 80,-792 80,-792 80,-663.5 80,-595.1363 165.4476,-563.6688 225.1942,-550.0221"/>
-<polygon fill="#191970" stroke="#191970" points="744.0422,-908.58 754.4647,-906.6771 745.1304,-901.6651 744.0422,-908.58"/>
+<path fill="none" stroke="#191970" d="M1254.2638,-895.3561C1245.8285,-840.0678 1223.8542,-703.9373 1196,-592 1192.3375,-577.2815 1186.9031,-560.5679 1183.4061,-550.2782"/>
+<polygon fill="#191970" stroke="#191970" points="1250.8123,-895.9404 1255.7681,-905.3047 1257.7336,-894.8938 1250.8123,-895.9404"/>
</g>
<!-- Node190->Node185 -->
-<g id="edge109" class="edge">
+<g id="edge110" class="edge">
<title>Node190->Node185</title>
-<path fill="none" stroke="#191970" d="M797.3322,-900.0445C777.714,-887.7601 753.6667,-872.7025 738.3376,-863.1039"/>
-<polygon fill="#191970" stroke="#191970" points="795.6541,-903.1232 805.9871,-905.4639 799.3691,-897.1903 795.6541,-903.1232"/>
+<path fill="none" stroke="#191970" d="M1207.471,-902.0003C1172.9291,-889.3538 1128.5701,-873.1131 1100.9622,-863.0053"/>
+<polygon fill="#191970" stroke="#191970" points="1206.3375,-905.3125 1216.9312,-905.4639 1208.7441,-898.7392 1206.3375,-905.3125"/>
</g>
</g>
</svg>
diff --git a/docs/reference/api/doxygen/greedy_8h.html b/docs/reference/api/doxygen/greedy_8h.html
index a2159325d..afbfb7b6d 100644
--- a/docs/reference/api/doxygen/greedy_8h.html
+++ b/docs/reference/api/doxygen/greedy_8h.html
@@ -83,7 +83,7 @@ $(function() {
</div><div class="textblock"><div class="dynheader">
Include dependency graph for greedy.h:</div>
<div class="dyncontent">
-<div class="center"><iframe scrolling="no" frameborder="0" src="greedy_8h__incl.svg" width="4512" height="947"><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="greedy_8h__incl.svg" width="4571" height="1022"><p><b>This browser is not able to show SVG: try Firefox, Chrome, Safari, or Opera instead.</b></p></iframe>
</div>
</div>
</div>
diff --git a/docs/reference/api/doxygen/greedy_8h__incl.svg b/docs/reference/api/doxygen/greedy_8h__incl.svg
index 05d2d1b84..9e9e011b3 100644
--- a/docs/reference/api/doxygen/greedy_8h__incl.svg
+++ b/docs/reference/api/doxygen/greedy_8h__incl.svg
@@ -4,1516 +4,1522 @@
<!-- Generated by graphviz version 2.40.1 (20161225.0304)
-->
<!-- Title: include/tvm/tir/usmp/algo/greedy.h Pages: 1 -->
-<svg width="3384pt" height="710pt"
- viewBox="0.00 0.00 3384.00 710.00" 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(4 706)">
+<svg width="3428pt" height="766pt"
+ viewBox="0.00 0.00 3427.91 766.00" 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(4 762)">
<title>include/tvm/tir/usmp/algo/greedy.h</title>
-<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-706 3380,-706 3380,4 -4,4"/>
+<polygon fill="#ffffff" stroke="transparent" points="-4,4 -4,-762 3423.9125,-762 3423.9125,4 -4,4"/>
<!-- Node0 -->
<g id="node1" class="node">
<title>Node0</title>
-<polygon fill="#bfbfbf" stroke="#000000" points="1956,-671.5 1956,-701.5 2074,-701.5 2074,-671.5 1956,-671.5"/>
-<text text-anchor="start" x="1964" y="-689.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/usmp</text>
-<text text-anchor="middle" x="2015" y="-678.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/algo/greedy.h</text>
+<polygon fill="#bfbfbf" stroke="#000000" points="1682,-727.5 1682,-757.5 1800,-757.5 1800,-727.5 1682,-727.5"/>
+<text text-anchor="start" x="1690" y="-745.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/usmp</text>
+<text text-anchor="middle" x="1741" y="-734.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/algo/greedy.h</text>
</g>
<!-- Node1 -->
<g id="node2" class="node">
<title>Node1</title>
<g id="a_node2"><a xlink:href="analyzer_8h.html" target="_top" xlink:title="Algebra expression simplifications. ">
-<polygon fill="#ffffff" stroke="#000000" points="2409.5,-615.5 2409.5,-634.5 2524.5,-634.5 2524.5,-615.5 2409.5,-615.5"/>
-<text text-anchor="middle" x="2467" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/arith/analyzer.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="936.5,-615.5 936.5,-634.5 1051.5,-634.5 1051.5,-615.5 936.5,-615.5"/>
+<text text-anchor="middle" x="994" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/arith/analyzer.h</text>
</a>
</g>
</g>
<!-- Node0->Node1 -->
<g id="edge1" class="edge">
<title>Node0->Node1</title>
-<path fill="none" stroke="#191970" d="M2074.0911,-678.4599C2157.9628,-667.0482 2311.1281,-646.2082 2399.4129,-634.196"/>
-<polygon fill="#191970" stroke="#191970" points="2399.8867,-637.6639 2409.3235,-632.8476 2398.9429,-630.7278 2399.8867,-637.6639"/>
+<path fill="none" stroke="#191970" d="M1681.6641,-737.636C1595.6339,-730.141 1431.4091,-714.1683 1293,-691 1203.2917,-675.9837 1099.9322,-651.5571 1041.2434,-637.0073"/>
+<polygon fill="#191970" stroke="#191970" points="1042.067,-633.6055 1031.5175,-634.585 1040.3752,-640.398 1042.067,-633.6055"/>
</g>
<!-- Node20 -->
<g id="node17" class="node">
<title>Node20</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="1582.5,-123.5 1582.5,-142.5 1629.5,-142.5 1629.5,-123.5 1582.5,-123.5"/>
-<text text-anchor="middle" x="1606" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">vector</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2122.5,-123.5 2122.5,-142.5 2169.5,-142.5 2169.5,-123.5 2122.5,-123.5"/>
+<text text-anchor="middle" x="2146" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">vector</text>
</g>
<!-- Node0->Node20 -->
-<g id="edge183" class="edge">
+<g id="edge184" class="edge">
<title>Node0->Node20</title>
-<path fill="none" stroke="#191970" d="M2074.1713,-680.7991C2170.5404,-671.4759 2367.3815,-652.2728 2534,-635 2765.4466,-611.0067 3198.1486,-740.063 3052,-559 2909.7805,-382.8047 2270.5398,-416.4144 2056,-344 1969.4065,-314.7718 1717.994,-189.4373 1633.8259,-147.0628"/>
-<polygon fill="#191970" stroke="#191970" points="1635.3639,-143.9186 1624.8587,-142.5435 1632.2135,-150.1696 1635.3639,-143.9186"/>
+<path fill="none" stroke="#191970" d="M1800.3506,-741.8657C2016.6357,-738.7805 2750,-721.225 2750,-625 2750,-625 2750,-625 2750,-457 2750,-397.222 2652.077,-305.8205 2355,-179 2296.2518,-153.9207 2222.1454,-141.7798 2179.8081,-136.5103"/>
+<polygon fill="#191970" stroke="#191970" points="2180.101,-133.0206 2169.7576,-135.315 2179.2742,-139.9716 2180.101,-133.0206"/>
</g>
<!-- Node36 -->
<g id="node24" class="node">
<title>Node36</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="3177.5,-123.5 3177.5,-142.5 3270.5,-142.5 3270.5,-123.5 3177.5,-123.5"/>
-<text text-anchor="middle" x="3224" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">unordered_map</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="256.5,-123.5 256.5,-142.5 349.5,-142.5 349.5,-123.5 256.5,-123.5"/>
+<text text-anchor="middle" x="303" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">unordered_map</text>
</g>
<!-- Node0->Node36 -->
-<g id="edge182" class="edge">
+<g id="edge183" class="edge">
<title>Node0->Node36</title>
-<path fill="none" stroke="#191970" d="M2074.0332,-685.9205C2278.4747,-683.5926 2955.4091,-673.1226 3169,-635 3264.0609,-618.0331 3376,-665.5632 3376,-569 3376,-569 3376,-569 3376,-261.5 3376,-202.1038 3307.4743,-164.5379 3262.4443,-146.2644"/>
-<polygon fill="#191970" stroke="#191970" points="3263.5358,-142.9332 3252.9478,-142.5525 3260.9874,-149.4529 3263.5358,-142.9332"/>
+<path fill="none" stroke="#191970" d="M1681.8256,-741.6701C1377.2006,-737.0185 0,-711.2209 0,-625 0,-625 0,-625 0,-457 0,-413.1619 43.6917,-264.8586 149,-179 176.5154,-156.5665 214.7193,-145.0695 246.1911,-139.1794"/>
+<polygon fill="#191970" stroke="#191970" points="246.8931,-142.61 256.1505,-137.4572 245.7004,-135.7123 246.8931,-142.61"/>
</g>
<!-- Node54 -->
<g id="node33" class="node">
<title>Node54</title>
<g id="a_node33"><a xlink:href="device__api_8h.html" target="_top" xlink:title="Abstract device memory management API. ">
-<polygon fill="#ffffff" stroke="#000000" points="2841.5,-380.5 2841.5,-410.5 2954.5,-410.5 2954.5,-380.5 2841.5,-380.5"/>
-<text text-anchor="start" x="2849.5" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/device</text>
-<text text-anchor="middle" x="2898" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_api.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2532.5,-380.5 2532.5,-410.5 2645.5,-410.5 2645.5,-380.5 2532.5,-380.5"/>
+<text text-anchor="start" x="2540.5" y="-398.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/device</text>
+<text text-anchor="middle" x="2589" y="-387.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_api.h</text>
</a>
</g>
</g>
<!-- Node0->Node54 -->
<g id="edge91" class="edge">
<title>Node0->Node54</title>
-<path fill="none" stroke="#191970" d="M2074.3803,-685.6117C2301.2161,-681.9854 3102.5178,-666.9549 3146,-635 3171.1512,-616.5165 3170,-600.2125 3170,-569 3170,-569 3170,-569 3170,-513 3170,-472.7538 3142.7538,-467.2959 3108,-447 3104.7854,-445.1227 3024.8855,-425.7891 2964.5932,-411.3589"/>
-<polygon fill="#191970" stroke="#191970" points="2965.2625,-407.9204 2954.7226,-408.9984 2963.6343,-414.7284 2965.2625,-407.9204"/>
+<path fill="none" stroke="#191970" d="M1800.0711,-741.2994C2005.1636,-736.359 2674,-713.4658 2674,-625 2674,-625 2674,-625 2674,-513 2674,-473.6968 2642.3936,-438.6355 2617.7271,-417.2574"/>
+<polygon fill="#191970" stroke="#191970" points="2619.9289,-414.5361 2610.0112,-410.8092 2615.4401,-419.9074 2619.9289,-414.5361"/>
</g>
<!-- Node55 -->
<g id="node41" class="node">
<title>Node55</title>
<g id="a_node41"><a xlink:href="builtin_8h.html" target="_top" xlink:title="TIR builtin intrinsics. ">
-<polygon fill="#ffffff" stroke="#000000" points="894,-559.5 894,-578.5 986,-578.5 986,-559.5 894,-559.5"/>
-<text text-anchor="middle" x="940" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/builtin.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1291,-559.5 1291,-578.5 1383,-578.5 1383,-559.5 1291,-559.5"/>
+<text text-anchor="middle" x="1337" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/builtin.h</text>
</a>
</g>
</g>
<!-- Node0->Node55 -->
<g id="edge134" class="edge">
<title>Node0->Node55</title>
-<path fill="none" stroke="#191970" d="M1955.7196,-683.8945C1836.7507,-678.2572 1560.5496,-663.1134 1330,-635 1210.5342,-620.4322 1071.8578,-594.8781 996.2642,-580.2077"/>
-<polygon fill="#191970" stroke="#191970" points="996.5097,-576.6897 986.0248,-578.2121 995.1706,-583.5605 996.5097,-576.6897"/>
+<path fill="none" stroke="#191970" d="M1705.8544,-727.4065C1628.1386,-694.0311 1442.7243,-614.4039 1368.5049,-582.53"/>
+<polygon fill="#191970" stroke="#191970" points="1369.7699,-579.2642 1359.2003,-578.534 1367.0076,-585.6961 1369.7699,-579.2642"/>
</g>
<!-- Node72 -->
<g id="node43" class="node">
<title>Node72</title>
<g id="a_node43"><a xlink:href="tir_2function_8h.html" target="_top" xlink:title="TIR Function. ">
-<polygon fill="#ffffff" stroke="#000000" points="1338.5,-615.5 1338.5,-634.5 1439.5,-634.5 1439.5,-615.5 1338.5,-615.5"/>
-<text text-anchor="middle" x="1389" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/function.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1133.5,-615.5 1133.5,-634.5 1234.5,-634.5 1234.5,-615.5 1133.5,-615.5"/>
+<text text-anchor="middle" x="1184" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/function.h</text>
</a>
</g>
</g>
<!-- Node0->Node72 -->
<g id="edge142" class="edge">
<title>Node0->Node72</title>
-<path fill="none" stroke="#191970" d="M1955.8255,-680.6865C1836.9218,-669.0051 1570.1214,-642.7939 1449.8459,-630.9777"/>
-<polygon fill="#191970" stroke="#191970" points="1450.0044,-627.4765 1439.7101,-629.9819 1449.3199,-634.4429 1450.0044,-627.4765"/>
+<path fill="none" stroke="#191970" d="M1681.8065,-738.5071C1606.5734,-732.5125 1473.1233,-718.7909 1362,-691 1308.6622,-677.6607 1249.2552,-653.6812 1214.3368,-638.5924"/>
+<polygon fill="#191970" stroke="#191970" points="1215.728,-635.3808 1205.1629,-634.5886 1212.928,-641.7964 1215.728,-635.3808"/>
</g>
<!-- Node74 -->
<g id="node46" class="node">
<title>Node74</title>
<g id="a_node46"><a xlink:href="stmt__functor_8h.html" target="_top" xlink:title="Functors for tir stmts utility functions to call common functors. ">
-<polygon fill="#ffffff" stroke="#000000" points="1885.5,-615.5 1885.5,-634.5 2008.5,-634.5 2008.5,-615.5 1885.5,-615.5"/>
-<text text-anchor="middle" x="1947" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/stmt_functor.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1718.5,-671.5 1718.5,-690.5 1841.5,-690.5 1841.5,-671.5 1718.5,-671.5"/>
+<text text-anchor="middle" x="1780" y="-678.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/stmt_functor.h</text>
</a>
</g>
</g>
<!-- Node0->Node74 -->
<g id="edge160" class="edge">
<title>Node0->Node74</title>
-<path fill="none" stroke="#191970" d="M1998.191,-671.2977C1988.2603,-662.3163 1975.6743,-650.9334 1965.4673,-641.702"/>
-<polygon fill="#191970" stroke="#191970" points="1967.652,-638.9588 1957.8877,-634.8469 1962.9566,-644.1505 1967.652,-638.9588"/>
+<path fill="none" stroke="#191970" d="M1750.6405,-727.2977C1755.9977,-718.8498 1762.7022,-708.2773 1768.3513,-699.369"/>
+<polygon fill="#191970" stroke="#191970" points="1771.3559,-701.1664 1773.7556,-690.8469 1765.4444,-697.4176 1771.3559,-701.1664"/>
</g>
<!-- Node76 -->
<g id="node48" class="node">
<title>Node76</title>
<g id="a_node48"><a xlink:href="tir_2usmp_2utils_8h.html" target="_top" xlink:title="Utilities for Unified Static Memory Planner. ">
-<polygon fill="#ffffff" stroke="#000000" points="1510.5,-615.5 1510.5,-634.5 1623.5,-634.5 1623.5,-615.5 1510.5,-615.5"/>
-<text text-anchor="middle" x="1567" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/usmp/utils.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2061.5,-615.5 2061.5,-634.5 2174.5,-634.5 2174.5,-615.5 2061.5,-615.5"/>
+<text text-anchor="middle" x="2118" y="-622.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/usmp/utils.h</text>
</a>
</g>
</g>
<!-- Node0->Node76 -->
-<g id="edge170" class="edge">
+<g id="edge171" class="edge">
<title>Node0->Node76</title>
-<path fill="none" stroke="#191970" d="M1955.9711,-678.3967C1872.6643,-666.9606 1721.046,-646.1469 1633.7425,-634.1622"/>
-<polygon fill="#191970" stroke="#191970" points="1634.0663,-630.6739 1623.6832,-632.7813 1633.1142,-637.6089 1634.0663,-630.6739"/>
+<path fill="none" stroke="#191970" d="M1789.1302,-727.4992C1864.0534,-704.1478 2007.2329,-659.5229 2077.3178,-637.6795"/>
+<polygon fill="#191970" stroke="#191970" points="2078.6054,-640.9443 2087.111,-634.6272 2076.5225,-634.2613 2078.6054,-640.9443"/>
</g>
<!-- Node2 -->
<g id="node3" class="node">
<title>Node2</title>
<g id="a_node3"><a xlink:href="int__set_8h.html" target="_top" xlink:title="Integer set. ">
-<polygon fill="#ffffff" stroke="#000000" points="1671.5,-559.5 1671.5,-578.5 1778.5,-578.5 1778.5,-559.5 1671.5,-559.5"/>
-<text text-anchor="middle" x="1725" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/arith/int_set.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1010.5,-559.5 1010.5,-578.5 1117.5,-578.5 1117.5,-559.5 1010.5,-559.5"/>
+<text text-anchor="middle" x="1064" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/arith/int_set.h</text>
</a>
</g>
</g>
<!-- Node1->Node2 -->
<g id="edge2" class="edge">
<title>Node1->Node2</title>
-<path fill="none" stroke="#191970" d="M2409.1637,-622.8346C2293.4355,-618.1209 2024.9765,-605.2757 1788.7311,-578.9801"/>
-<polygon fill="#191970" stroke="#191970" points="1788.8999,-575.4772 1778.5717,-577.8387 1788.1183,-582.4334 1788.8999,-575.4772"/>
+<path fill="none" stroke="#191970" d="M1006.1931,-615.2455C1016.6321,-606.8943 1031.8655,-594.7076 1044.074,-584.9408"/>
+<polygon fill="#191970" stroke="#191970" points="1046.3244,-587.6227 1051.9467,-578.6427 1041.9515,-582.1566 1046.3244,-587.6227"/>
</g>
<!-- Node3 -->
<g id="node4" class="node">
<title>Node3</title>
<g id="a_node4"><a xlink:href="ir_2expr_8h.html" target="_top" xlink:title="Base expr nodes in TVM. ">
-<polygon fill="#ffffff" stroke="#000000" points="959.5,-386 959.5,-405 1038.5,-405 1038.5,-386 959.5,-386"/>
-<text text-anchor="middle" x="999" y="-393" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/expr.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1056.5,-386 1056.5,-405 1135.5,-405 1135.5,-386 1056.5,-386"/>
+<text text-anchor="middle" x="1096" y="-393" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/expr.h</text>
</a>
</g>
</g>
<!-- Node1->Node3 -->
<g id="edge83" class="edge">
<title>Node1->Node3</title>
-<path fill="none" stroke="#191970" d="M2409.3508,-623.3647C2277.1281,-619.2506 1941.945,-606.7537 1663,-579 1618.9995,-574.6222 1312.5006,-530.9315 1269,-523 1156.5555,-502.4979 1100.4299,-546.1796 1018,-467 1004.152,-453.6981 999.9983,-431.5482 998.939,-415.4209"/>
-<polygon fill="#191970" stroke="#191970" points="1002.4365,-415.284 998.6188,-405.4009 995.4401,-415.5076 1002.4365,-415.284"/>
+<path fill="none" stroke="#191970" d="M993.9766,-615.3395C994.2104,-602.2424 995.5867,-578.2753 1002,-559 1021.028,-501.8111 1061.2934,-442.33 1082.4163,-413.4222"/>
+<polygon fill="#191970" stroke="#191970" points="1085.4039,-415.2693 1088.5505,-405.1525 1079.7818,-411.0989 1085.4039,-415.2693"/>
</g>
<!-- Node1->Node20 -->
<g id="edge90" class="edge">
<title>Node1->Node20</title>
-<path fill="none" stroke="#191970" d="M2442.6797,-615.433C2360.7627,-582.774 2088.4866,-470.6705 1882,-344 1778.6136,-280.577 1667.3473,-186.5978 1624.3749,-149.2024"/>
-<polygon fill="#191970" stroke="#191970" points="1626.5456,-146.4511 1616.7136,-142.5036 1621.9379,-151.7208 1626.5456,-146.4511"/>
+<path fill="none" stroke="#191970" d="M1051.8344,-620.258C1144.5803,-612.5162 1332.2791,-596.2829 1491,-579 1648.2628,-561.8759 2053.0324,-538.7842 2194,-467 2227.3836,-450.0002 2241.6532,-445.175 2257,-411 2274.8428,-371.2669 2269.3863,-354.7572 2257,-313 2241.4699,-260.6441 2218.4479,-256.6159 2190,-210 2178.0257,-190.3784 2165.0531,-167.4545 2156.3192,-151.7536"/>
+<polygon fill="#191970" stroke="#191970" points="2159.1499,-149.6403 2151.2449,-142.5861 2153.0255,-153.0303 2159.1499,-149.6403"/>
</g>
<!-- Node23 -->
<g id="node21" class="node">
<title>Node23</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2568,-123.5 2568,-142.5 2626,-142.5 2626,-123.5 2568,-123.5"/>
-<text text-anchor="middle" x="2597" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">memory</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="890,-123.5 890,-142.5 948,-142.5 948,-123.5 890,-123.5"/>
+<text text-anchor="middle" x="919" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">memory</text>
</g>
<!-- Node1->Node23 -->
<g id="edge88" class="edge">
<title>Node1->Node23</title>
-<path fill="none" stroke="#191970" d="M2524.5935,-624.1074C2664.0705,-621.4378 3011.5108,-611.5596 3052,-579 3072.2696,-562.7001 3062.7075,-548.2378 3069,-523 3091.7028,-431.9434 3139.1316,-393.5609 3091,-313 3026.5431,-205.1145 2960.4946,-220.869 2842,-179 2771.8603,-154.2168 2685.4567,-141.9762 2636.2504,-136.6081"/>
-<polygon fill="#191970" stroke="#191970" points="2636.3519,-133.0996 2626.0408,-135.5352 2635.6202,-140.0612 2636.3519,-133.0996"/>
+<path fill="none" stroke="#191970" d="M936.2566,-620.1792C784.4805,-606.927 383.331,-567.9312 345,-523 172.2535,-320.5081 725.1431,-177.0617 879.8959,-141.5542"/>
+<polygon fill="#191970" stroke="#191970" points="880.9423,-144.9057 889.919,-139.2781 879.3922,-138.0795 880.9423,-144.9057"/>
</g>
<!-- Node1->Node36 -->
<g id="edge89" class="edge">
<title>Node1->Node36</title>
-<path fill="none" stroke="#191970" d="M2524.5841,-623.3262C2713.1368,-617.0906 3300,-591.3861 3300,-513 3300,-513 3300,-513 3300,-261.5 3300,-223.3133 3296.44,-211.2557 3276,-179 3268.5403,-167.2281 3257.2113,-156.7061 3247.0299,-148.7064"/>
-<polygon fill="#191970" stroke="#191970" points="3248.9653,-145.7836 3238.8576,-142.6079 3244.7788,-151.3937 3248.9653,-145.7836"/>
+<path fill="none" stroke="#191970" d="M936.2607,-623.7426C740.4844,-618.7236 114,-596.1655 114,-513 114,-513 114,-513 114,-457 114,-330.409 92.3984,-274.0582 176,-179 193.9102,-158.6354 221.5566,-147.2826 246.4874,-140.9555"/>
+<polygon fill="#191970" stroke="#191970" points="247.2927,-144.3617 256.2429,-138.6921 245.7106,-137.5428 247.2927,-144.3617"/>
</g>
<!-- Node44 -->
<g id="node25" class="node">
<title>Node44</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="1634,-252 1634,-271 1678,-271 1678,-252 1634,-252"/>
-<text text-anchor="middle" x="1656" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">limits</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="856,-252 856,-271 900,-271 900,-252 856,-252"/>
+<text text-anchor="middle" x="878" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">limits</text>
</g>
<!-- Node1->Node44 -->
<g id="edge87" class="edge">
<title>Node1->Node44</title>
-<path fill="none" stroke="#191970" d="M2432.5376,-615.4755C2340.3597,-589.3545 2081.5919,-511.7729 1882,-411 1801.5226,-370.3674 1714.8885,-306.7925 1676.4127,-277.3931"/>
-<polygon fill="#191970" stroke="#191970" points="1678.1758,-274.3337 1668.1155,-271.0108 1673.9078,-279.8821 1678.1758,-274.3337"/>
+<path fill="none" stroke="#191970" d="M940.0585,-615.4785C825.1442,-593.9817 562.4636,-537.9821 511,-467 418.7952,-339.825 740.6759,-281.2775 845.848,-265.8196"/>
+<polygon fill="#191970" stroke="#191970" points="846.4902,-269.2634 855.8916,-264.3784 845.4958,-262.3344 846.4902,-269.2634"/>
</g>
<!-- Node52 -->
<g id="node31" class="node">
<title>Node52</title>
<g id="a_node31"><a xlink:href="with_8h.html" target="_top" xlink:title="RAII wrapper function to enter and exit a context object similar to python's with syntax...">
-<polygon fill="#ffffff" stroke="#000000" points="2972.5,-386 2972.5,-405 3081.5,-405 3081.5,-386 2972.5,-386"/>
-<text text-anchor="middle" x="3027" y="-393" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/support/with.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2138.5,-386 2138.5,-405 2247.5,-405 2247.5,-386 2138.5,-386"/>
+<text text-anchor="middle" x="2193" y="-393" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/support/with.h</text>
</a>
</g>
</g>
<!-- Node1->Node52 -->
<g id="edge84" class="edge">
<title>Node1->Node52</title>
-<path fill="none" stroke="#191970" d="M2524.5842,-622.3977C2665.0261,-615.7734 3014.9586,-597.4853 3032,-579 3073.8773,-533.5747 3049.4617,-451.7167 3035.1421,-414.6555"/>
-<polygon fill="#191970" stroke="#191970" points="3038.2603,-413.0297 3031.2702,-405.068 3031.7696,-415.651 3038.2603,-413.0297"/>
+<path fill="none" stroke="#191970" d="M1051.7319,-618.8254C1129.2172,-610.4242 1271.1086,-594.6404 1392,-579 1435.525,-573.3689 2141.1984,-496.158 2174,-467 2188.5074,-454.1041 2192.5077,-431.5168 2193.3507,-415.1744"/>
+<polygon fill="#191970" stroke="#191970" points="2196.8521,-415.1113 2193.5201,-405.0541 2189.8531,-414.9941 2196.8521,-415.1113"/>
</g>
<!-- Node2->Node3 -->
<g id="edge3" class="edge">
<title>Node2->Node3</title>
-<path fill="none" stroke="#191970" d="M1671.1986,-560.9396C1666.3997,-560.2656 1661.6136,-559.6096 1657,-559 1439.924,-530.3172 1375.5094,-550.2745 1173,-467 1154.9841,-459.5916 1151.7934,-454.9278 1134,-447 1101.5522,-432.5429 1063.6938,-418.3217 1036.3393,-408.4991"/>
-<polygon fill="#191970" stroke="#191970" points="1037.3608,-405.1475 1026.7661,-405.0869 1035.0106,-411.7412 1037.3608,-405.1475"/>
+<path fill="none" stroke="#191970" d="M1065.766,-559.4248C1070.8244,-531.999 1085.4442,-452.732 1092.3303,-415.3964"/>
+<polygon fill="#191970" stroke="#191970" points="1095.8281,-415.7285 1094.2,-405.2595 1088.9442,-414.4588 1095.8281,-415.7285"/>
</g>
<!-- Node2->Node36 -->
<g id="edge82" class="edge">
<title>Node2->Node36</title>
-<path fill="none" stroke="#191970" d="M1778.5806,-560.2371C1781.7646,-559.7966 1784.9225,-559.38 1788,-559 2248.0897,-502.1841 2367.4067,-527.7007 2827,-467 2844.7841,-464.6512 3131.1361,-422.3835 3145,-411 3225.0338,-345.285 3227.3674,-204.8362 3225.2663,-152.8698"/>
-<polygon fill="#191970" stroke="#191970" points="3228.7508,-152.4764 3224.7452,-142.668 3221.76,-152.8336 3228.7508,-152.4764"/>
+<path fill="none" stroke="#191970" d="M1010.4019,-567.1367C831.4803,-560.2135 266,-532.3709 266,-457 266,-457 266,-457 266,-261.5 266,-221.2793 282.946,-176.6448 293.9014,-151.9918"/>
+<polygon fill="#191970" stroke="#191970" points="297.1665,-153.2686 298.1629,-142.7207 290.8062,-150.345 297.1665,-153.2686"/>
</g>
<!-- Node49 -->
<g id="node26" class="node">
<title>Node49</title>
<g id="a_node26"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1358.5,-503.5 1358.5,-522.5 1441.5,-522.5 1441.5,-503.5 1358.5,-503.5"/>
-<text text-anchor="middle" x="1400" y="-510.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/expr.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1139.5,-503.5 1139.5,-522.5 1222.5,-522.5 1222.5,-503.5 1139.5,-503.5"/>
+<text text-anchor="middle" x="1181" y="-510.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/expr.h</text>
</a>
</g>
</g>
<!-- Node2->Node49 -->
<g id="edge54" class="edge">
<title>Node2->Node49</title>
-<path fill="none" stroke="#191970" d="M1671.3329,-559.7527C1610.4888,-549.2688 1511.9709,-532.2934 1451.8775,-521.9389"/>
-<polygon fill="#191970" stroke="#191970" points="1452.2978,-518.4598 1441.8486,-520.2108 1451.1091,-525.3581 1452.2978,-518.4598"/>
+<path fill="none" stroke="#191970" d="M1084.113,-559.3733C1102.8103,-550.4241 1130.8744,-536.9917 1151.9934,-526.8835"/>
+<polygon fill="#191970" stroke="#191970" points="1153.6405,-529.9755 1161.1494,-522.5011 1150.6183,-523.6614 1153.6405,-529.9755"/>
</g>
<!-- Node4 -->
<g id="node5" class="node">
<title>Node4</title>
<g id="a_node5"><a xlink:href="ir_2span_8h.html" target="_top" xlink:title="Span information for debugging purposes. ">
-<polygon fill="#ffffff" stroke="#000000" points="850.5,-252 850.5,-271 931.5,-271 931.5,-252 850.5,-252"/>
-<text text-anchor="middle" x="891" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/span.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1726.5,-252 1726.5,-271 1807.5,-271 1807.5,-252 1726.5,-252"/>
+<text text-anchor="middle" x="1767" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/span.h</text>
</a>
</g>
</g>
<!-- Node3->Node4 -->
<g id="edge4" class="edge">
<title>Node3->Node4</title>
-<path fill="none" stroke="#191970" d="M991.2329,-385.8631C973.1742,-363.4569 928.1499,-307.5934 905.0929,-278.9857"/>
-<polygon fill="#191970" stroke="#191970" points="907.6971,-276.6393 898.6967,-271.0496 902.2469,-281.032 907.6971,-276.6393"/>
+<path fill="none" stroke="#191970" d="M1135.9083,-392.973C1275.6775,-384.0433 1735.2987,-353.9755 1747,-344 1764.9069,-328.7341 1768.2569,-300.4191 1768.2424,-281.2704"/>
+<polygon fill="#191970" stroke="#191970" points="1771.7392,-281.1093 1767.9297,-271.2231 1764.7425,-281.3272 1771.7392,-281.1093"/>
</g>
<!-- Node5 -->
<g id="node6" class="node">
<title>Node5</title>
<g id="a_node6"><a xlink:href="node_8h.html" target="_top" xlink:title="Definitions and helper macros for IR/AST nodes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="978.5,-185 978.5,-204 1077.5,-204 1077.5,-185 978.5,-185"/>
-<text text-anchor="middle" x="1028" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/node/node.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1890.5,-185 1890.5,-204 1989.5,-204 1989.5,-185 1890.5,-185"/>
+<text text-anchor="middle" x="1940" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/node/node.h</text>
</a>
</g>
</g>
<!-- Node3->Node5 -->
<g id="edge38" class="edge">
<title>Node3->Node5</title>
-<path fill="none" stroke="#191970" d="M1000.3987,-385.8054C1004.8967,-354.6298 1019.0504,-256.5298 1025.1533,-214.2302"/>
-<polygon fill="#191970" stroke="#191970" points="1028.6514,-214.4947 1026.6153,-204.0974 1021.7231,-213.4951 1028.6514,-214.4947"/>
+<path fill="none" stroke="#191970" d="M1135.6143,-393.0374C1270.4979,-384.574 1706.9714,-356.5133 1769,-344 1815.153,-334.6894 1942.4702,-315.1956 1970,-277 1984.3207,-257.1311 1968.8399,-229.697 1955.3299,-212.0255"/>
+<polygon fill="#191970" stroke="#191970" points="1958.0111,-209.775 1948.9952,-204.2105 1952.5732,-214.1829 1958.0111,-209.775"/>
</g>
<!-- Node10 -->
<g id="node11" class="node">
<title>Node10</title>
<g id="a_node11"><a xlink:href="object_8h.html" target="_top" xlink:title="A managed object in the TVM runtime. ">
-<polygon fill="#ffffff" stroke="#000000" points="1213.5,-123.5 1213.5,-142.5 1332.5,-142.5 1332.5,-123.5 1213.5,-123.5"/>
-<text text-anchor="middle" x="1273" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/object.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1764.5,-123.5 1764.5,-142.5 1883.5,-142.5 1883.5,-123.5 1764.5,-123.5"/>
+<text text-anchor="middle" x="1824" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/object.h</text>
</a>
</g>
</g>
<!-- Node3->Node10 -->
<g id="edge49" class="edge">
<title>Node3->Node10</title>
-<path fill="none" stroke="#191970" d="M993.1171,-385.8462C973.6557,-352.5485 915.5543,-241.1412 969,-179 984.0929,-161.4515 1118.2866,-146.5659 1203.2624,-138.7906"/>
-<polygon fill="#191970" stroke="#191970" points="1203.7253,-142.2631 1213.37,-137.8779 1203.0958,-135.2915 1203.7253,-142.2631"/>
+<path fill="none" stroke="#191970" d="M1135.5954,-392.7511C1243.0639,-385.0777 1534.8029,-362.7588 1574,-344 1667.5479,-299.2302 1655.8601,-240.9945 1739,-179 1755.7543,-166.5069 1776.3267,-155.2814 1793.0358,-147.0474"/>
+<polygon fill="#191970" stroke="#191970" points="1794.8747,-150.0468 1802.3693,-142.558 1791.8405,-143.7385 1794.8747,-150.0468"/>
</g>
<!-- Node16 -->
<g id="node13" class="node">
<title>Node16</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="358,-62 358,-81 402,-81 402,-62 358,-62"/>
-<text text-anchor="middle" x="380" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">string</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="830,-62 830,-81 874,-81 874,-62 830,-62"/>
+<text text-anchor="middle" x="852" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">string</text>
</g>
<!-- Node3->Node16 -->
<g id="edge52" class="edge">
<title>Node3->Node16</title>
-<path fill="none" stroke="#191970" d="M959.4167,-392.1046C874.0698,-384.4827 675.2597,-365.0261 611,-344 552.0212,-324.7018 397.6467,-263.3797 366,-210 343.7055,-172.395 359.9078,-118.7821 371.3343,-90.5596"/>
-<polygon fill="#191970" stroke="#191970" points="374.6774,-91.6434 375.3896,-81.0726 368.2407,-88.892 374.6774,-91.6434"/>
+<path fill="none" stroke="#191970" d="M1056.4587,-394.145C989.5523,-390.9013 850.4205,-380.0286 739,-344 677.9036,-324.2441 645.4696,-331.1748 611,-277 574.2299,-219.2097 561.7906,-176.1479 605,-123 631.7194,-90.1349 759.0244,-77.5347 819.6974,-73.3372"/>
+<polygon fill="#191970" stroke="#191970" points="820.1201,-76.8171 829.8693,-72.6694 819.6615,-69.8321 820.1201,-76.8171"/>
</g>
<!-- Node17 -->
<g id="node14" class="node">
<title>Node17</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="842.5,-62 842.5,-81 911.5,-81 911.5,-62 842.5,-62"/>
-<text text-anchor="middle" x="877" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">type_traits</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="1359.5,-62 1359.5,-81 1428.5,-81 1428.5,-62 1359.5,-62"/>
+<text text-anchor="middle" x="1394" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">type_traits</text>
</g>
<!-- Node3->Node17 -->
<g id="edge53" class="edge">
<title>Node3->Node17</title>
-<path fill="none" stroke="#191970" d="M959.4097,-391.6925C853.207,-380.3918 570,-342.649 570,-261.5 570,-261.5 570,-261.5 570,-194.5 570,-143.7608 613.4093,-145.2701 659,-123 715.0827,-95.6047 786.531,-82.3175 832.0342,-76.2164"/>
-<polygon fill="#191970" stroke="#191970" points="832.7012,-79.6597 842.1758,-74.9182 831.8123,-72.7163 832.7012,-79.6597"/>
+<path fill="none" stroke="#191970" d="M1081.7095,-385.8275C1046.0254,-360.9438 952.2776,-290.9541 902,-210 881.0139,-176.2094 854.315,-152.498 881,-123 912.1486,-88.5679 1229.7996,-76.1193 1349.0714,-72.6378"/>
+<polygon fill="#191970" stroke="#191970" points="1349.4186,-76.1295 1359.3152,-72.3467 1349.2197,-69.1323 1349.4186,-76.1295"/>
</g>
<!-- Node48 -->
<g id="node18" class="node">
<title>Node48</title>
<g id="a_node18"><a xlink:href="ir_2type_8h.html" target="_top" xlink:title="IR/AST nodes for the unified type system in TVM. ">
-<polygon fill="#ffffff" stroke="#000000" points="798,-319 798,-338 878,-338 878,-319 798,-319"/>
-<text text-anchor="middle" x="838" y="-326" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/type.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1658,-319 1658,-338 1738,-338 1738,-319 1658,-319"/>
+<text text-anchor="middle" x="1698" y="-326" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/type.h</text>
</a>
</g>
</g>
<!-- Node3->Node48 -->
<g id="edge23" class="edge">
<title>Node3->Node48</title>
-<path fill="none" stroke="#191970" d="M975.9326,-385.9005C948.3154,-374.4077 901.8804,-355.0838 870.6037,-342.068"/>
-<polygon fill="#191970" stroke="#191970" points="871.6018,-338.6924 861.0246,-338.0817 868.9123,-345.1552 871.6018,-338.6924"/>
+<path fill="none" stroke="#191970" d="M1135.8793,-392.72C1244.8905,-384.9664 1546.2276,-362.4603 1644,-344 1648.7801,-343.0975 1653.7473,-341.9583 1658.6473,-340.7089"/>
+<polygon fill="#191970" stroke="#191970" points="1659.7933,-344.0247 1668.5292,-338.0302 1657.9618,-337.2685 1659.7933,-344.0247"/>
</g>
<!-- Node22 -->
<g id="node20" class="node">
<title>Node22</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="668,-123.5 668,-142.5 732,-142.5 732,-123.5 668,-123.5"/>
-<text text-anchor="middle" x="700" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">algorithm</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="614,-123.5 614,-142.5 678,-142.5 678,-123.5 614,-123.5"/>
+<text text-anchor="middle" x="646" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">algorithm</text>
</g>
<!-- Node3->Node22 -->
<g id="edge50" class="edge">
<title>Node3->Node22</title>
-<path fill="none" stroke="#191970" d="M959.4743,-390.1077C906.018,-382.1079 814.4219,-365.6775 789,-344 729.1432,-292.9595 708.6469,-194.7741 702.4244,-152.8086"/>
-<polygon fill="#191970" stroke="#191970" points="705.8737,-152.1981 701.0535,-142.7632 698.938,-153.1447 705.8737,-152.1981"/>
+<path fill="none" stroke="#191970" d="M1056.1407,-386.8217C945.2753,-362.5015 640.4968,-294.3312 627,-277 598.3245,-240.178 621.0424,-181.3329 635.8307,-151.5542"/>
+<polygon fill="#191970" stroke="#191970" points="638.9553,-153.1314 640.4561,-142.6434 632.7425,-149.9064 638.9553,-153.1314"/>
</g>
<!-- Node33 -->
<g id="node23" class="node">
<title>Node33</title>
<g id="a_node23"><a xlink:href="string_8h.html" target="_top" xlink:title="Runtime String container types. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1210,-179.5 1210,-209.5 1336,-209.5 1336,-179.5 1210,-179.5"/>
-<text text-anchor="start" x="1218" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
-<text text-anchor="middle" x="1273" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/string.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="956,-179.5 956,-209.5 1082,-209.5 1082,-179.5 956,-179.5"/>
+<text text-anchor="start" x="964" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
+<text text-anchor="middle" x="1019" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/string.h</text>
</a>
</g>
</g>
<!-- Node3->Node33 -->
<g id="edge39" class="edge">
<title>Node3->Node33</title>
-<path fill="none" stroke="#191970" d="M1008.6093,-385.8859C1034.0249,-360.7828 1105.7535,-292.0244 1174,-246 1191.6451,-234.1004 1212.3761,-222.9303 1230.2956,-214.0548"/>
-<polygon fill="#191970" stroke="#191970" points="1231.921,-217.1563 1239.3788,-209.631 1228.8559,-210.8631 1231.921,-217.1563"/>
+<path fill="none" stroke="#191970" d="M1092.2861,-385.8054C1080.7758,-355.7588 1045.4508,-263.547 1028.4006,-219.0393"/>
+<polygon fill="#191970" stroke="#191970" points="1031.6146,-217.645 1024.7688,-209.5588 1025.0778,-220.1492 1031.6146,-217.645"/>
</g>
<!-- Node3->Node44 -->
<g id="edge51" class="edge">
<title>Node3->Node44</title>
-<path fill="none" stroke="#191970" d="M1038.6187,-387.4195C1158.2307,-363.0237 1514.1783,-290.4256 1623.6203,-268.1041"/>
-<polygon fill="#191970" stroke="#191970" points="1624.6168,-271.473 1633.7156,-266.0451 1623.2178,-264.6142 1624.6168,-271.473"/>
+<path fill="none" stroke="#191970" d="M1066.411,-385.9316C1040.8997,-377.0127 1003.6083,-362.344 974,-344 943.3825,-325.0307 912.6197,-296.5347 894.4476,-278.4858"/>
+<polygon fill="#191970" stroke="#191970" points="896.5548,-275.6398 887.0299,-271 891.5824,-280.5669 896.5548,-275.6398"/>
</g>
<!-- Node4->Node5 -->
<g id="edge5" class="edge">
<title>Node4->Node5</title>
-<path fill="none" stroke="#191970" d="M910.6288,-251.9005C933.8239,-240.5569 972.6185,-221.5844 999.2131,-208.5783"/>
-<polygon fill="#191970" stroke="#191970" points="1000.962,-211.6191 1008.4077,-204.0817 997.8867,-205.3308 1000.962,-211.6191"/>
+<path fill="none" stroke="#191970" d="M1791.7867,-251.9005C1821.7192,-240.3082 1872.2239,-220.7486 1905.8355,-207.7313"/>
+<polygon fill="#191970" stroke="#191970" points="1907.1982,-210.957 1915.2593,-204.0817 1904.6702,-204.4294 1907.1982,-210.957"/>
</g>
<!-- Node4->Node10 -->
<g id="edge21" class="edge">
<title>Node4->Node10</title>
-<path fill="none" stroke="#191970" d="M893.1955,-251.943C897.7922,-234.2958 910.2095,-196.7792 936,-179 978.1043,-149.9746 1117.5066,-139.1682 1203.269,-135.2142"/>
-<polygon fill="#191970" stroke="#191970" points="1203.6157,-138.7024 1213.4518,-134.7653 1203.3073,-131.7092 1203.6157,-138.7024"/>
+<path fill="none" stroke="#191970" d="M1789.8113,-251.9436C1806.5226,-243.5792 1827.9682,-229.6696 1838,-210 1847.3177,-191.7307 1840.7811,-168.2372 1833.8466,-152.0308"/>
+<polygon fill="#191970" stroke="#191970" points="1836.8168,-150.122 1829.3948,-142.5613 1830.4819,-153.1002 1836.8168,-150.122"/>
</g>
<!-- Node4->Node16 -->
<g id="edge22" class="edge">
<title>Node4->Node16</title>
-<path fill="none" stroke="#191970" d="M861.0449,-251.9472C829.9062,-241.9084 779.8248,-225.4579 737,-210 615.8674,-166.2764 473.6398,-109.4194 411.6028,-84.3412"/>
-<polygon fill="#191970" stroke="#191970" points="412.5877,-80.9641 402.005,-80.4561 409.9611,-87.4526 412.5877,-80.9641"/>
+<path fill="none" stroke="#191970" d="M1743.9185,-251.8891C1702.5226,-234.9788 1612.6304,-199.7363 1534,-179 1288.7874,-114.3328 983.0732,-83.1299 884.3407,-74.2467"/>
+<polygon fill="#191970" stroke="#191970" points="884.3937,-70.7378 874.1237,-73.3412 883.7756,-77.7104 884.3937,-70.7378"/>
</g>
<!-- Node11 -->
<g id="node7" class="node">
<title>Node11</title>
<g id="a_node7"><a xlink:href="c__runtime__api_8h.html" target="_top" xlink:title="tvm/runtime/c_runtime\l_api.h">
-<polygon fill="#ffffff" stroke="#000000" points="2589.5,-56.5 2589.5,-86.5 2718.5,-86.5 2718.5,-56.5 2589.5,-56.5"/>
-<text text-anchor="start" x="2597.5" y="-74.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/c_runtime</text>
-<text text-anchor="middle" x="2654" y="-63.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_api.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="3023.5,-56.5 3023.5,-86.5 3152.5,-86.5 3152.5,-56.5 3023.5,-56.5"/>
+<text text-anchor="start" x="3031.5" y="-74.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/c_runtime</text>
+<text text-anchor="middle" x="3088" y="-63.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_api.h</text>
</a>
</g>
</g>
<!-- Node5->Node11 -->
<g id="edge6" class="edge">
<title>Node5->Node11</title>
-<path fill="none" stroke="#191970" d="M1077.5352,-186.618C1173.2129,-171.7042 1390.2192,-139.3869 1574,-123 1950.7215,-89.4095 2403.4549,-76.7921 2579.1861,-72.9418"/>
-<polygon fill="#191970" stroke="#191970" points="2579.4334,-76.4373 2589.3554,-72.722 2579.2821,-69.439 2579.4334,-76.4373"/>
+<path fill="none" stroke="#191970" d="M1957.6292,-184.899C1988.338,-168.7219 2054.0986,-136.4804 2114,-123 2135.8757,-118.077 2788.6756,-86.0282 3013.095,-75.1248"/>
+<polygon fill="#191970" stroke="#191970" points="3013.3168,-78.6183 3023.1352,-74.6372 3012.9771,-71.6265 3013.3168,-78.6183"/>
</g>
<!-- Node5->Node10 -->
<g id="edge10" class="edge">
<title>Node5->Node10</title>
-<path fill="none" stroke="#191970" d="M1066.0124,-184.9581C1108.8371,-174.2082 1178.5208,-156.7162 1225.0499,-145.0365"/>
-<polygon fill="#191970" stroke="#191970" points="1226.0515,-148.3937 1234.8984,-142.5643 1224.3471,-141.6043 1226.0515,-148.3937"/>
+<path fill="none" stroke="#191970" d="M1921.875,-184.8906C1902.8917,-174.8262 1872.8627,-158.9057 1851.0693,-147.3514"/>
+<polygon fill="#191970" stroke="#191970" points="1852.4723,-144.1338 1841.9977,-142.5419 1849.1934,-150.3184 1852.4723,-144.1338"/>
</g>
<!-- Node5->Node16 -->
<g id="edge17" class="edge">
<title>Node5->Node16</title>
-<path fill="none" stroke="#191970" d="M994.2335,-184.8914C940.84,-169.9695 833.4778,-141.0666 741,-123 620.5919,-99.477 475.6145,-82.0561 412.2628,-74.9784"/>
-<polygon fill="#191970" stroke="#191970" points="412.4323,-71.4758 402.108,-73.8547 411.6623,-78.4334 412.4323,-71.4758"/>
+<path fill="none" stroke="#191970" d="M1899.0848,-184.9618C1830.9986,-169.4631 1689.9605,-138.9428 1569,-123 1306.0657,-88.3449 985.5914,-75.7372 884.2867,-72.4535"/>
+<polygon fill="#191970" stroke="#191970" points="884.205,-68.9493 874.0993,-72.1311 883.9835,-75.9458 884.205,-68.9493"/>
</g>
<!-- Node5->Node17 -->
<g id="edge18" class="edge">
<title>Node5->Node17</title>
-<path fill="none" stroke="#191970" d="M978.3849,-186.522C950.4968,-179.554 917.1444,-166.6619 896,-143 883.3284,-128.8197 878.9588,-107.1247 877.5268,-91.3511"/>
-<polygon fill="#191970" stroke="#191970" points="881.0054,-90.8823 876.9265,-81.1041 874.0174,-91.2917 881.0054,-90.8823"/>
+<path fill="none" stroke="#191970" d="M1936.7834,-184.8408C1930.9059,-168.8773 1916.7958,-137.3324 1893,-123 1820.7183,-79.4642 1598.8591,-96.3619 1515,-87 1489.7204,-84.1778 1461.5276,-80.592 1438.7195,-77.5803"/>
+<polygon fill="#191970" stroke="#191970" points="1439.0228,-74.0899 1428.6488,-76.2417 1438.1004,-81.0289 1439.0228,-74.0899"/>
</g>
<!-- Node18 -->
<g id="node15" class="node">
<title>Node18</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="1957.5,-62 1957.5,-81 2002.5,-81 2002.5,-62 1957.5,-62"/>
-<text text-anchor="middle" x="1980" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">utility</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2370.5,-62 2370.5,-81 2415.5,-81 2415.5,-62 2370.5,-62"/>
+<text text-anchor="middle" x="2393" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">utility</text>
</g>
<!-- Node5->Node18 -->
<g id="edge19" class="edge">
<title>Node5->Node18</title>
-<path fill="none" stroke="#191970" d="M1077.6334,-188.0399C1153.2327,-177.9159 1293.5634,-157.928 1342,-143 1362.1423,-136.7922 1364.6358,-128.4358 1385,-123 1492.6134,-94.275 1838.6733,-77.4916 1947.1567,-72.8378"/>
-<polygon fill="#191970" stroke="#191970" points="1947.3432,-76.3332 1957.1862,-72.4133 1947.0471,-69.3394 1947.3432,-76.3332"/>
+<path fill="none" stroke="#191970" d="M1950.0459,-184.9744C1967.3179,-169.2076 2004.4227,-137.945 2042,-123 2099.6836,-100.0584 2284.1664,-81.3137 2359.9857,-74.3817"/>
+<polygon fill="#191970" stroke="#191970" points="2360.7108,-77.8305 2370.3556,-73.4455 2360.0814,-70.8589 2360.7108,-77.8305"/>
</g>
<!-- Node5->Node20 -->
<g id="edge20" class="edge">
<title>Node5->Node20</title>
-<path fill="none" stroke="#191970" d="M1077.5587,-188.9445C1104.3355,-185.9601 1137.9734,-182.2389 1168,-179 1318.7075,-162.7435 1498.6711,-144.0786 1572.1962,-136.4847"/>
-<polygon fill="#191970" stroke="#191970" points="1572.668,-139.9547 1582.2557,-135.4462 1571.9492,-132.9917 1572.668,-139.9547"/>
+<path fill="none" stroke="#191970" d="M1972.1875,-184.8906C2010.4141,-173.4783 2073.8622,-154.5363 2112.6091,-142.9686"/>
+<polygon fill="#191970" stroke="#191970" points="2113.7883,-146.2693 2122.3692,-140.0548 2111.7858,-139.5618 2113.7883,-146.2693"/>
</g>
<!-- Node12 -->
<g id="node8" class="node">
<title>Node12</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2514.5,-.5 2514.5,-19.5 2607.5,-19.5 2607.5,-.5 2514.5,-.5"/>
-<text text-anchor="middle" x="2561" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">dlpack/dlpack.h</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2948.5,-.5 2948.5,-19.5 3041.5,-19.5 3041.5,-.5 2948.5,-.5"/>
+<text text-anchor="middle" x="2995" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">dlpack/dlpack.h</text>
</g>
<!-- Node11->Node12 -->
<g id="edge7" class="edge">
<title>Node11->Node12</title>
-<path fill="none" stroke="#191970" d="M2631.0112,-56.2977C2616.8034,-46.9022 2598.6215,-34.8787 2584.3475,-25.4395"/>
-<polygon fill="#191970" stroke="#191970" points="2585.8845,-22.2598 2575.6128,-19.6633 2582.0233,-28.0986 2585.8845,-22.2598"/>
+<path fill="none" stroke="#191970" d="M3065.0112,-56.2977C3050.8034,-46.9022 3032.6215,-34.8787 3018.3475,-25.4395"/>
+<polygon fill="#191970" stroke="#191970" points="3019.8845,-22.2598 3009.6128,-19.6633 3016.0233,-28.0986 3019.8845,-22.2598"/>
</g>
<!-- Node13 -->
<g id="node9" class="node">
<title>Node13</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2626,-.5 2626,-19.5 2682,-19.5 2682,-.5 2626,-.5"/>
-<text text-anchor="middle" x="2654" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">stddef.h</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="3060,-.5 3060,-19.5 3116,-19.5 3116,-.5 3060,-.5"/>
+<text text-anchor="middle" x="3088" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">stddef.h</text>
</g>
<!-- Node11->Node13 -->
<g id="edge8" class="edge">
<title>Node11->Node13</title>
-<path fill="none" stroke="#191970" d="M2654,-56.2977C2654,-48.3834 2654,-38.6043 2654,-30.0759"/>
-<polygon fill="#191970" stroke="#191970" points="2657.5001,-29.8469 2654,-19.8469 2650.5001,-29.847 2657.5001,-29.8469"/>
+<path fill="none" stroke="#191970" d="M3088,-56.2977C3088,-48.3834 3088,-38.6043 3088,-30.0759"/>
+<polygon fill="#191970" stroke="#191970" points="3091.5001,-29.8469 3088,-19.8469 3084.5001,-29.847 3091.5001,-29.8469"/>
</g>
<!-- Node14 -->
<g id="node10" class="node">
<title>Node14</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2700.5,-.5 2700.5,-19.5 2753.5,-19.5 2753.5,-.5 2700.5,-.5"/>
-<text text-anchor="middle" x="2727" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">stdint.h</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="3134.5,-.5 3134.5,-19.5 3187.5,-19.5 3187.5,-.5 3134.5,-.5"/>
+<text text-anchor="middle" x="3161" y="-7.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">stdint.h</text>
</g>
<!-- Node11->Node14 -->
<g id="edge9" class="edge">
<title>Node11->Node14</title>
-<path fill="none" stroke="#191970" d="M2672.0449,-56.2977C2682.8114,-47.2274 2696.4851,-35.7077 2707.4995,-26.4285"/>
-<polygon fill="#191970" stroke="#191970" points="2709.9191,-28.9667 2715.3118,-19.8469 2705.4089,-23.6132 2709.9191,-28.9667"/>
+<path fill="none" stroke="#191970" d="M3106.0449,-56.2977C3116.8114,-47.2274 3130.4851,-35.7077 3141.4995,-26.4285"/>
+<polygon fill="#191970" stroke="#191970" points="3143.9191,-28.9667 3149.3118,-19.8469 3139.4089,-23.6132 3143.9191,-28.9667"/>
</g>
<!-- Node10->Node11 -->
<g id="edge11" class="edge">
<title>Node10->Node11</title>
-<path fill="none" stroke="#191970" d="M1332.537,-127.1439C1349.4765,-125.6245 1367.9543,-124.1062 1385,-123 1837.8621,-93.611 2382.7945,-78.1897 2579.0344,-73.2731"/>
-<polygon fill="#191970" stroke="#191970" points="2579.321,-76.7671 2589.2308,-73.0192 2579.1467,-69.7693 2579.321,-76.7671"/>
+<path fill="none" stroke="#191970" d="M1883.6918,-130.0957C2091.5306,-119.9833 2783.27,-86.3267 3013.0643,-75.146"/>
+<polygon fill="#191970" stroke="#191970" points="3013.5108,-78.6285 3023.3289,-74.6466 3013.1706,-71.6368 3013.5108,-78.6285"/>
</g>
<!-- Node15 -->
<g id="node12" class="node">
<title>Node15</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="1166.5,-62 1166.5,-81 1291.5,-81 1291.5,-62 1166.5,-62"/>
-<text text-anchor="middle" x="1229" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/logging.h</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="1523.5,-62 1523.5,-81 1648.5,-81 1648.5,-62 1523.5,-62"/>
+<text text-anchor="middle" x="1586" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/logging.h</text>
</g>
<!-- Node10->Node15 -->
<g id="edge12" class="edge">
<title>Node10->Node15</title>
-<path fill="none" stroke="#191970" d="M1266.125,-123.3906C1259.6407,-114.3273 1249.7586,-100.5149 1241.8199,-89.4188"/>
-<polygon fill="#191970" stroke="#191970" points="1244.6371,-87.3412 1235.9719,-81.2449 1238.9441,-91.4143 1244.6371,-87.3412"/>
+<path fill="none" stroke="#191970" d="M1787.0736,-123.4581C1745.5598,-112.7308 1678.0635,-95.2895 1632.8647,-83.61"/>
+<polygon fill="#191970" stroke="#191970" points="1633.5707,-80.1775 1623.013,-81.0643 1631.8193,-86.9549 1633.5707,-80.1775"/>
</g>
<!-- Node10->Node16 -->
<g id="edge13" class="edge">
<title>Node10->Node16</title>
-<path fill="none" stroke="#191970" d="M1213.3691,-128.8933C1040.637,-116.9974 543.8717,-82.7857 412.2988,-73.7244"/>
-<polygon fill="#191970" stroke="#191970" points="412.3003,-70.2163 402.0834,-73.0209 411.8193,-77.1997 412.3003,-70.2163"/>
+<path fill="none" stroke="#191970" d="M1764.2959,-129.2224C1580.1272,-117.5698 1024.8379,-82.4357 884.4964,-73.5561"/>
+<polygon fill="#191970" stroke="#191970" points="884.5877,-70.055 874.3867,-72.9164 884.1457,-77.041 884.5877,-70.055"/>
</g>
<!-- Node10->Node17 -->
<g id="edge14" class="edge">
<title>Node10->Node17</title>
-<path fill="none" stroke="#191970" d="M1213.286,-123.7262C1133.2789,-111.3009 993.1035,-89.5312 921.6374,-78.4323"/>
-<polygon fill="#191970" stroke="#191970" points="922.0258,-74.9508 911.6071,-76.8746 920.9515,-81.8678 922.0258,-74.9508"/>
+<path fill="none" stroke="#191970" d="M1764.2103,-124.4487C1676.864,-111.9561 1516.3356,-88.9968 1438.5641,-77.8737"/>
+<polygon fill="#191970" stroke="#191970" points="1438.9981,-74.4002 1428.6033,-76.4491 1438.007,-81.3297 1438.9981,-74.4002"/>
</g>
<!-- Node10->Node18 -->
<g id="edge15" class="edge">
<title>Node10->Node18</title>
-<path fill="none" stroke="#191970" d="M1332.7412,-124.6496C1337.5617,-124.0604 1342.36,-123.5019 1347,-123 1576.3108,-98.1955 1853.5292,-79.541 1947.2456,-73.5481"/>
-<polygon fill="#191970" stroke="#191970" points="1947.4924,-77.0396 1957.2501,-72.912 1947.0481,-70.0537 1947.4924,-77.0396"/>
+<path fill="none" stroke="#191970" d="M1883.5499,-126.5636C2002.7968,-113.6749 2267.3958,-85.0758 2360.1236,-75.0534"/>
+<polygon fill="#191970" stroke="#191970" points="2360.744,-78.5068 2370.31,-73.9524 2359.9918,-71.5474 2360.744,-78.5068"/>
</g>
<!-- Node19 -->
<g id="node16" class="node">
<title>Node19</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="1605,-62 1605,-81 1655,-81 1655,-62 1605,-62"/>
-<text text-anchor="middle" x="1630" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">atomic</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2685,-62 2685,-81 2735,-81 2735,-62 2685,-62"/>
+<text text-anchor="middle" x="2710" y="-69" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">atomic</text>
</g>
<!-- Node10->Node19 -->
<g id="edge16" class="edge">
<title>Node10->Node19</title>
-<path fill="none" stroke="#191970" d="M1328.3895,-123.4581C1402.6301,-110.6688 1532.2673,-88.3363 1594.5779,-77.6021"/>
-<polygon fill="#191970" stroke="#191970" points="1595.4318,-81.0067 1604.6924,-75.8597 1594.2434,-74.1083 1595.4318,-81.0067"/>
+<path fill="none" stroke="#191970" d="M1883.5087,-128.8693C2053.6906,-117.0565 2538.8551,-83.3797 2674.4245,-73.9694"/>
+<polygon fill="#191970" stroke="#191970" points="2674.9226,-77.4433 2684.6562,-73.2592 2674.4379,-70.4601 2674.9226,-77.4433"/>
</g>
<!-- Node48->Node4 -->
<g id="edge24" class="edge">
<title>Node48->Node4</title>
-<path fill="none" stroke="#191970" d="M845.5936,-318.9005C853.7798,-308.552 866.9885,-291.8542 877.0543,-279.1295"/>
-<polygon fill="#191970" stroke="#191970" points="879.9614,-281.0959 883.4205,-271.0817 874.4714,-276.7531 879.9614,-281.0959"/>
+<path fill="none" stroke="#191970" d="M1707.886,-318.9005C1718.7484,-308.353 1736.4035,-291.2096 1749.5957,-278.3998"/>
+<polygon fill="#191970" stroke="#191970" points="1752.3962,-280.559 1757.1323,-271.0817 1747.5198,-275.537 1752.3962,-280.559"/>
</g>
<!-- Node48->Node5 -->
<g id="edge25" class="edge">
<title>Node48->Node5</title>
-<path fill="none" stroke="#191970" d="M861.7521,-318.827C883.1397,-309.5899 914.9662,-294.5099 940,-277 967.9771,-257.4314 996.0101,-229.3278 1012.6813,-211.4964"/>
-<polygon fill="#191970" stroke="#191970" points="1015.2992,-213.8202 1019.5004,-204.094 1010.1508,-209.0775 1015.2992,-213.8202"/>
+<path fill="none" stroke="#191970" d="M1738.0208,-326.3984C1802.851,-322.0644 1926.2938,-309.639 1953,-277 1967.6425,-259.1047 1958.9982,-231.4913 1950.3971,-213.2112"/>
+<polygon fill="#191970" stroke="#191970" points="1953.4312,-211.4564 1945.7714,-204.1367 1947.1947,-214.6355 1953.4312,-211.4564"/>
</g>
<!-- Node48->Node10 -->
<g id="edge36" class="edge">
<title>Node48->Node10</title>
-<path fill="none" stroke="#191970" d="M836.2035,-318.8681C833.6948,-302.9603 830.5379,-270.5195 841,-246 857.2823,-207.8402 868.8726,-197.516 906,-179 956.6798,-153.7252 1111.7513,-141.4417 1203.2999,-136.2461"/>
-<polygon fill="#191970" stroke="#191970" points="1203.5278,-139.739 1213.319,-135.6914 1203.1408,-132.7497 1203.5278,-139.739"/>
+<path fill="none" stroke="#191970" d="M1698.744,-318.777C1700.3124,-302.7383 1704.9628,-270.1136 1718,-246 1739.7428,-205.7844 1779.3374,-169.1602 1803.4961,-149.0847"/>
+<polygon fill="#191970" stroke="#191970" points="1805.7986,-151.7234 1811.343,-142.6951 1801.3786,-146.2953 1805.7986,-151.7234"/>
</g>
<!-- Node48->Node16 -->
<g id="edge37" class="edge">
<title>Node48->Node16</title>
-<path fill="none" stroke="#191970" d="M797.7774,-323.4304C727.3685,-312.8808 579.7966,-283.1534 480,-210 435.9225,-177.6901 403.4868,-120.1326 388.7711,-90.4158"/>
-<polygon fill="#191970" stroke="#191970" points="391.8042,-88.6469 384.315,-81.1527 385.4961,-91.6815 391.8042,-88.6469"/>
+<path fill="none" stroke="#191970" d="M1681.3403,-318.9764C1626.318,-287.5789 1452.3508,-188.7746 1425,-179 1373.2938,-160.5213 997.5332,-96.1287 884.2436,-76.938"/>
+<polygon fill="#191970" stroke="#191970" points="884.563,-73.4424 874.1192,-75.225 883.3951,-80.3443 884.563,-73.4424"/>
</g>
<!-- Node21 -->
<g id="node19" class="node">
<title>Node21</title>
<g id="a_node19"><a xlink:href="array_8h.html" target="_top" xlink:title="Runtime Array container types. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1330,-246.5 1330,-276.5 1456,-276.5 1456,-246.5 1330,-246.5"/>
-<text text-anchor="start" x="1338" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
-<text text-anchor="middle" x="1393" y="-253.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/array.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1070,-246.5 1070,-276.5 1196,-276.5 1196,-246.5 1070,-246.5"/>
+<text text-anchor="start" x="1078" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
+<text text-anchor="middle" x="1133" y="-253.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/array.h</text>
</a>
</g>
</g>
<!-- Node48->Node21 -->
<g id="edge26" class="edge">
<title>Node48->Node21</title>
-<path fill="none" stroke="#191970" d="M878.3613,-323.6276C971.1925,-312.4209 1200.971,-284.6819 1319.5678,-270.3648"/>
-<polygon fill="#191970" stroke="#191970" points="1320.1397,-273.8213 1329.6481,-269.1479 1319.3007,-266.8717 1320.1397,-273.8213"/>
+<path fill="none" stroke="#191970" d="M1657.8019,-323.3916C1633.5389,-320.329 1602.0001,-316.3849 1574,-313 1444.6285,-297.3605 1293.8059,-279.9179 1206.2824,-269.874"/>
+<polygon fill="#191970" stroke="#191970" points="1206.5229,-266.3787 1196.1892,-268.7164 1205.7252,-273.3332 1206.5229,-266.3787"/>
</g>
<!-- Node28 -->
<g id="node22" class="node">
<title>Node28</title>
<g id="a_node22"><a xlink:href="data__type_8h.html" target="_top" xlink:title="tvm/runtime/data_type.h">
-<polygon fill="#ffffff" stroke="#000000" points="905,-123.5 905,-142.5 1043,-142.5 1043,-123.5 905,-123.5"/>
-<text text-anchor="middle" x="974" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/data_type.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1422,-123.5 1422,-142.5 1560,-142.5 1560,-123.5 1422,-123.5"/>
+<text text-anchor="middle" x="1491" y="-130.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/data_type.h</text>
</a>
</g>
</g>
<!-- Node48->Node28 -->
<g id="edge31" class="edge">
<title>Node48->Node28</title>
-<path fill="none" stroke="#191970" d="M830.3286,-318.948C822.9557,-309.1047 812.3159,-292.9833 808,-277 796.5303,-234.5236 792.2855,-211.4476 822,-179 834.6375,-165.2001 876.607,-153.1019 913.2503,-144.7805"/>
-<polygon fill="#191970" stroke="#191970" points="914.4057,-148.1097 923.4155,-142.5353 912.896,-141.2745 914.4057,-148.1097"/>
+<path fill="none" stroke="#191970" d="M1687.7348,-318.8051C1654.6954,-287.6013 1550.8273,-189.5036 1508.8566,-149.8646"/>
+<polygon fill="#191970" stroke="#191970" points="1511.0005,-147.0751 1501.3271,-142.7534 1506.1941,-152.1642 1511.0005,-147.0751"/>
</g>
<!-- Node21->Node18 -->
<g id="edge29" class="edge">
<title>Node21->Node18</title>
-<path fill="none" stroke="#191970" d="M1389.6609,-246.1914C1386.6807,-228.2933 1384.7222,-198.6255 1399,-179 1418.4699,-152.2375 1504.9384,-131.2076 1537,-123 1687.135,-84.5661 1873.3092,-74.7498 1947.389,-72.2961"/>
-<polygon fill="#191970" stroke="#191970" points="1947.6003,-75.7913 1957.4878,-71.985 1947.3847,-68.7947 1947.6003,-75.7913"/>
+<path fill="none" stroke="#191970" d="M1150.6367,-246.2887C1172.326,-228.2098 1210.8481,-198.0944 1248,-179 1316.8773,-143.6002 1336.9496,-137.6141 1413,-123 1600.5079,-86.9679 2210.9271,-74.5523 2359.9561,-72.0202"/>
+<polygon fill="#191970" stroke="#191970" points="2360.3235,-75.5146 2370.2637,-71.8484 2360.2068,-68.5156 2360.3235,-75.5146"/>
</g>
<!-- Node21->Node20 -->
<g id="edge30" class="edge">
<title>Node21->Node20</title>
-<path fill="none" stroke="#191970" d="M1394.0829,-246.362C1396.241,-227.7827 1402.6965,-196.5861 1422,-179 1443.8374,-159.1055 1525.4545,-144.5012 1572.3247,-137.5446"/>
-<polygon fill="#191970" stroke="#191970" points="1572.9261,-140.9941 1582.322,-136.0985 1571.9239,-134.0662 1572.9261,-140.9941"/>
+<path fill="none" stroke="#191970" d="M1196.3677,-253.4617C1389.4824,-228.965 1966.3508,-155.7887 2112.2047,-137.287"/>
+<polygon fill="#191970" stroke="#191970" points="2112.8522,-140.733 2122.3322,-136.0023 2111.9712,-133.7886 2112.8522,-140.733"/>
</g>
<!-- Node21->Node22 -->
<g id="edge27" class="edge">
<title>Node21->Node22</title>
-<path fill="none" stroke="#191970" d="M1329.9612,-258.4776C1184.3132,-251.1654 833.1055,-231.5273 784,-210 754.4156,-197.0305 728.3979,-169.2809 713.4325,-150.8814"/>
-<polygon fill="#191970" stroke="#191970" points="715.9378,-148.4049 707.0064,-142.7057 710.4343,-152.7307 715.9378,-148.4049"/>
+<path fill="none" stroke="#191970" d="M1069.9551,-248.739C1023.3477,-238.9917 958.448,-224.7607 902,-210 824.4961,-189.7334 735.125,-161.791 685.124,-145.7371"/>
+<polygon fill="#191970" stroke="#191970" points="685.9587,-142.329 675.3672,-142.5948 683.8127,-148.9919 685.9587,-142.329"/>
</g>
<!-- Node21->Node23 -->
<g id="edge28" class="edge">
<title>Node21->Node23</title>
-<path fill="none" stroke="#191970" d="M1456.1869,-257.2981C1625.5389,-245.9922 2078.9808,-215.4125 2111,-210 2163.1951,-201.177 2174.0507,-189.1706 2226,-179 2345.7094,-155.5635 2489.6131,-141.7457 2557.6636,-136.0553"/>
-<polygon fill="#191970" stroke="#191970" points="2558.2394,-139.5199 2567.9187,-135.2116 2557.6653,-132.5434 2558.2394,-139.5199"/>
+<path fill="none" stroke="#191970" d="M1069.827,-248.6385C1006.9123,-235.5288 918.9565,-216.2572 914,-210 901.1877,-193.8253 905.6035,-169.2388 911.1539,-152.2187"/>
+<polygon fill="#191970" stroke="#191970" points="914.472,-153.3358 914.6126,-142.7419 907.8962,-150.9359 914.472,-153.3358"/>
</g>
<!-- Node28->Node11 -->
<g id="edge32" class="edge">
<title>Node28->Node11</title>
-<path fill="none" stroke="#191970" d="M1043.2348,-129.8628C1089.1189,-127.8182 1150.6502,-125.1422 1205,-123 1731.1897,-102.2601 2365.0181,-81.0257 2579.1807,-73.9548"/>
-<polygon fill="#191970" stroke="#191970" points="2579.4837,-77.4468 2589.3628,-73.6189 2579.2528,-70.4507 2579.4837,-77.4468"/>
+<path fill="none" stroke="#191970" d="M1560.2462,-124.1891C1564.8964,-123.7419 1569.5134,-123.3386 1574,-123 2092.7706,-83.846 2224.1274,-106.7093 2744,-87 2836.8115,-83.4814 2943.6668,-78.5249 3013.3383,-75.1716"/>
+<polygon fill="#191970" stroke="#191970" points="3013.5221,-78.6669 3023.3418,-74.6889 3013.1846,-71.6751 3013.5221,-78.6669"/>
</g>
<!-- Node28->Node15 -->
<g id="edge33" class="edge">
<title>Node28->Node15</title>
-<path fill="none" stroke="#191970" d="M1013.564,-123.4581C1058.23,-112.6857 1130.9687,-95.1428 1179.397,-83.4631"/>
-<polygon fill="#191970" stroke="#191970" points="1180.4426,-86.8113 1189.3432,-81.0643 1178.8013,-80.0065 1180.4426,-86.8113"/>
+<path fill="none" stroke="#191970" d="M1505.8438,-123.3906C1521.0418,-113.5519 1544.8848,-98.1167 1562.6206,-86.6351"/>
+<polygon fill="#191970" stroke="#191970" points="1564.768,-89.4144 1571.2605,-81.0419 1560.9639,-83.5382 1564.768,-89.4144"/>
</g>
<!-- Node28->Node16 -->
<g id="edge34" class="edge">
<title>Node28->Node16</title>
-<path fill="none" stroke="#191970" d="M904.9895,-125.855C776.4976,-112.5515 505.1843,-84.461 412.1224,-74.8258"/>
-<polygon fill="#191970" stroke="#191970" points="412.4827,-71.3445 402.1754,-73.7959 411.7618,-78.3073 412.4827,-71.3445"/>
+<path fill="none" stroke="#191970" d="M1421.7043,-126.3307C1284.5413,-113.1296 983.1614,-84.1235 884.1907,-74.5982"/>
+<polygon fill="#191970" stroke="#191970" points="884.4853,-71.1104 874.1959,-73.6362 883.8146,-78.0782 884.4853,-71.1104"/>
</g>
<!-- Node28->Node17 -->
<g id="edge35" class="edge">
<title>Node28->Node17</title>
-<path fill="none" stroke="#191970" d="M958.8438,-123.3906C943.3257,-113.5519 918.9807,-98.1167 900.8716,-86.6351"/>
-<polygon fill="#191970" stroke="#191970" points="902.3696,-83.4407 892.0498,-81.0419 898.6213,-89.3526 902.3696,-83.4407"/>
+<path fill="none" stroke="#191970" d="M1475.8438,-123.3906C1460.3257,-113.5519 1435.9807,-98.1167 1417.8716,-86.6351"/>
+<polygon fill="#191970" stroke="#191970" points="1419.3696,-83.4407 1409.0498,-81.0419 1415.6213,-89.3526 1419.3696,-83.4407"/>
</g>
<!-- Node33->Node10 -->
<g id="edge41" class="edge">
<title>Node33->Node10</title>
-<path fill="none" stroke="#191970" d="M1273,-179.2977C1273,-171.3834 1273,-161.6043 1273,-153.0759"/>
-<polygon fill="#191970" stroke="#191970" points="1276.5001,-152.8469 1273,-142.8469 1269.5001,-152.847 1276.5001,-152.8469"/>
+<path fill="none" stroke="#191970" d="M1082.167,-189.6742C1230.3566,-178.3529 1599.4647,-150.1539 1754.1037,-138.3399"/>
+<polygon fill="#191970" stroke="#191970" points="1754.7138,-141.8036 1764.4181,-137.5519 1754.1805,-134.8239 1754.7138,-141.8036"/>
</g>
<!-- Node33->Node15 -->
<g id="edge40" class="edge">
<title>Node33->Node15</title>
-<path fill="none" stroke="#191970" d="M1241.8046,-179.3802C1227.8335,-170.8126 1212.7783,-158.6483 1205,-143 1196.3122,-125.5222 1205.8789,-104.1565 1215.3873,-89.3563"/>
-<polygon fill="#191970" stroke="#191970" points="1218.3041,-91.2922 1221.1364,-81.083 1212.5557,-87.2977 1218.3041,-91.2922"/>
+<path fill="none" stroke="#191970" d="M1045.7184,-179.4416C1081.257,-159.5205 1140.9062,-126.501 1152,-123 1168.962,-117.6471 1393.9734,-92.5 1513.1164,-79.4319"/>
+<polygon fill="#191970" stroke="#191970" points="1513.6971,-82.8893 1523.2563,-78.3208 1512.9346,-75.931 1513.6971,-82.8893"/>
</g>
<!-- Node33->Node16 -->
<g id="edge44" class="edge">
<title>Node33->Node16</title>
-<path fill="none" stroke="#191970" d="M1209.7117,-181.7849C1172.5284,-173.1882 1125.1045,-160.246 1085,-143 1069.2449,-136.2249 1068.329,-128.2428 1052,-123 930.7312,-84.0639 529.735,-74.1065 412.4746,-71.9987"/>
-<polygon fill="#191970" stroke="#191970" points="412.3983,-68.4969 402.3393,-71.8236 412.2773,-75.4959 412.3983,-68.4969"/>
+<path fill="none" stroke="#191970" d="M1008.591,-179.2277C997.3149,-163.5582 978.0886,-139.2396 957,-123 934.6796,-105.8118 905.7098,-92.127 883.8553,-83.197"/>
+<polygon fill="#191970" stroke="#191970" points="884.9501,-79.866 874.3641,-79.4331 882.3696,-86.3731 884.9501,-79.866"/>
</g>
<!-- Node33->Node17 -->
<g id="edge47" class="edge">
<title>Node33->Node17</title>
-<path fill="none" stroke="#191970" d="M1230.5518,-179.4131C1202.7026,-169.3853 1165.5499,-155.7615 1133,-143 1111.4835,-134.5642 1106.9469,-130.2426 1085,-123 1029.681,-104.7444 964.214,-89.5532 921.711,-80.5079"/>
-<polygon fill="#191970" stroke="#191970" points="922.191,-77.0323 911.6845,-78.3976 920.7492,-83.8822 922.191,-77.0323"/>
+<path fill="none" stroke="#191970" d="M1034.3345,-179.308C1051.9927,-162.7457 1082.6221,-136.77 1114,-123 1192.1859,-88.6885 1292.3509,-77.233 1349.2301,-73.411"/>
+<polygon fill="#191970" stroke="#191970" points="1349.5686,-76.8968 1359.3319,-72.7824 1349.1338,-69.9103 1349.5686,-76.8968"/>
</g>
<!-- Node33->Node18 -->
<g id="edge46" class="edge">
<title>Node33->Node18</title>
-<path fill="none" stroke="#191970" d="M1304.7859,-179.4278C1341.527,-162.717 1404.2353,-136.3609 1461,-123 1554.3309,-101.0324 1848.4385,-80.1654 1947.2518,-73.6124"/>
-<polygon fill="#191970" stroke="#191970" points="1947.5078,-77.1032 1957.2563,-72.9537 1947.0479,-70.1183 1947.5078,-77.1032"/>
+<path fill="none" stroke="#191970" d="M1059.8306,-179.448C1108.2578,-162.3572 1191.7833,-135.2604 1266,-123 1484.6568,-86.8786 2197.4068,-74.3751 2359.8424,-71.9601"/>
+<polygon fill="#191970" stroke="#191970" points="2360.2072,-75.4553 2370.155,-71.8095 2360.1049,-68.456 2360.2072,-75.4553"/>
</g>
<!-- Node33->Node20 -->
<g id="edge48" class="edge">
<title>Node33->Node20</title>
-<path fill="none" stroke="#191970" d="M1336.0262,-182.86C1406.1731,-169.905 1516.9166,-149.4523 1572.2264,-139.2375"/>
-<polygon fill="#191970" stroke="#191970" points="1573.0058,-142.6528 1582.2039,-137.3948 1571.7345,-135.7692 1573.0058,-142.6528"/>
+<path fill="none" stroke="#191970" d="M1082.2579,-190.6517C1135.3194,-187.4491 1213.1003,-182.8139 1281,-179 1604.604,-160.8234 1996.417,-140.6487 2112.2321,-134.7232"/>
+<polygon fill="#191970" stroke="#191970" points="2112.4531,-138.2165 2122.2613,-134.2104 2112.0956,-131.2257 2112.4531,-138.2165"/>
</g>
<!-- Node33->Node22 -->
<g id="edge42" class="edge">
<title>Node33->Node22</title>
-<path fill="none" stroke="#191970" d="M1209.765,-187.713C1091.9821,-175.0714 841.9306,-148.2334 742.2781,-137.5377"/>
-<polygon fill="#191970" stroke="#191970" points="742.5923,-134.0514 732.2758,-136.4642 741.8452,-141.0114 742.5923,-134.0514"/>
+<path fill="none" stroke="#191970" d="M955.7256,-184.0674C879.5702,-171.5109 753.7967,-150.7734 688.421,-139.9943"/>
+<polygon fill="#191970" stroke="#191970" points="688.9002,-136.5262 678.464,-138.3526 687.7614,-143.4329 688.9002,-136.5262"/>
</g>
<!-- Node33->Node23 -->
<g id="edge43" class="edge">
<title>Node33->Node23</title>
-<path fill="none" stroke="#191970" d="M1336.3611,-186.9973C1362.8007,-184.1037 1393.8536,-181.0128 1422,-179 1866.9936,-147.1772 2408.4799,-136.1646 2557.4537,-133.6227"/>
-<polygon fill="#191970" stroke="#191970" points="2557.9529,-137.1149 2567.8928,-133.4475 2557.8355,-130.1158 2557.9529,-137.1149"/>
+<path fill="none" stroke="#191970" d="M994.5372,-179.4554C978.8765,-169.8241 958.6468,-157.3828 943.0644,-147.7996"/>
+<polygon fill="#191970" stroke="#191970" points="944.848,-144.7876 934.4965,-142.5303 941.181,-150.7503 944.848,-144.7876"/>
</g>
<!-- Node33->Node36 -->
<g id="edge45" class="edge">
<title>Node33->Node36</title>
-<path fill="none" stroke="#191970" d="M1336.3481,-186.8066C1362.7857,-183.883 1393.8405,-180.8213 1422,-179 1773.5172,-156.2646 2900.2667,-137.9436 3167.0015,-133.8542"/>
-<polygon fill="#191970" stroke="#191970" points="3167.2823,-137.3504 3177.2276,-133.6979 3167.1753,-130.3512 3167.2823,-137.3504"/>
+<path fill="none" stroke="#191970" d="M955.7739,-189.0693C817.8542,-177.2228 491.7911,-149.216 359.8721,-137.885"/>
+<polygon fill="#191970" stroke="#191970" points="359.8521,-134.3704 349.5892,-137.0017 359.253,-141.3448 359.8521,-134.3704"/>
</g>
<!-- Node49->Node3 -->
<g id="edge55" class="edge">
<title>Node49->Node3</title>
-<path fill="none" stroke="#191970" d="M1383.9933,-503.3527C1356.4868,-486.9188 1301.1671,-454.5365 1280,-447 1201.9244,-419.2013 1106.0509,-405.7508 1048.879,-399.7613"/>
-<polygon fill="#191970" stroke="#191970" points="1048.9687,-396.2525 1038.6671,-398.7279 1048.2639,-403.217 1048.9687,-396.2525"/>
+<path fill="none" stroke="#191970" d="M1174.0441,-503.3845C1159.8484,-483.761 1127.3433,-438.8274 1109.0771,-413.5772"/>
+<polygon fill="#191970" stroke="#191970" points="1111.749,-411.2992 1103.052,-405.2484 1106.0775,-415.4021 1111.749,-411.2992"/>
</g>
<!-- Node49->Node5 -->
<g id="edge61" class="edge">
<title>Node49->Node5</title>
-<path fill="none" stroke="#191970" d="M1388.8668,-503.4679C1336.4652,-458.6026 1114.1436,-268.2547 1047.1699,-210.913"/>
-<polygon fill="#191970" stroke="#191970" points="1049.327,-208.1523 1039.4545,-204.3072 1044.7744,-213.4696 1049.327,-208.1523"/>
+<path fill="none" stroke="#191970" d="M1220.6404,-503.4273C1258.7082,-494.2044 1317.8227,-479.8041 1369,-467 1647.3502,-397.359 1834.4328,-520.631 1986,-277 1999.8973,-254.6614 1978.3109,-227.9211 1960.2567,-211.069"/>
+<polygon fill="#191970" stroke="#191970" points="1962.1589,-208.0787 1952.3503,-204.0733 1957.5203,-213.3211 1962.1589,-208.0787"/>
</g>
<!-- Node49->Node11 -->
<g id="edge62" class="edge">
<title>Node49->Node11</title>
-<path fill="none" stroke="#191970" d="M1441.6261,-507.9071C1456.8312,-506.1703 1474.1777,-504.3327 1490,-503 1534.3884,-499.2613 3060.4184,-443.3894 3091,-411 3164.2724,-333.3961 3099.0637,-259.5127 3029,-179 2998.8131,-144.3111 2984.7984,-139.8186 2942,-123 2872.6209,-95.7359 2788.1379,-82.8063 2728.6699,-76.7361"/>
-<polygon fill="#191970" stroke="#191970" points="2728.9017,-73.2421 2718.608,-75.75 2728.2189,-80.2088 2728.9017,-73.2421"/>
+<path fill="none" stroke="#191970" d="M1222.6895,-512.1077C1359.3143,-508.9613 1805.8911,-496.9181 2174,-467 2731.504,-421.6889 3069.2926,-656.6452 3406,-210 3473.1193,-120.9657 3276.7854,-88.2706 3162.99,-76.9937"/>
+<polygon fill="#191970" stroke="#191970" points="3163.2097,-73.4988 3152.9217,-76.0302 3162.5428,-80.467 3163.2097,-73.4988"/>
</g>
<!-- Node49->Node16 -->
<g id="edge79" class="edge">
<title>Node49->Node16</title>
-<path fill="none" stroke="#191970" d="M1358.1838,-511.5775C1154.0076,-504.5999 271.1272,-474.0445 259,-467 116.8404,-384.4217 18.7937,-250.1597 123,-123 151.0837,-88.7303 285.1513,-76.7975 347.7127,-73.0487"/>
-<polygon fill="#191970" stroke="#191970" points="348.0316,-76.5364 357.8181,-72.4776 347.6365,-69.5475 348.0316,-76.5364"/>
+<path fill="none" stroke="#191970" d="M1139.3637,-508.8119C1119.059,-506.8636 1094.2695,-504.6237 1072,-503 924.3977,-492.2378 549.8536,-505.6667 407,-467 316.9811,-442.6343 278.8277,-427.3519 237,-344 192.9094,-256.1386 153.0723,-186.6353 228,-123 273.3903,-84.4505 698.3005,-74.1618 819.7327,-71.9981"/>
+<polygon fill="#191970" stroke="#191970" points="819.9411,-75.4951 829.8795,-71.8238 819.8209,-68.4961 819.9411,-75.4951"/>
</g>
<!-- Node49->Node18 -->
<g id="edge81" class="edge">
<title>Node49->Node18</title>
-<path fill="none" stroke="#191970" d="M1441.6784,-508.2625C1504.7468,-500.6478 1621.098,-484.7307 1659,-467 1703.715,-446.0821 1744,-444.8659 1744,-395.5 1744,-395.5 1744,-395.5 1744,-328.5 1744,-207.5011 1887.6775,-118.8606 1950.8214,-85.7472"/>
-<polygon fill="#191970" stroke="#191970" points="1952.5862,-88.7754 1959.8788,-81.0897 1949.3851,-82.5502 1952.5862,-88.7754"/>
+<path fill="none" stroke="#191970" d="M1222.5155,-508.5495C1366.5887,-492.4298 1842.8086,-433.304 1964,-344 2059.5662,-273.5788 2014.6067,-187.9079 2114,-123 2154.238,-96.723 2295.2401,-80.5824 2360.0114,-74.3929"/>
+<polygon fill="#191970" stroke="#191970" points="2360.4584,-77.8665 2370.0896,-73.4517 2359.8075,-70.8968 2360.4584,-77.8665"/>
</g>
<!-- Node49->Node21 -->
<g id="edge63" class="edge">
<title>Node49->Node21</title>
-<path fill="none" stroke="#191970" d="M1399.7341,-503.448C1398.7412,-467.7725 1395.2293,-341.5945 1393.7096,-286.9932"/>
-<polygon fill="#191970" stroke="#191970" points="1397.2023,-286.6791 1393.4253,-276.7804 1390.205,-286.8739 1397.2023,-286.6791"/>
+<path fill="none" stroke="#191970" d="M1179.177,-503.448C1172.3539,-467.698 1148.1858,-341.0673 1137.8004,-286.6522"/>
+<polygon fill="#191970" stroke="#191970" points="1141.2291,-285.9469 1135.9163,-276.7804 1134.3532,-287.2593 1141.2291,-285.9469"/>
</g>
<!-- Node49->Node22 -->
<g id="edge77" class="edge">
<title>Node49->Node22</title>
-<path fill="none" stroke="#191970" d="M1358.353,-511.6649C1170.8169,-505.5877 411.743,-480.2646 363,-467 297.4868,-449.1717 228,-463.3957 228,-395.5 228,-395.5 228,-395.5 228,-328.5 228,-270.5392 280.2445,-277.3426 329,-246 385.6799,-209.5632 401.1573,-200.5489 465,-179 530.5041,-156.8903 610.0708,-144.0944 657.7884,-137.8417"/>
-<polygon fill="#191970" stroke="#191970" points="658.2829,-141.3071 667.7597,-136.5699 657.3972,-134.3633 658.2829,-141.3071"/>
+<path fill="none" stroke="#191970" d="M1139.3563,-508.9097C1119.0499,-506.9841 1094.2611,-504.7351 1072,-503 922.6074,-491.356 481.5951,-539.1902 404,-411 335.3628,-297.6086 536.2407,-185.7069 615.4187,-147.1228"/>
+<polygon fill="#191970" stroke="#191970" points="617.3799,-150.0638 624.8796,-142.5801 614.3499,-143.7535 617.3799,-150.0638"/>
</g>
<!-- Node49->Node28 -->
<g id="edge71" class="edge">
<title>Node49->Node28</title>
-<path fill="none" stroke="#191970" d="M1358.0924,-512.0345C1244.2458,-509.0637 919.8246,-498.2849 652,-467 506.8518,-450.0451 422.1231,-522.7868 328,-411 165.1132,-217.5449 688.7021,-155.1484 894.55,-138.3753"/>
-<polygon fill="#191970" stroke="#191970" points="895.0094,-141.8498 904.6988,-137.5644 894.4518,-134.8721 895.0094,-141.8498"/>
+<path fill="none" stroke="#191970" d="M1194.0141,-503.2742C1216.9109,-484.9385 1262,-442.8235 1262,-395.5 1262,-395.5 1262,-395.5 1262,-261.5 1262,-223.8735 1254.2519,-205.463 1281,-179 1299.5311,-160.6664 1360.3641,-148.6445 1411.8172,-141.5048"/>
+<polygon fill="#191970" stroke="#191970" points="1412.5224,-144.9416 1421.9675,-140.1416 1411.5906,-138.0039 1412.5224,-144.9416"/>
</g>
<!-- Node49->Node33 -->
<g id="edge70" class="edge">
<title>Node49->Node33</title>
-<path fill="none" stroke="#191970" d="M1396.1991,-503.4679C1379.0792,-460.5332 1308.834,-284.3671 1282.782,-219.0321"/>
-<polygon fill="#191970" stroke="#191970" points="1285.9502,-217.5276 1278.9952,-209.5352 1279.448,-220.1204 1285.9502,-217.5276"/>
+<path fill="none" stroke="#191970" d="M1147.4753,-503.4317C1072.6762,-479.8795 894.9672,-411.0319 847,-277 842.3576,-264.0279 838.7727,-257.0516 847,-246 859.3209,-229.4495 904.8965,-216.2441 946.0374,-207.3502"/>
+<polygon fill="#191970" stroke="#191970" points="946.7597,-210.7749 955.8272,-205.2947 945.3213,-203.9243 946.7597,-210.7749"/>
</g>
<!-- Node49->Node36 -->
<g id="edge80" class="edge">
<title>Node49->Node36</title>
-<path fill="none" stroke="#191970" d="M1441.6067,-507.6558C1456.8099,-505.8937 1474.1601,-504.1042 1490,-503 2049.3154,-464.0092 2192.8144,-519.7524 2751,-467 2806.5757,-461.7477 2819.8182,-455.4386 2875,-447 2930.2662,-438.5485 3080.687,-446.352 3124,-411 3134.6365,-402.3186 3197.4505,-213.6381 3217.7071,-152.1696"/>
-<polygon fill="#191970" stroke="#191970" points="3221.0747,-153.1328 3220.875,-142.5399 3214.4252,-150.9453 3221.0747,-153.1328"/>
+<path fill="none" stroke="#191970" d="M1139.3616,-508.8404C1119.0564,-506.8986 1094.2671,-504.6561 1072,-503 1001.9683,-497.7914 504.5451,-496.8929 441,-467 319.6774,-409.9275 304.7436,-215.3982 303.1209,-152.6695"/>
+<polygon fill="#191970" stroke="#191970" points="306.6185,-152.4928 302.9501,-142.5533 299.6195,-152.611 306.6185,-152.4928"/>
</g>
<!-- Node49->Node44 -->
<g id="edge78" class="edge">
<title>Node49->Node44</title>
-<path fill="none" stroke="#191970" d="M1441.6781,-507.7383C1506.6229,-499.1363 1625.9267,-481.5616 1639,-467 1686.0001,-414.6494 1670.8575,-321.5701 1661.2844,-281.1456"/>
-<polygon fill="#191970" stroke="#191970" points="1664.621,-280.0624 1658.7882,-271.2177 1657.8323,-281.7693 1664.621,-280.0624"/>
+<path fill="none" stroke="#191970" d="M1160.9365,-503.3385C1115.3552,-480.6426 1002.2867,-420.0545 929,-344 910.5519,-324.8552 895.4599,-297.9435 886.6016,-280.128"/>
+<polygon fill="#191970" stroke="#191970" points="889.722,-278.5406 882.234,-271.0452 883.4135,-281.5742 889.722,-278.5406"/>
</g>
<!-- Node8 -->
<g id="node27" class="node">
<title>Node8</title>
<g id="a_node27"><a xlink:href="functor_8h.html" target="_top" xlink:title="Defines the Functor data structures. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1430.5,-185 1430.5,-204 1539.5,-204 1539.5,-185 1430.5,-185"/>
-<text text-anchor="middle" x="1485" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/node/functor.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="2236.5,-185 2236.5,-204 2345.5,-204 2345.5,-185 2236.5,-185"/>
+<text text-anchor="middle" x="2291" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/node/functor.h</text>
</a>
</g>
</g>
<!-- Node49->Node8 -->
<g id="edge56" class="edge">
<title>Node49->Node8</title>
-<path fill="none" stroke="#191970" d="M1402.8343,-503.1369C1412.1003,-470.8041 1442.2093,-365.0435 1465,-277 1470.5214,-255.6701 1476.4148,-231.1172 1480.3626,-214.38"/>
-<polygon fill="#191970" stroke="#191970" points="1483.8796,-214.7122 1482.7564,-204.1771 1477.0647,-213.1132 1483.8796,-214.7122"/>
+<path fill="none" stroke="#191970" d="M1222.7432,-512.1801C1376.0423,-508.9497 1902.6604,-495.8405 1972,-467 2066.7398,-427.5948 2065.8095,-379.4427 2144,-313 2188.3703,-275.2962 2241.9375,-232.865 2270.5782,-210.4181"/>
+<polygon fill="#191970" stroke="#191970" points="2272.9086,-213.0388 2278.6287,-204.1208 2268.5957,-207.5253 2272.9086,-213.0388"/>
</g>
<!-- Node42 -->
<g id="node28" class="node">
<title>Node42</title>
<g id="a_node28"><a xlink:href="map_8h.html" target="_top" xlink:title="Runtime Map container types. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="2691,-246.5 2691,-276.5 2817,-276.5 2817,-246.5 2691,-246.5"/>
-<text text-anchor="start" x="2699" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
-<text text-anchor="middle" x="2754" y="-253.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/map.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="636,-246.5 636,-276.5 762,-276.5 762,-246.5 636,-246.5"/>
+<text text-anchor="start" x="644" y="-264.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
+<text text-anchor="middle" x="699" y="-253.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/map.h</text>
</a>
</g>
</g>
<!-- Node49->Node42 -->
<g id="edge64" class="edge">
<title>Node49->Node42</title>
-<path fill="none" stroke="#191970" d="M1441.6496,-508.0508C1522.4674,-498.3807 1697.8935,-477.0475 1758,-467 2105.539,-408.9046 2516.5199,-316.4984 2680.814,-278.584"/>
-<polygon fill="#191970" stroke="#191970" points="2681.8511,-281.9367 2690.8063,-276.275 2680.2751,-275.1164 2681.8511,-281.9367"/>
+<path fill="none" stroke="#191970" d="M1139.3354,-509.1583C1119.024,-507.2905 1094.2372,-505.0182 1072,-503 1027.4484,-498.9565 701.1849,-500.0162 671,-467 624.8088,-416.4762 662.6696,-327.654 685.2147,-285.285"/>
+<polygon fill="#191970" stroke="#191970" points="688.2893,-286.9575 690.0242,-276.5057 682.1501,-283.5943 688.2893,-286.9575"/>
</g>
<!-- Node50 -->
<g id="node30" class="node">
<title>Node50</title>
<g id="a_node30"><a xlink:href="buffer_8h.html" target="_top" xlink:title="Symbolic n-dimensional array, to represent a memory buffer. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1181.5,-447.5 1181.5,-466.5 1270.5,-466.5 1270.5,-447.5 1181.5,-447.5"/>
-<text text-anchor="middle" x="1226" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/buffer.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="680.5,-447.5 680.5,-466.5 769.5,-466.5 769.5,-447.5 680.5,-447.5"/>
+<text text-anchor="middle" x="725" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/buffer.h</text>
</a>
</g>
</g>
<!-- Node49->Node50 -->
<g id="edge72" class="edge">
<title>Node49->Node50</title>
-<path fill="none" stroke="#191970" d="M1370.0883,-503.3733C1341.1548,-494.0613 1297.1389,-479.8953 1265.3707,-469.671"/>
-<polygon fill="#191970" stroke="#191970" points="1266.1128,-466.2331 1255.5213,-466.5011 1263.9682,-472.8965 1266.1128,-466.2331"/>
+<path fill="none" stroke="#191970" d="M1139.4711,-508.9138C1067.1174,-501.6327 913.309,-485.4369 784,-467 782.6983,-466.8144 781.3811,-466.6227 780.0531,-466.426"/>
+<polygon fill="#191970" stroke="#191970" points="780.1557,-462.9006 769.7392,-464.8368 779.0896,-469.8189 780.1557,-462.9006"/>
</g>
<!-- Node8->Node10 -->
<g id="edge57" class="edge">
<title>Node8->Node10</title>
-<path fill="none" stroke="#191970" d="M1451.875,-184.8906C1415.1589,-174.2395 1355.8336,-157.0296 1315.6662,-145.3772"/>
-<polygon fill="#191970" stroke="#191970" points="1316.4716,-141.9666 1305.8924,-142.5419 1314.5213,-148.6895 1316.4716,-141.9666"/>
+<path fill="none" stroke="#191970" d="M2236.0559,-187.2643C2150.8825,-176.0477 1987.0862,-154.4771 1893.7055,-142.1796"/>
+<polygon fill="#191970" stroke="#191970" points="1893.8799,-138.6725 1883.5084,-140.8368 1892.9658,-145.6125 1893.8799,-138.6725"/>
</g>
<!-- Node8->Node17 -->
<g id="edge58" class="edge">
<title>Node8->Node17</title>
-<path fill="none" stroke="#191970" d="M1430.354,-185.4036C1373.7845,-175.7637 1282.9399,-159.6582 1205,-143 1168.3157,-135.1594 1159.5856,-131.2889 1123,-123 1052.8771,-107.1128 971.1058,-90.3409 921.7342,-80.4053"/>
-<polygon fill="#191970" stroke="#191970" points="922.1998,-76.929 911.7065,-78.3918 920.8217,-83.7921 922.1998,-76.929"/>
+<path fill="none" stroke="#191970" d="M2280.2303,-184.8563C2261.3359,-168.6146 2220.2868,-136.2785 2179,-123 2038.3243,-77.7565 1662.1637,-100.397 1515,-87 1489.6681,-84.6939 1461.4721,-81.1399 1438.6752,-78.0166"/>
+<polygon fill="#191970" stroke="#191970" points="1438.9973,-74.5278 1428.6107,-76.6179 1438.0337,-81.4611 1438.9973,-74.5278"/>
</g>
<!-- Node8->Node18 -->
<g id="edge59" class="edge">
<title>Node8->Node18</title>
-<path fill="none" stroke="#191970" d="M1493.4761,-184.8738C1508.1202,-168.959 1539.8838,-137.4838 1574,-123 1641.3442,-94.4095 1863.0825,-78.4903 1947.2248,-73.3578"/>
-<polygon fill="#191970" stroke="#191970" points="1947.47,-76.8495 1957.2431,-72.7583 1947.0518,-69.862 1947.47,-76.8495"/>
+<path fill="none" stroke="#191970" d="M2299.0038,-184.8484C2316.1346,-164.1906 2356.5651,-115.4362 2378.4014,-89.1042"/>
+<polygon fill="#191970" stroke="#191970" points="2381.3442,-91.0386 2385.0334,-81.1068 2375.9559,-86.5702 2381.3442,-91.0386"/>
</g>
<!-- Node8->Node20 -->
<g id="edge60" class="edge">
<title>Node8->Node20</title>
-<path fill="none" stroke="#191970" d="M1503.9063,-184.8906C1523.7966,-174.7811 1555.3123,-158.7628 1578.0692,-147.1962"/>
-<polygon fill="#191970" stroke="#191970" points="1579.8978,-150.193 1587.2265,-142.5419 1576.726,-143.9528 1579.8978,-150.193"/>
+<path fill="none" stroke="#191970" d="M2268.3438,-184.8906C2244.0826,-174.6005 2205.3882,-158.1888 2178.02,-146.5809"/>
+<polygon fill="#191970" stroke="#191970" points="2179.07,-143.2245 2168.4972,-142.5419 2176.3367,-149.6688 2179.07,-143.2245"/>
</g>
<!-- Node42->Node18 -->
<g id="edge67" class="edge">
<title>Node42->Node18</title>
-<path fill="none" stroke="#191970" d="M2798.6929,-246.3931C2815.3418,-238.3419 2832.4898,-226.5846 2842,-210 2848.8537,-198.0479 2850.3742,-189.9407 2842,-179 2790.4723,-111.6802 2164.4539,-79.6858 2012.8864,-72.8971"/>
-<polygon fill="#191970" stroke="#191970" points="2012.9527,-69.3967 2002.8079,-72.4517 2012.6436,-76.3899 2012.9527,-69.3967"/>
+<path fill="none" stroke="#191970" d="M762.3519,-254.9333C856.8739,-244.8372 1030.2103,-225.0886 1091,-210 1198.6878,-183.2708 1216.3022,-145.2686 1325,-123 1530.1197,-80.9778 2203.0455,-73.0251 2360.0442,-71.7286"/>
+<polygon fill="#191970" stroke="#191970" points="2360.471,-75.2254 2370.4429,-71.6464 2360.4156,-68.2257 2360.471,-75.2254"/>
</g>
<!-- Node42->Node22 -->
<g id="edge65" class="edge">
<title>Node42->Node22</title>
-<path fill="none" stroke="#191970" d="M2690.9977,-259.1476C2604.7785,-255.9644 2445.1832,-250.201 2309,-246 2011.2564,-236.8152 1263.4134,-255.3473 969,-210 882.8056,-196.7238 785.2086,-164.199 734.5938,-145.9529"/>
-<polygon fill="#191970" stroke="#191970" points="735.7086,-142.6341 725.1146,-142.504 733.3152,-149.2122 735.7086,-142.6341"/>
+<path fill="none" stroke="#191970" d="M692.7121,-246.2548C683.024,-222.7656 664.4623,-177.7623 653.8875,-152.1235"/>
+<polygon fill="#191970" stroke="#191970" points="657.0641,-150.6458 650.0155,-142.7358 650.5929,-153.3149 657.0641,-150.6458"/>
</g>
<!-- Node42->Node36 -->
<g id="edge66" class="edge">
<title>Node42->Node36</title>
-<path fill="none" stroke="#191970" d="M2809.2971,-246.3816C2903.4803,-220.6314 3093.1864,-168.765 3179.4647,-145.1761"/>
-<polygon fill="#191970" stroke="#191970" points="3180.4913,-148.524 3189.2142,-142.5106 3178.6451,-141.7718 3180.4913,-148.524"/>
+<path fill="none" stroke="#191970" d="M652.4093,-246.3816C573.6188,-220.8144 415.4852,-169.5009 342.0863,-145.6833"/>
+<polygon fill="#191970" stroke="#191970" points="342.901,-142.2681 332.3089,-142.5106 340.7403,-148.9263 342.901,-142.2681"/>
</g>
<!-- Node31 -->
<g id="node29" class="node">
<title>Node31</title>
<g id="a_node29"><a xlink:href="optional_8h.html" target="_top" xlink:title="Runtime Optional container types. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="2318,-179.5 2318,-209.5 2444,-209.5 2444,-179.5 2318,-179.5"/>
-<text text-anchor="start" x="2326" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
-<text text-anchor="middle" x="2381" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/optional.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="2516,-179.5 2516,-209.5 2642,-209.5 2642,-179.5 2516,-179.5"/>
+<text text-anchor="start" x="2524" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
+<text text-anchor="middle" x="2579" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/optional.h</text>
</a>
</g>
</g>
<!-- Node42->Node31 -->
<g id="edge68" class="edge">
<title>Node42->Node31</title>
-<path fill="none" stroke="#191970" d="M2690.7256,-250.1344C2625.2517,-238.3736 2523.1023,-220.0251 2454.1905,-207.6468"/>
-<polygon fill="#191970" stroke="#191970" points="2454.734,-204.1885 2444.2727,-205.8653 2453.4964,-211.0782 2454.734,-204.1885"/>
+<path fill="none" stroke="#191970" d="M762.2834,-253.7407C788.464,-250.8255 819.1555,-247.7875 847,-246 1516.036,-203.0501 1685.3108,-241.1477 2355,-210 2405.3955,-207.6561 2462.0853,-203.7207 2505.8255,-200.4019"/>
+<polygon fill="#191970" stroke="#191970" points="2506.0983,-203.8914 2515.8021,-199.6387 2505.5643,-196.9118 2506.0983,-203.8914"/>
</g>
<!-- Node31->Node18 -->
<g id="edge69" class="edge">
<title>Node31->Node18</title>
-<path fill="none" stroke="#191970" d="M2352.0449,-179.4489C2320.3551,-163.5009 2267.6948,-138.5232 2220,-123 2147.6307,-99.4459 2059.4336,-83.6919 2012.7116,-76.318"/>
-<polygon fill="#191970" stroke="#191970" points="2012.9934,-72.82 2002.5755,-74.7477 2011.9217,-79.7375 2012.9934,-72.82"/>
+<path fill="none" stroke="#191970" d="M2556.1927,-179.4178C2520.5732,-155.8629 2452.1099,-110.5888 2416.2315,-86.8628"/>
+<polygon fill="#191970" stroke="#191970" points="2417.7121,-83.6458 2407.4403,-81.0493 2413.8509,-89.4846 2417.7121,-83.6458"/>
</g>
<!-- Node50->Node3 -->
<g id="edge73" class="edge">
<title>Node50->Node3</title>
-<path fill="none" stroke="#191970" d="M1190.5313,-447.3906C1151.0507,-436.6944 1087.1557,-419.3836 1044.1395,-407.7294"/>
-<polygon fill="#191970" stroke="#191970" points="1044.787,-404.2787 1034.2197,-405.0419 1042.9565,-411.0352 1044.787,-404.2787"/>
+<path fill="none" stroke="#191970" d="M769.7513,-449.5817C839.4616,-438.0259 973.5051,-415.8058 1046.1911,-403.7567"/>
+<polygon fill="#191970" stroke="#191970" points="1047.1622,-407.1436 1056.4551,-402.0553 1046.0173,-400.2378 1047.1622,-407.1436"/>
</g>
<!-- Node50->Node16 -->
<g id="edge76" class="edge">
<title>Node50->Node16</title>
-<path fill="none" stroke="#191970" d="M1181.1213,-451.6683C1166.1502,-450.0108 1149.3701,-448.2837 1134,-447 997.3533,-435.5878 649.2047,-450.8483 518,-411 376.2161,-367.9387 312.9792,-345.0499 252,-210 236.0877,-174.7593 230.3547,-155.0405 252,-123 273.2907,-91.4845 317.411,-79.2514 347.8514,-74.5052"/>
-<polygon fill="#191970" stroke="#191970" points="348.5545,-77.9418 357.9882,-73.1194 347.6062,-71.0063 348.5545,-77.9418"/>
+<path fill="none" stroke="#191970" d="M691.3596,-447.405C628.7817,-427.5201 495.9495,-375.2518 442,-277 409.057,-217.005 396.7187,-174.3249 442,-123 466.9384,-94.7331 727.2902,-78.1671 819.6926,-73.1424"/>
+<polygon fill="#191970" stroke="#191970" points="820.0543,-76.6281 829.8534,-72.5998 819.681,-69.6381 820.0543,-76.6281"/>
</g>
<!-- Node50->Node21 -->
<g id="edge74" class="edge">
<title>Node50->Node21</title>
-<path fill="none" stroke="#191970" d="M1240.3886,-447.3176C1266.0428,-429.459 1320.1588,-389.0207 1354,-344 1367.3685,-326.2151 1377.8064,-303.1153 1384.5058,-285.8842"/>
-<polygon fill="#191970" stroke="#191970" points="1387.79,-287.094 1388.0122,-276.5015 1381.233,-284.6435 1387.79,-287.094"/>
+<path fill="none" stroke="#191970" d="M769.7132,-447.8545C803.1178,-440.1998 849.462,-427.8173 888,-411 971.275,-374.6602 1060.2188,-314.2409 1104.5069,-282.4815"/>
+<polygon fill="#191970" stroke="#191970" points="1106.695,-285.2186 1112.7541,-276.5273 1102.5976,-279.5431 1106.695,-285.2186"/>
</g>
<!-- Node50->Node33 -->
<g id="edge75" class="edge">
<title>Node50->Node33</title>
-<path fill="none" stroke="#191970" d="M1227.5402,-447.2068C1231.2237,-423.9129 1240.9134,-363.3838 1250,-313 1255.8059,-280.807 1263.0826,-243.8164 1267.878,-219.8476"/>
-<polygon fill="#191970" stroke="#191970" points="1271.3552,-220.3086 1269.8925,-209.8152 1264.4922,-218.9304 1271.3552,-220.3086"/>
+<path fill="none" stroke="#191970" d="M726.0207,-447.1979C730.0791,-413.4604 748.2062,-301.189 814,-246 834.0647,-229.1693 895.4832,-215.097 945.9038,-205.9393"/>
+<polygon fill="#191970" stroke="#191970" points="946.5935,-209.3716 955.8257,-204.1739 945.3672,-202.4799 946.5935,-209.3716"/>
</g>
<!-- Node52->Node18 -->
<g id="edge86" class="edge">
<title>Node52->Node18</title>
-<path fill="none" stroke="#191970" d="M3013.2218,-385.8928C3001.3534,-376.7 2985.0543,-361.6481 2978,-344 2972.8862,-331.2064 2973.1036,-325.8784 2978,-313 2985.6077,-292.9902 3002.3923,-297.0098 3010,-277 3025.4788,-236.2877 3029.6435,-217.8744 3010,-179 2991.0774,-141.5523 2973.4922,-137.1692 2934,-123 2845.8648,-91.3785 2170.9457,-75.4721 2012.9933,-72.1604"/>
-<polygon fill="#191970" stroke="#191970" points="2013.0081,-68.66 2002.9377,-71.952 2012.863,-75.6585 2013.0081,-68.66"/>
+<path fill="none" stroke="#191970" d="M2206.8572,-385.9853C2220.4439,-376.2947 2241.3397,-360.4001 2257,-344 2307.9552,-290.6378 2322.9324,-276.4501 2355,-210 2374.2394,-170.1324 2385.0697,-118.9493 2389.9012,-91.3203"/>
+<polygon fill="#191970" stroke="#191970" points="2393.3741,-91.7717 2391.5629,-81.3328 2386.4691,-90.6228 2393.3741,-91.7717"/>
</g>
<!-- Node53 -->
<g id="node32" class="node">
<title>Node53</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2986.5,-319 2986.5,-338 3081.5,-338 3081.5,-319 2986.5,-319"/>
-<text text-anchor="middle" x="3034" y="-326" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">dmlc/common.h</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2152.5,-319 2152.5,-338 2247.5,-338 2247.5,-319 2152.5,-319"/>
+<text text-anchor="middle" x="2200" y="-326" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">dmlc/common.h</text>
</g>
<!-- Node52->Node53 -->
<g id="edge85" class="edge">
<title>Node52->Node53</title>
-<path fill="none" stroke="#191970" d="M3028.0029,-385.9005C3029.0217,-376.149 3030.6296,-360.7597 3031.9241,-348.3695"/>
-<polygon fill="#191970" stroke="#191970" points="3035.4408,-348.3912 3032.9989,-338.0817 3028.4787,-347.6638 3035.4408,-348.3912"/>
+<path fill="none" stroke="#191970" d="M2194.0029,-385.9005C2195.0217,-376.149 2196.6296,-360.7597 2197.9241,-348.3695"/>
+<polygon fill="#191970" stroke="#191970" points="2201.4408,-348.3912 2198.9989,-338.0817 2194.4787,-347.6638 2201.4408,-348.3912"/>
</g>
<!-- Node54->Node11 -->
<g id="edge92" class="edge">
<title>Node54->Node11</title>
-<path fill="none" stroke="#191970" d="M2903.381,-380.467C2910.409,-362.5636 2924.1958,-332.621 2944,-313 2967.7363,-289.4833 2992.9437,-305.732 3010,-277 3032.2336,-239.5466 3035.8764,-214.0356 3010,-179 2976.5079,-133.653 2823.0417,-100.1487 2728.874,-83.4323"/>
-<polygon fill="#191970" stroke="#191970" points="2729.2792,-79.9501 2718.8252,-81.6721 2728.0714,-86.8451 2729.2792,-79.9501"/>
+<path fill="none" stroke="#191970" d="M2645.8512,-382.3257C2756.154,-356.6608 2995.5149,-300.43 3077,-277 3167.4988,-250.9783 3222.9404,-286.3911 3278,-210 3302.274,-176.3217 3276.4883,-148.845 3244,-123 3220.5402,-104.3373 3190.0471,-92.318 3162.2508,-84.6329"/>
+<polygon fill="#191970" stroke="#191970" points="3163.1079,-81.2394 3152.5474,-82.0915 3161.3343,-88.011 3163.1079,-81.2394"/>
</g>
<!-- Node54->Node16 -->
<g id="edge133" class="edge">
<title>Node54->Node16</title>
-<path fill="none" stroke="#191970" d="M2841.3102,-394.9229C2526.1941,-391.6075 996.9183,-374.0709 789,-344 626.2008,-320.4546 556.1063,-330.375 444,-210 411.1848,-174.7644 392.7031,-119.737 384.7446,-90.7904"/>
-<polygon fill="#191970" stroke="#191970" points="388.1185,-89.8578 382.2018,-81.069 381.3464,-91.6293 388.1185,-89.8578"/>
+<path fill="none" stroke="#191970" d="M2532.3131,-392.8083C2466.222,-389.6805 2353.6951,-384.3875 2257,-380 1724.0253,-355.8162 1572.7175,-427.9706 1061,-277 1051.8062,-274.2876 908.5983,-216.9531 902,-210 870.1532,-176.4407 858.3458,-121.1129 854.1603,-91.5664"/>
+<polygon fill="#191970" stroke="#191970" points="857.5866,-90.7599 852.8674,-81.2742 850.6412,-91.6324 857.5866,-90.7599"/>
</g>
<!-- Node30 -->
<g id="node34" class="node">
<title>Node30</title>
<g id="a_node34"><a xlink:href="ndarray_8h.html" target="_top" xlink:title="A device-independent managed NDArray abstraction. ">
-<polygon fill="#ffffff" stroke="#000000" points="2076.5,-252 2076.5,-271 2201.5,-271 2201.5,-252 2076.5,-252"/>
-<text text-anchor="middle" x="2139" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/ndarray.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2771.5,-252 2771.5,-271 2896.5,-271 2896.5,-252 2771.5,-252"/>
+<text text-anchor="middle" x="2834" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/ndarray.h</text>
</a>
</g>
</g>
<!-- Node54->Node30 -->
<g id="edge93" class="edge">
<title>Node54->Node30</title>
-<path fill="none" stroke="#191970" d="M2841.1816,-382.2179C2836.4015,-381.3682 2831.6281,-380.6104 2827,-380 2785.0934,-374.4728 2096.1734,-374.588 2067,-344 2045.0974,-321.0354 2082.3749,-292.9403 2110.965,-276.1707"/>
-<polygon fill="#191970" stroke="#191970" points="2112.873,-279.1127 2119.8562,-271.1448 2109.4284,-273.0188 2112.873,-279.1127"/>
+<path fill="none" stroke="#191970" d="M2616.6284,-380.389C2664.1621,-354.3909 2760.8008,-301.5355 2807.526,-275.9796"/>
+<polygon fill="#191970" stroke="#191970" points="2809.3462,-278.9735 2816.4401,-271.1042 2805.9871,-272.832 2809.3462,-278.9735"/>
</g>
<!-- Node41 -->
<g id="node38" class="node">
<title>Node41</title>
<g id="a_node38"><a xlink:href="packed__func_8h.html" target="_top" xlink:title="Type-erased function used across TVM API. ">
-<polygon fill="#ffffff" stroke="#000000" points="2076,-313.5 2076,-343.5 2192,-343.5 2192,-313.5 2076,-313.5"/>
-<text text-anchor="start" x="2084" y="-331.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/packed</text>
-<text text-anchor="middle" x="2134" y="-320.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_func.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1449,-313.5 1449,-343.5 1565,-343.5 1565,-313.5 1449,-313.5"/>
+<text text-anchor="start" x="1457" y="-331.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/packed</text>
+<text text-anchor="middle" x="1507" y="-320.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">_func.h</text>
</a>
</g>
</g>
<!-- Node54->Node41 -->
<g id="edge109" class="edge">
<title>Node54->Node41</title>
-<path fill="none" stroke="#191970" d="M2841.1627,-382.3542C2836.3868,-381.4743 2831.6195,-380.6725 2827,-380 2599.6923,-346.9101 2327.1678,-334.5536 2202.4207,-330.3936"/>
-<polygon fill="#191970" stroke="#191970" points="2202.3634,-326.89 2192.2545,-330.0619 2202.1351,-333.8862 2202.3634,-326.89"/>
+<path fill="none" stroke="#191970" d="M2532.4131,-392.8022C2356.7618,-384.3644 1822.2781,-358.1934 1649,-344 1624.9937,-342.0336 1598.7631,-339.3282 1575.4517,-336.7275"/>
+<polygon fill="#191970" stroke="#191970" points="1575.6407,-333.2267 1565.3111,-335.5826 1574.8553,-340.1825 1575.6407,-333.2267"/>
</g>
<!-- Node30->Node11 -->
<g id="edge94" class="edge">
<title>Node30->Node11</title>
-<path fill="none" stroke="#191970" d="M2201.7479,-252.2303C2217.4842,-250.0455 2234.3451,-247.824 2250,-246 2292.6757,-241.0278 2603.972,-239.7187 2635,-210 2665.2603,-181.0165 2663.3473,-128.135 2658.8933,-96.7706"/>
-<polygon fill="#191970" stroke="#191970" points="2662.3084,-95.9634 2657.2604,-86.6484 2655.3977,-97.0782 2662.3084,-95.9634"/>
+<path fill="none" stroke="#191970" d="M2896.7331,-252.7513C2949.9719,-244.1363 3020.9709,-229.4505 3042,-210 3074.1123,-180.2982 3083.833,-127.6672 3086.7588,-96.5421"/>
+<polygon fill="#191970" stroke="#191970" points="3090.2536,-96.7428 3087.5358,-86.5025 3083.2745,-96.2026 3090.2536,-96.7428"/>
</g>
<!-- Node30->Node10 -->
<g id="edge101" class="edge">
<title>Node30->Node10</title>
-<path fill="none" stroke="#191970" d="M2080.9602,-251.9602C2021.3927,-242.0304 1926.0752,-225.7553 1844,-210 1777.9998,-197.3305 1762.3745,-189.5343 1696,-179 1572.3862,-159.3812 1427.1074,-145.5817 1342.9896,-138.4925"/>
-<polygon fill="#191970" stroke="#191970" points="1343.046,-134.9851 1332.7894,-137.6405 1342.4633,-141.9608 1343.046,-134.9851"/>
+<path fill="none" stroke="#191970" d="M2826.271,-251.826C2811.0636,-233.6276 2775.2173,-194.719 2735,-179 2696.0187,-163.7641 2100.0223,-142.3177 1893.6791,-135.3119"/>
+<polygon fill="#191970" stroke="#191970" points="1893.6792,-131.81 1883.5665,-134.9696 1893.4423,-138.806 1893.6792,-131.81"/>
</g>
<!-- Node30->Node18 -->
<g id="edge107" class="edge">
<title>Node30->Node18</title>
-<path fill="none" stroke="#191970" d="M2150.7225,-251.8841C2161.1935,-242.4695 2175.8061,-227.0823 2182,-210 2195.1805,-173.6491 2205.725,-153.5326 2182,-123 2161.381,-96.4646 2064.379,-81.3269 2012.6629,-75.0144"/>
-<polygon fill="#191970" stroke="#191970" points="2012.9999,-71.5299 2002.6588,-73.8345 2012.18,-78.4818 2012.9999,-71.5299"/>
+<path fill="none" stroke="#191970" d="M2896.5544,-252.3052C2952.9126,-241.264 3022.699,-219.0198 2993,-179 2941.8051,-110.0142 2892.0426,-140.7988 2808,-123 2667.4797,-93.2401 2496.2916,-78.7197 2425.7649,-73.6583"/>
+<polygon fill="#191970" stroke="#191970" points="2425.9376,-70.1619 2415.717,-72.9528 2425.4473,-77.1447 2425.9376,-70.1619"/>
</g>
<!-- Node30->Node19 -->
<g id="edge105" class="edge">
<title>Node30->Node19</title>
-<path fill="none" stroke="#191970" d="M2138.6584,-251.9782C2137.5013,-234.7264 2132.4941,-198.3029 2111,-179 2044.354,-119.1482 1764.9864,-85.2912 1665.4272,-74.9432"/>
-<polygon fill="#191970" stroke="#191970" points="1665.5835,-71.441 1655.2793,-73.9055 1664.8714,-78.4047 1665.5835,-71.441"/>
+<path fill="none" stroke="#191970" d="M2896.8946,-252.9681C2946.6136,-244.7826 3010.1932,-230.6054 3026,-210 3034.386,-199.0683 3030.0466,-192.1701 3026,-179 3017.1659,-150.2485 3013.1775,-139.4555 2988,-123 2948.3979,-97.1169 2810.8073,-81.0127 2745.2542,-74.6364"/>
+<polygon fill="#191970" stroke="#191970" points="2745.2917,-71.1243 2735.0053,-73.6619 2744.6291,-78.0929 2745.2917,-71.1243"/>
</g>
<!-- Node30->Node20 -->
<g id="edge108" class="edge">
<title>Node30->Node20</title>
-<path fill="none" stroke="#191970" d="M2108.6481,-251.9352C2055.9788,-235.554 1944.0764,-201.7034 1848,-179 1774.0954,-161.5359 1686.2833,-146.1673 1639.4147,-138.3896"/>
-<polygon fill="#191970" stroke="#191970" points="1639.9429,-134.9296 1629.5069,-136.7576 1638.8051,-141.8365 1639.9429,-134.9296"/>
+<path fill="none" stroke="#191970" d="M2833.938,-251.9332C2833.2512,-234.2698 2829.0024,-196.7271 2806,-179 2756.1631,-140.5927 2308.1307,-134.233 2180.0542,-133.1977"/>
+<polygon fill="#191970" stroke="#191970" points="2179.7147,-129.6951 2169.6887,-133.1201 2179.6622,-136.6949 2179.7147,-129.6951"/>
</g>
<!-- Node30->Node28 -->
<g id="edge100" class="edge">
<title>Node30->Node28</title>
-<path fill="none" stroke="#191970" d="M2076.3266,-256.9722C2030.3127,-253.7175 1966.2862,-249.3336 1910,-246 1595.0349,-227.3457 1509.7082,-275.194 1201,-210 1163.2854,-202.0353 1156.3989,-191.6864 1120,-179 1085.3175,-166.9119 1045.5305,-154.4688 1016.2588,-145.5771"/>
-<polygon fill="#191970" stroke="#191970" points="1016.9543,-142.131 1006.3692,-142.5866 1014.9281,-148.8313 1016.9543,-142.131"/>
+<path fill="none" stroke="#191970" d="M2771.1797,-259.2557C2587.7478,-252.5192 2054.969,-231.6015 1882,-210 1757.5839,-194.4621 1613.3729,-162.3352 1539.7434,-144.8932"/>
+<polygon fill="#191970" stroke="#191970" points="1540.3706,-141.4448 1529.8318,-142.5328 1538.7489,-148.2543 1540.3706,-141.4448"/>
</g>
<!-- Node30->Node33 -->
<g id="edge99" class="edge">
<title>Node30->Node33</title>
-<path fill="none" stroke="#191970" d="M2076.3102,-257.2559C1926.0265,-247.0398 1548.3407,-221.1059 1422,-210 1397.2905,-207.8279 1370.3113,-205.1229 1346.1832,-202.5835"/>
-<polygon fill="#191970" stroke="#191970" points="1346.3454,-199.0812 1336.0321,-201.5074 1345.6074,-206.0422 1346.3454,-199.0812"/>
+<path fill="none" stroke="#191970" d="M2771.4729,-260.0089C2561.1563,-254.869 1860.0439,-236.7247 1281,-210 1217.3115,-207.0606 1144.9739,-202.7039 1092.4681,-199.3574"/>
+<polygon fill="#191970" stroke="#191970" points="1092.4431,-195.8487 1082.2398,-198.7022 1091.9956,-202.8344 1092.4431,-195.8487"/>
</g>
<!-- Node30->Node31 -->
<g id="edge95" class="edge">
<title>Node30->Node31</title>
-<path fill="none" stroke="#191970" d="M2173.4157,-251.9717C2210.6166,-241.6723 2270.7473,-225.0245 2316.7398,-212.291"/>
-<polygon fill="#191970" stroke="#191970" points="2317.9174,-215.5968 2326.621,-209.5553 2316.0496,-208.8505 2317.9174,-215.5968"/>
+<path fill="none" stroke="#191970" d="M2797.7355,-251.9717C2758.3718,-241.6291 2694.6436,-224.8848 2646.1033,-212.1311"/>
+<polygon fill="#191970" stroke="#191970" points="2646.8614,-208.7115 2636.3002,-209.5553 2645.0825,-215.4817 2646.8614,-208.7115"/>
</g>
<!-- Node32 -->
<g id="node35" class="node">
<title>Node32</title>
<g id="a_node35"><a xlink:href="shape__tuple_8h.html" target="_top" xlink:title="Runtime ShapeTuple container types. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1976,-179.5 1976,-209.5 2102,-209.5 2102,-179.5 1976,-179.5"/>
-<text text-anchor="start" x="1984" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
-<text text-anchor="middle" x="2039" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/shape_tuple.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="2858,-179.5 2858,-209.5 2984,-209.5 2984,-179.5 2858,-179.5"/>
+<text text-anchor="start" x="2866" y="-197.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/container</text>
+<text text-anchor="middle" x="2921" y="-186.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">/shape_tuple.h</text>
</a>
</g>
</g>
<!-- Node30->Node32 -->
<g id="edge96" class="edge">
<title>Node30->Node32</title>
-<path fill="none" stroke="#191970" d="M2124.6724,-251.9005C2110.4706,-242.3853 2088.2572,-227.5023 2070.0082,-215.2755"/>
-<polygon fill="#191970" stroke="#191970" points="2071.6764,-212.1803 2061.4205,-209.5218 2067.7801,-217.9957 2071.6764,-212.1803"/>
+<path fill="none" stroke="#191970" d="M2846.465,-251.9005C2858.5959,-242.5583 2877.4458,-228.0417 2893.1535,-215.945"/>
+<polygon fill="#191970" stroke="#191970" points="2895.7068,-218.3963 2901.4941,-209.5218 2891.4357,-212.8503 2895.7068,-218.3963"/>
</g>
<!-- Node37 -->
<g id="node36" class="node">
<title>Node37</title>
<g id="a_node36"><a xlink:href="serializer_8h.html" target="_top" xlink:title="Serializer extension to support TVM data types Include this file to enable serialization of DLDataTyp...">
-<polygon fill="#ffffff" stroke="#ff0000" points="2682,-185 2682,-204 2816,-204 2816,-185 2682,-185"/>
-<text text-anchor="middle" x="2749" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/serializer.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="3116,-185 3116,-204 3250,-204 3250,-185 3116,-185"/>
+<text text-anchor="middle" x="3183" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/serializer.h</text>
</a>
</g>
</g>
<!-- Node30->Node37 -->
<g id="edge102" class="edge">
<title>Node30->Node37</title>
-<path fill="none" stroke="#191970" d="M2194.1709,-251.9882C2209.3073,-249.8587 2225.7252,-247.723 2241,-246 2426.2904,-225.0988 2474.2942,-235.5569 2659,-210 2667.4497,-208.8309 2676.3535,-207.3664 2685.156,-205.8074"/>
-<polygon fill="#191970" stroke="#191970" points="2685.8068,-209.2465 2695.0215,-204.0177 2684.5573,-202.3589 2685.8068,-209.2465"/>
+<path fill="none" stroke="#191970" d="M2878.7458,-251.9717C2939.6063,-239.8335 3048.6503,-218.8781 3118.3741,-205.9039"/>
+<polygon fill="#191970" stroke="#191970" points="3119.1721,-209.3157 3128.3655,-204.0496 3117.8947,-202.4332 3119.1721,-209.3157"/>
</g>
<!-- Node40 -->
<g id="node37" class="node">
<title>Node40</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2234.5,-185 2234.5,-204 2299.5,-204 2299.5,-185 2234.5,-185"/>
-<text text-anchor="middle" x="2267" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">functional</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="2660.5,-185 2660.5,-204 2725.5,-204 2725.5,-185 2660.5,-185"/>
+<text text-anchor="middle" x="2693" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">functional</text>
</g>
<!-- Node30->Node40 -->
<g id="edge106" class="edge">
<title>Node30->Node40</title>
-<path fill="none" stroke="#191970" d="M2157.3393,-251.9005C2178.9156,-240.6067 2214.9391,-221.7506 2239.7766,-208.7497"/>
-<polygon fill="#191970" stroke="#191970" points="2241.4582,-211.8201 2248.6947,-204.0817 2238.2119,-205.6183 2241.4582,-211.8201"/>
+<path fill="none" stroke="#191970" d="M2813.7981,-251.9005C2789.821,-240.5072 2749.6479,-221.4178 2722.2679,-208.4075"/>
+<polygon fill="#191970" stroke="#191970" points="2723.6987,-205.2123 2713.1644,-204.0817 2720.6944,-211.5348 2723.6987,-205.2123"/>
</g>
<!-- Node32->Node18 -->
<g id="edge97" class="edge">
<title>Node32->Node18</title>
-<path fill="none" stroke="#191970" d="M2031.7654,-179.4178C2020.9956,-156.9655 2000.7602,-114.7796 1989.0331,-90.3317"/>
-<polygon fill="#191970" stroke="#191970" points="1992.0613,-88.5519 1984.5805,-81.0493 1985.7498,-91.5794 1992.0613,-88.5519"/>
+<path fill="none" stroke="#191970" d="M2895.143,-179.4571C2865.4599,-162.9018 2814.8233,-136.8107 2768,-123 2645.3026,-86.8098 2492.3745,-75.8799 2426.1358,-72.72"/>
+<polygon fill="#191970" stroke="#191970" points="2426.0461,-69.2125 2415.8994,-72.2614 2425.7328,-76.2055 2426.0461,-69.2125"/>
</g>
<!-- Node32->Node20 -->
<g id="edge98" class="edge">
<title>Node32->Node20</title>
-<path fill="none" stroke="#191970" d="M1975.5744,-185.4915C1882.4356,-172.2628 1711.9127,-148.043 1639.7173,-137.7889"/>
-<polygon fill="#191970" stroke="#191970" points="1640.1927,-134.3214 1629.7999,-136.3804 1639.2083,-141.2519 1640.1927,-134.3214"/>
+<path fill="none" stroke="#191970" d="M2857.6629,-181.0568C2853.0508,-180.2912 2848.4607,-179.5933 2844,-179 2590.2098,-145.2455 2281.0559,-135.8896 2179.8355,-133.6385"/>
+<polygon fill="#191970" stroke="#191970" points="2179.6845,-130.1346 2169.6119,-133.4199 2179.5348,-137.133 2179.6845,-130.1346"/>
</g>
<!-- Node37->Node11 -->
<g id="edge103" class="edge">
<title>Node37->Node11</title>
-<path fill="none" stroke="#191970" d="M2741.5455,-184.8484C2726.705,-165.6339 2693.0902,-122.1115 2672.0997,-94.9344"/>
-<polygon fill="#191970" stroke="#191970" points="2674.6507,-92.5113 2665.768,-86.7365 2669.1107,-96.7902 2674.6507,-92.5113"/>
+<path fill="none" stroke="#191970" d="M3175.5455,-184.8484C3160.705,-165.6339 3127.0902,-122.1115 3106.0997,-94.9344"/>
+<polygon fill="#191970" stroke="#191970" points="3108.6507,-92.5113 3099.768,-86.7365 3103.1107,-96.7902 3108.6507,-92.5113"/>
</g>
<!-- Node37->Node30 -->
<g id="edge104" class="edge">
<title>Node37->Node30</title>
-<path fill="none" stroke="#191970" d="M2712.4299,-204.0177C2701.1197,-206.2223 2688.6297,-208.3909 2677,-210 2492.2942,-235.5569 2444.2904,-225.0988 2259,-246 2243.7735,-247.7176 2227.4112,-249.8452 2211.7906,-251.968"/>
-<polygon fill="#191970" stroke="#191970" points="2211.0841,-248.5321 2201.6522,-253.358 2212.035,-255.4672 2211.0841,-248.5321"/>
+<path fill="none" stroke="#191970" d="M3138.1473,-204.0496C3077.2305,-216.1981 2968.1814,-237.154 2898.5036,-250.1188"/>
+<polygon fill="#191970" stroke="#191970" points="2897.7128,-246.7057 2888.5193,-251.9717 2898.9901,-253.5882 2897.7128,-246.7057"/>
</g>
<!-- Node41->Node11 -->
<g id="edge110" class="edge">
<title>Node41->Node11</title>
-<path fill="none" stroke="#191970" d="M2192.1239,-327.4824C2349.2695,-324.272 2774.257,-312.3804 2826,-277 2863.8223,-251.1382 2881.6712,-218.2305 2858,-179 2830.0023,-132.5991 2774.5716,-105.2117 2728.5495,-89.7042"/>
-<polygon fill="#191970" stroke="#191970" points="2729.4769,-86.3256 2718.8849,-86.5728 2727.3192,-92.9848 2729.4769,-86.3256"/>
+<path fill="none" stroke="#191970" d="M1565.2852,-321.0735C1590.9205,-318.0825 1621.4141,-314.8975 1649,-313 2206.5782,-274.6462 2350.3667,-337.2997 2906,-277 3064.7577,-259.7709 3157.153,-332.996 3259,-210 3300.8902,-159.4111 3209.391,-114.305 3144.7723,-90.134"/>
+<polygon fill="#191970" stroke="#191970" points="3145.6487,-86.7278 3135.0549,-86.5791 3143.2438,-93.3017 3145.6487,-86.7278"/>
</g>
<!-- Node41->Node10 -->
<g id="edge124" class="edge">
<title>Node41->Node10</title>
-<path fill="none" stroke="#191970" d="M2088.7645,-313.4455C2059.3468,-303.4866 2020.232,-289.9273 1986,-277 1951.8727,-264.1123 1945.0893,-255.9754 1910,-246 1817.2285,-219.6264 1790.7616,-227.9546 1696,-210 1630.3969,-197.5701 1614.6125,-191.3802 1549,-179 1479.0086,-165.7936 1398.6526,-152.6251 1342.7648,-143.7835"/>
-<polygon fill="#191970" stroke="#191970" points="1343.0453,-140.2845 1332.6221,-142.1838 1341.9547,-147.199 1343.0453,-140.2845"/>
+<path fill="none" stroke="#191970" d="M1524.6282,-313.3389C1564.1692,-279.5363 1659.337,-199.3692 1696,-179 1721.8765,-164.6235 1753.1036,-153.2966 1778.2362,-145.4798"/>
+<polygon fill="#191970" stroke="#191970" points="1779.34,-148.8026 1787.8958,-142.5536 1777.3105,-142.1032 1779.34,-148.8026"/>
</g>
<!-- Node41->Node15 -->
<g id="edge114" class="edge">
<title>Node41->Node15</title>
-<path fill="none" stroke="#191970" d="M2114.6009,-313.4864C2085.6654,-291.1635 2033.5652,-251.2706 2024,-246 1950.6935,-205.607 1927.6379,-204.7535 1848,-179 1756.4999,-149.4105 1733.3731,-141.4803 1639,-123 1521.9109,-100.0715 1383.9139,-85.1984 1301.5936,-77.5914"/>
-<polygon fill="#191970" stroke="#191970" points="1301.8662,-74.1018 1291.5892,-76.6774 1301.2293,-81.0728 1301.8662,-74.1018"/>
+<path fill="none" stroke="#191970" d="M1512.611,-313.2537C1524.0954,-281.7361 1550.7358,-206.9988 1569,-143 1573.9666,-125.5966 1578.6485,-105.5283 1581.8364,-91.091"/>
+<polygon fill="#191970" stroke="#191970" points="1585.2638,-91.8005 1583.963,-81.2858 1578.4229,-90.3167 1585.2638,-91.8005"/>
</g>
<!-- Node41->Node16 -->
<g id="edge128" class="edge">
<title>Node41->Node16</title>
-<path fill="none" stroke="#191970" d="M2075.7345,-327.2326C1839.9005,-321.9593 962.9411,-300.9121 841,-277 661.9215,-241.8835 469.1193,-127.9034 403.4351,-86.6118"/>
-<polygon fill="#191970" stroke="#191970" points="405.0033,-83.462 394.6832,-81.0651 401.2561,-89.3746 405.0033,-83.462"/>
+<path fill="none" stroke="#191970" d="M1468.4721,-313.383C1350.0033,-266.8998 992.5548,-126.649 883.7618,-83.9623"/>
+<polygon fill="#191970" stroke="#191970" points="884.9021,-80.65 874.3146,-80.2555 882.3453,-87.1663 884.9021,-80.65"/>
</g>
<!-- Node41->Node17 -->
<g id="edge130" class="edge">
<title>Node41->Node17</title>
-<path fill="none" stroke="#191970" d="M2075.8807,-321.9032C2008.2142,-313.7024 1893.2983,-298.1605 1796,-277 1746.7849,-266.2967 1736.6931,-254.2027 1687,-246 1529.377,-219.9815 1119.0476,-264.8424 969,-210 922.8304,-193.125 904.0466,-186.4195 881,-143 872.6392,-127.2484 872.3918,-106.7015 873.7907,-91.6779"/>
-<polygon fill="#191970" stroke="#191970" points="877.3034,-91.7946 875.0742,-81.4369 870.3577,-90.924 877.3034,-91.7946"/>
+<path fill="none" stroke="#191970" d="M1497.8056,-313.1269C1479.4968,-281.9783 1438.0891,-208.78 1413,-143 1406.5075,-125.9777 1401.3767,-105.8681 1398.0918,-91.3183"/>
+<polygon fill="#191970" stroke="#191970" points="1401.4853,-90.4514 1395.9476,-81.419 1394.6439,-91.9333 1401.4853,-90.4514"/>
</g>
<!-- Node41->Node18 -->
<g id="edge131" class="edge">
<title>Node41->Node18</title>
-<path fill="none" stroke="#191970" d="M2192.2532,-313.7017C2226.6389,-304.3886 2270.7765,-291.4244 2309,-277 2375.0424,-252.0777 2415.0477,-269.5176 2453,-210 2460.4077,-198.3831 2461.2895,-190.0051 2453,-179 2399.4026,-107.8441 2111.3343,-80.8775 2012.955,-73.6603"/>
-<polygon fill="#191970" stroke="#191970" points="2012.9366,-70.1502 2002.7131,-72.9306 2012.439,-77.1325 2012.9366,-70.1502"/>
+<path fill="none" stroke="#191970" d="M1565.0146,-321.9796C1590.7346,-319.1374 1621.3682,-315.8183 1649,-313 1769.5826,-300.7013 2653.4394,-299.6621 2735,-210 2827.0063,-108.8546 2527.836,-80.0917 2425.8344,-73.3242"/>
+<polygon fill="#191970" stroke="#191970" points="2425.9734,-69.8261 2415.7719,-72.6861 2425.5304,-76.812 2425.9734,-69.8261"/>
</g>
<!-- Node41->Node20 -->
<g id="edge132" class="edge">
<title>Node41->Node20</title>
-<path fill="none" stroke="#191970" d="M2102.9453,-313.4878C2082.987,-303.6005 2056.65,-290.1045 2034,-277 2012.0183,-264.2822 2009.1489,-256.4435 1986,-246 1932.1321,-221.6979 1911.8679,-234.3021 1858,-210 1834.8511,-199.5565 1833.5064,-188.6119 1810,-179 1739.5155,-150.1783 1713.7087,-162.8287 1639.4612,-144.0286"/>
-<polygon fill="#191970" stroke="#191970" points="1640.2239,-140.61 1629.6615,-141.438 1638.4349,-147.3775 1640.2239,-140.61"/>
+<path fill="none" stroke="#191970" d="M1565.307,-322.8958C1642.4683,-314.7861 1773.784,-298.3786 1817,-277 1835.0368,-268.0774 1832.2982,-255.5699 1850,-246 1909.9304,-213.6005 1935.2345,-233.9863 1999,-210 2044.3039,-192.9583 2093.4572,-165.1207 2121.8947,-147.9973"/>
+<polygon fill="#191970" stroke="#191970" points="2123.9869,-150.8213 2130.7091,-142.6321 2120.3472,-144.8419 2123.9869,-150.8213"/>
</g>
<!-- Node41->Node21 -->
<g id="edge111" class="edge">
<title>Node41->Node21</title>
-<path fill="none" stroke="#191970" d="M2075.855,-323.2426C1942.2486,-311.1622 1613.5553,-281.4422 1466.3567,-268.1328"/>
-<polygon fill="#191970" stroke="#191970" points="1466.3519,-264.6182 1456.0773,-267.2033 1465.7214,-271.5897 1466.3519,-264.6182"/>
+<path fill="none" stroke="#191970" d="M1448.9729,-318.1048C1383.6566,-306.4037 1277.4106,-287.3703 1206.3829,-274.6461"/>
+<polygon fill="#191970" stroke="#191970" points="1206.6285,-271.1345 1196.168,-272.8162 1205.3941,-278.0248 1206.6285,-271.1345"/>
</g>
<!-- Node41->Node23 -->
<g id="edge127" class="edge">
<title>Node41->Node23</title>
-<path fill="none" stroke="#191970" d="M2192.0635,-315.4752C2196.7678,-314.5802 2201.4569,-313.741 2206,-313 2348.1245,-289.8175 2385.9912,-300.881 2528,-277 2536.3402,-275.5975 2819.7186,-216.6055 2825,-210 2833.6039,-199.239 2833.726,-189.6623 2825,-179 2801.5976,-150.4046 2695.1322,-139.1852 2636.3257,-135.1121"/>
-<polygon fill="#191970" stroke="#191970" points="2636.4845,-131.6151 2626.277,-134.4536 2636.0266,-138.6002 2636.4845,-131.6151"/>
+<path fill="none" stroke="#191970" d="M1453.2762,-313.4215C1373.2136,-291.0119 1229.357,-251.0048 1205,-246 1148.2959,-234.3486 992.0006,-246.415 947,-210 929.7748,-196.0612 923.145,-170.5746 920.5941,-152.7552"/>
+<polygon fill="#191970" stroke="#191970" points="924.0466,-152.1351 919.4553,-142.5868 917.0901,-152.9143 924.0466,-152.1351"/>
</g>
<!-- Node41->Node28 -->
<g id="edge113" class="edge">
<title>Node41->Node28</title>
-<path fill="none" stroke="#191970" d="M2075.8345,-315.7889C2030.5308,-305.7287 1966.1217,-291.0762 1910,-277 1859.071,-264.2262 1847.8413,-254.3319 1796,-246 1647.471,-222.1285 1258.7791,-268.0609 1120,-210 1101.1352,-202.1075 1103.2768,-189.9401 1086,-179 1063.9769,-165.0544 1037.0094,-153.8256 1015.0553,-145.9539"/>
-<polygon fill="#191970" stroke="#191970" points="1015.9517,-142.5599 1005.3569,-142.5757 1013.6491,-149.1704 1015.9517,-142.5599"/>
+<path fill="none" stroke="#191970" d="M1505.7711,-313.4841C1502.9205,-278.654 1495.8571,-192.3473 1492.6414,-153.0563"/>
+<polygon fill="#191970" stroke="#191970" points="1496.1135,-152.571 1491.8094,-142.8898 1489.1368,-153.142 1496.1135,-152.571"/>
</g>
<!-- Node41->Node44 -->
<g id="edge126" class="edge">
<title>Node41->Node44</title>
-<path fill="none" stroke="#191970" d="M2075.8657,-320.3515C1974.3987,-306.1291 1768.1913,-277.2256 1688.4743,-266.0518"/>
-<polygon fill="#191970" stroke="#191970" points="1688.4785,-262.5183 1678.0895,-264.5962 1687.5068,-269.4505 1688.4785,-262.5183"/>
+<path fill="none" stroke="#191970" d="M1448.9526,-322.3169C1319.5467,-308.5328 1010.9033,-275.6566 910.3048,-264.9411"/>
+<polygon fill="#191970" stroke="#191970" points="910.4701,-261.4389 900.1556,-263.86 909.7286,-268.3996 910.4701,-261.4389"/>
</g>
<!-- Node41->Node42 -->
<g id="edge112" class="edge">
<title>Node41->Node42</title>
-<path fill="none" stroke="#191970" d="M2192.0363,-315.2985C2196.7467,-314.4431 2201.4447,-313.6613 2206,-313 2413.766,-282.8385 2469.0366,-305.769 2677,-277 2678.1144,-276.8458 2679.2369,-276.6855 2680.3659,-276.5196"/>
-<polygon fill="#191970" stroke="#191970" points="2681.3263,-279.9112 2690.6557,-274.89 2680.2313,-272.9974 2681.3263,-279.9112"/>
+<path fill="none" stroke="#191970" d="M1448.6863,-324.4707C1333.5765,-316.4018 1069.0426,-297.2867 847,-277 822.6872,-274.7787 796.1569,-272.0852 772.3592,-269.5709"/>
+<polygon fill="#191970" stroke="#191970" points="772.6559,-266.0828 762.3419,-268.5062 771.916,-273.0436 772.6559,-266.0828"/>
</g>
<!-- Node41->Node30 -->
<g id="edge123" class="edge">
<title>Node41->Node30</title>
-<path fill="none" stroke="#191970" d="M2135.1346,-313.2967C2135.8455,-303.7699 2136.769,-291.3954 2137.535,-281.1306"/>
-<polygon fill="#191970" stroke="#191970" points="2141.0329,-281.2877 2138.2869,-271.055 2134.0524,-280.7667 2141.0329,-281.2877"/>
+<path fill="none" stroke="#191970" d="M1565.2898,-321.1394C1590.926,-318.1604 1621.4189,-314.9665 1649,-313 2123.6198,-279.1602 2244.8453,-316.8292 2719,-277 2732.8926,-275.833 2747.6779,-274.1643 2761.8579,-272.3504"/>
+<polygon fill="#191970" stroke="#191970" points="2762.5968,-275.7834 2772.0549,-271.009 2761.6838,-268.8432 2762.5968,-275.7834"/>
</g>
<!-- Node41->Node40 -->
<g id="edge125" class="edge">
<title>Node41->Node40</title>
-<path fill="none" stroke="#191970" d="M2161.9118,-313.4123C2177.4446,-304.18 2196.5082,-291.3929 2211,-277 2230.4831,-257.6498 2247.3223,-230.6045 2257.3137,-212.8295"/>
-<polygon fill="#191970" stroke="#191970" points="2260.4061,-214.4692 2262.1332,-204.0161 2254.2644,-211.1106 2260.4061,-214.4692"/>
+<path fill="none" stroke="#191970" d="M1565.1824,-321.2522C1656.0832,-309.7053 1824.1163,-287.3804 1850,-277 1872.2028,-268.0958 1871.3741,-253.7663 1894,-246 2053.29,-191.3244 2485.9625,-243.546 2651,-210 2654.456,-209.2975 2658.0064,-208.3571 2661.5091,-207.2852"/>
+<polygon fill="#191970" stroke="#191970" points="2662.8,-210.543 2671.1364,-204.0044 2660.5419,-203.9172 2662.8,-210.543"/>
</g>
<!-- Node43 -->
<g id="node39" class="node">
<title>Node43</title>
<g id="a_node39"><a xlink:href="runtime_2module_8h.html" target="_top" xlink:title="Runtime container of the functions generated by TVM, This is used to support dynamically link...">
-<polygon fill="#ffffff" stroke="#ff0000" points="2500,-185 2500,-204 2626,-204 2626,-185 2500,-185"/>
-<text text-anchor="middle" x="2563" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/module.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1290,-185 1290,-204 1416,-204 1416,-185 1290,-185"/>
+<text text-anchor="middle" x="1353" y="-192" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/runtime/module.h</text>
</a>
</g>
</g>
<!-- Node41->Node43 -->
<g id="edge115" class="edge">
<title>Node41->Node43</title>
-<path fill="none" stroke="#191970" d="M2192.0347,-313.8758C2193.7042,-313.5737 2195.3617,-313.2811 2197,-313 2320.6611,-291.7784 2365.3567,-334.2554 2477,-277 2506.1571,-262.047 2530.0821,-231.4189 2545.6761,-211.9798"/>
-<polygon fill="#191970" stroke="#191970" points="2548.4354,-214.1343 2552.0416,-204.172 2543.0099,-209.711 2548.4354,-214.1343"/>
+<path fill="none" stroke="#191970" d="M1466.2313,-313.3249C1447.4784,-304.2347 1425.8639,-291.6136 1409,-277 1387.2424,-258.1456 1367.5977,-231.1661 1357.8475,-213.256"/>
+<polygon fill="#191970" stroke="#191970" points="1360.9245,-211.5801 1353.4029,-204.1184 1354.6297,-214.642 1360.9245,-211.5801"/>
</g>
<!-- Node45 -->
<g id="node40" class="node">
<title>Node45</title>
-<polygon fill="#ffffff" stroke="#bfbfbf" points="2258.5,-252 2258.5,-271 2299.5,-271 2299.5,-252 2258.5,-252"/>
-<text text-anchor="middle" x="2279" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tuple</text>
+<polygon fill="#ffffff" stroke="#bfbfbf" points="1902.5,-252 1902.5,-271 1943.5,-271 1943.5,-252 1902.5,-252"/>
+<text text-anchor="middle" x="1923" y="-259" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tuple</text>
</g>
<!-- Node41->Node45 -->
<g id="edge129" class="edge">
<title>Node41->Node45</title>
-<path fill="none" stroke="#191970" d="M2166.5408,-313.4639C2191.1987,-302.0702 2224.8248,-286.5327 2248.8706,-275.4219"/>
-<polygon fill="#191970" stroke="#191970" points="2250.6057,-278.4757 2258.2154,-271.1039 2247.6695,-272.1213 2250.6057,-278.4757"/>
+<path fill="none" stroke="#191970" d="M1565.0027,-321.5749C1663.9359,-309.6267 1857.6121,-285.5945 1888,-277 1890.1822,-276.3828 1892.4046,-275.6548 1894.6176,-274.8563"/>
+<polygon fill="#191970" stroke="#191970" points="1896.1402,-278.0163 1904.1064,-271.0313 1893.5231,-271.5239 1896.1402,-278.0163"/>
</g>
<!-- Node43->Node11 -->
<g id="edge116" class="edge">
<title>Node43->Node11</title>
-<path fill="none" stroke="#191970" d="M2559.2889,-184.9045C2554.3296,-170.3785 2547.4244,-142.4423 2559,-123 2567.1879,-109.2477 2580.4689,-98.942 2594.5391,-91.3129"/>
-<polygon fill="#191970" stroke="#191970" points="2596.4609,-94.266 2603.8537,-86.6767 2593.3417,-87.9994 2596.4609,-94.266"/>
+<path fill="none" stroke="#191970" d="M1395.3709,-184.9738C1468.5011,-168.9846 1623.3377,-137.116 1756,-123 2192.936,-76.5076 2304.9618,-104.8895 2744,-87 2836.8012,-83.2186 2943.6586,-78.3183 3013.3335,-75.0498"/>
+<polygon fill="#191970" stroke="#191970" points="3013.5128,-78.5454 3023.3375,-74.5798 3013.1842,-71.5531 3013.5128,-78.5454"/>
</g>
<!-- Node43->Node10 -->
<g id="edge117" class="edge">
<title>Node43->Node10</title>
-<path fill="none" stroke="#191970" d="M2503.4693,-184.9927C2487.1466,-182.6916 2469.4235,-180.4734 2453,-179 2342.84,-169.117 1579.9546,-143.2105 1343.0306,-135.3164"/>
-<polygon fill="#191970" stroke="#191970" points="1342.9277,-131.8111 1332.8167,-134.9764 1342.6947,-138.8073 1342.9277,-131.8111"/>
+<path fill="none" stroke="#191970" d="M1416.0324,-186.2697C1503.9795,-174.7861 1662.9632,-154.0271 1754.3309,-142.0969"/>
+<polygon fill="#191970" stroke="#191970" points="1754.8513,-145.5588 1764.3139,-140.7934 1753.9449,-138.6177 1754.8513,-145.5588"/>
</g>
<!-- Node43->Node16 -->
<g id="edge119" class="edge">
<title>Node43->Node16</title>
-<path fill="none" stroke="#191970" d="M2501.1379,-184.9945C2485.4603,-182.8027 2468.6325,-180.637 2453,-179 2092.3392,-141.2324 2001.0888,-142.857 1639,-123 1389.6035,-109.3231 583.3849,-79.0689 412.1546,-72.6937"/>
-<polygon fill="#191970" stroke="#191970" points="412.149,-69.1912 402.0258,-72.3169 411.8887,-76.1863 412.149,-69.1912"/>
+<path fill="none" stroke="#191970" d="M1314.1043,-184.9507C1218.3329,-161.438 972.5096,-101.0862 884.1971,-79.4047"/>
+<polygon fill="#191970" stroke="#191970" points="884.7749,-75.9427 874.2288,-76.9574 883.1059,-82.7408 884.7749,-75.9427"/>
</g>
<!-- Node43->Node20 -->
<g id="edge121" class="edge">
<title>Node43->Node20</title>
-<path fill="none" stroke="#191970" d="M2502.6008,-184.9597C2486.5171,-182.7014 2469.1294,-180.5126 2453,-179 2114.1107,-147.2192 2027.1707,-171.6218 1688,-143 1672.1305,-141.6608 1654.5994,-139.6018 1639.8341,-137.6995"/>
-<polygon fill="#191970" stroke="#191970" points="1640.0067,-134.192 1629.6349,-136.3545 1639.0915,-141.1319 1640.0067,-134.192"/>
+<path fill="none" stroke="#191970" d="M1416.2208,-189.597C1575.7903,-177.2218 1990.7725,-145.0385 2112.1732,-135.6234"/>
+<polygon fill="#191970" stroke="#191970" points="2112.6485,-139.0971 2122.3479,-134.8343 2112.1072,-132.1181 2112.6485,-139.0971"/>
</g>
<!-- Node43->Node23 -->
<g id="edge118" class="edge">
<title>Node43->Node23</title>
-<path fill="none" stroke="#191970" d="M2568.3125,-184.8906C2573.2239,-176.0068 2580.6579,-162.5601 2586.7275,-151.5811"/>
-<polygon fill="#191970" stroke="#191970" points="2589.8373,-153.1899 2591.6126,-142.7449 2583.7112,-149.8031 2589.8373,-153.1899"/>
+<path fill="none" stroke="#191970" d="M1289.8927,-185.5574C1198.8439,-172.6553 1033.3254,-149.2005 958.0918,-138.5395"/>
+<polygon fill="#191970" stroke="#191970" points="958.4845,-135.0603 948.0923,-137.1225 957.5023,-141.991 958.4845,-135.0603"/>
</g>
<!-- Node43->Node36 -->
<g id="edge120" class="edge">
<title>Node43->Node36</title>
-<path fill="none" stroke="#191970" d="M2626.0355,-184.9919C2641.3815,-182.8502 2657.7632,-180.7081 2673,-179 2853.1227,-158.8072 3067.2883,-143.3772 3167.2904,-136.6692"/>
-<polygon fill="#191970" stroke="#191970" points="3167.7271,-140.148 3177.4719,-135.99 3167.2611,-133.1635 3167.7271,-140.148"/>
+<path fill="none" stroke="#191970" d="M1289.7368,-190.7434C1236.6722,-187.5958 1158.8896,-182.9897 1091,-179 816.9506,-162.8947 489.2779,-143.8273 359.7149,-136.2955"/>
+<polygon fill="#191970" stroke="#191970" points="359.8159,-132.7956 349.6296,-135.7093 359.4097,-139.7838 359.8159,-132.7956"/>
</g>
<!-- Node43->Node41 -->
<g id="edge122" class="edge">
<title>Node43->Node41</title>
-<path fill="none" stroke="#191970" d="M2562.4478,-204.172C2555.0484,-222.0109 2528.4718,-259.8342 2495,-277 2383.3567,-334.2554 2338.6611,-291.7784 2215,-313 2210.8885,-313.7056 2206.6557,-314.483 2202.3804,-315.3026"/>
-<polygon fill="#191970" stroke="#191970" points="2201.4993,-311.9089 2192.3653,-317.2774 2202.8535,-318.7767 2201.4993,-311.9089"/>
+<path fill="none" stroke="#191970" d="M1364.274,-204.1184C1378.1821,-220.5807 1401.0933,-254.5501 1427,-277 1440.9654,-289.1019 1458.1887,-299.8373 1473.1568,-308.3326"/>
+<polygon fill="#191970" stroke="#191970" points="1471.6951,-311.5245 1482.1358,-313.3249 1475.0966,-305.4065 1471.6951,-311.5245"/>
</g>
<!-- Node55->Node49 -->
<g id="edge141" class="edge">
<title>Node55->Node49</title>
-<path fill="none" stroke="#191970" d="M986.2127,-563.3741C1072.6472,-552.8517 1258.1772,-530.2654 1348.492,-519.2705"/>
-<polygon fill="#191970" stroke="#191970" points="1348.995,-522.7352 1358.4987,-518.0523 1348.149,-515.7865 1348.995,-522.7352"/>
+<path fill="none" stroke="#191970" d="M1310.1826,-559.3733C1284.4668,-550.1419 1245.462,-536.1402 1217.0401,-525.9375"/>
+<polygon fill="#191970" stroke="#191970" points="1218.0619,-522.5857 1207.4674,-522.5011 1215.6968,-529.174 1218.0619,-522.5857"/>
</g>
<!-- Node56 -->
<g id="node42" class="node">
<title>Node56</title>
<g id="a_node42"><a xlink:href="ir_2op_8h.html" target="_top" xlink:title="Primitive operators(builtin intrinsics) and registry for them. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="661,-447.5 661,-466.5 731,-466.5 731,-447.5 661,-447.5"/>
-<text text-anchor="middle" x="696" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/op.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="1290,-447.5 1290,-466.5 1360,-466.5 1360,-447.5 1290,-447.5"/>
+<text text-anchor="middle" x="1325" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/op.h</text>
</a>
</g>
</g>
<!-- Node55->Node56 -->
<g id="edge135" class="edge">
<title>Node55->Node56</title>
-<path fill="none" stroke="#191970" d="M919.1967,-559.4509C875.9794,-539.6135 776.3366,-493.8758 726.1065,-470.8194"/>
-<polygon fill="#191970" stroke="#191970" points="727.2991,-467.5157 716.7506,-466.5249 724.3788,-473.8775 727.2991,-467.5157"/>
+<path fill="none" stroke="#191970" d="M1335.9769,-559.4509C1334.0197,-541.184 1329.7095,-500.9553 1327.1105,-476.6976"/>
+<polygon fill="#191970" stroke="#191970" points="1330.566,-476.0951 1326.0205,-466.5249 1323.6059,-476.8409 1330.566,-476.0951"/>
</g>
<!-- Node56->Node3 -->
<g id="edge136" class="edge">
<title>Node56->Node3</title>
-<path fill="none" stroke="#191970" d="M731.0232,-450.273C776.4836,-441.4914 858.2543,-425.5261 928,-411 934.805,-409.5827 941.9694,-408.0576 949.0233,-406.5366"/>
-<polygon fill="#191970" stroke="#191970" points="950.1933,-409.8643 959.2243,-404.3243 948.7096,-403.0233 950.1933,-409.8643"/>
+<path fill="none" stroke="#191970" d="M1289.7206,-447.5254C1249.9135,-436.8349 1185.0496,-419.4151 1141.471,-407.7117"/>
+<polygon fill="#191970" stroke="#191970" points="1142.2624,-404.3002 1131.6968,-405.0867 1140.4467,-411.0607 1142.2624,-404.3002"/>
</g>
<!-- Node56->Node16 -->
<g id="edge138" class="edge">
<title>Node56->Node16</title>
-<path fill="none" stroke="#191970" d="M660.9334,-455.6193C580.6295,-451.9815 385.8663,-440.3169 328,-411 273.1711,-383.2219 160.7204,-267.1105 138,-210 123.7067,-174.0721 113.6699,-153.0526 138,-123 164.071,-90.797 288.2771,-77.8831 347.9073,-73.4741"/>
-<polygon fill="#191970" stroke="#191970" points="348.185,-76.9633 357.9145,-72.7695 347.6934,-69.9805 348.185,-76.9633"/>
+<path fill="none" stroke="#191970" d="M1289.6482,-454.5476C1165.9315,-445.4511 753.8955,-410.6747 642,-344 555.1393,-292.2427 514.6574,-192.6022 588,-123 620.9806,-91.7014 756.5456,-78.2196 819.5432,-73.5592"/>
+<polygon fill="#191970" stroke="#191970" points="819.9871,-77.0366 829.7142,-72.8372 819.4914,-70.0542 819.9871,-77.0366"/>
</g>
<!-- Node56->Node18 -->
<g id="edge139" class="edge">
<title>Node56->Node18</title>
-<path fill="none" stroke="#191970" d="M697.1697,-447.4483C702.2836,-409.2979 725.7981,-265.7283 799,-179 831.1079,-140.9591 848.2312,-137.0059 896,-123 1059.9521,-74.929 1493.3421,-95.1852 1664,-87 1767.342,-82.0434 1889.6844,-75.9894 1947.1199,-73.1363"/>
-<polygon fill="#191970" stroke="#191970" points="1947.5797,-76.6179 1957.3936,-72.6257 1947.2322,-69.6265 1947.5797,-76.6179"/>
+<path fill="none" stroke="#191970" d="M1360.0467,-453.2178C1432.0109,-445.4371 1601.6487,-427.0218 1744,-411 2260.6052,-352.8556 2392.0634,-355.3001 2906,-277 2925.7335,-273.9935 3246.2691,-225.3745 3259,-210 3297.3429,-163.6951 3225.5567,-165.542 2893,-123 2717.379,-100.5339 2506.0709,-81.3301 2425.9541,-74.3291"/>
+<polygon fill="#191970" stroke="#191970" points="2426.237,-70.8406 2415.9713,-73.4607 2425.6303,-77.8142 2426.237,-70.8406"/>
</g>
<!-- Node56->Node20 -->
<g id="edge140" class="edge">
<title>Node56->Node20</title>
-<path fill="none" stroke="#191970" d="M713.4209,-447.3504C801.5373,-398.549 1194.2275,-181.1557 1201,-179 1236.0665,-167.8384 1481.5452,-144.4443 1572.3409,-136.0663"/>
-<polygon fill="#191970" stroke="#191970" points="1572.7382,-139.5446 1582.3756,-135.1435 1572.0971,-132.574 1572.7382,-139.5446"/>
+<path fill="none" stroke="#191970" d="M1360.0168,-455.7453C1466.1156,-450.3748 1789.7463,-422.2988 2002,-277 2056.4649,-239.7159 2108.0725,-180.2309 2132.0257,-150.7528"/>
+<polygon fill="#191970" stroke="#191970" points="2135.0177,-152.6163 2138.5441,-142.6255 2129.557,-148.2366 2135.0177,-152.6163"/>
</g>
<!-- Node56->Node48 -->
<g id="edge137" class="edge">
<title>Node56->Node48</title>
-<path fill="none" stroke="#191970" d="M706.6731,-447.3416C730.7274,-425.5742 789.4161,-372.465 819.5221,-345.2212"/>
-<polygon fill="#191970" stroke="#191970" points="822.1368,-347.5755 827.2031,-338.2705 817.4399,-342.3851 822.1368,-347.5755"/>
+<path fill="none" stroke="#191970" d="M1328.1256,-447.2501C1334.0804,-430.4063 1348.7598,-396.118 1374,-380 1476.0325,-314.8435 1525.4986,-368.7671 1644,-344 1648.4708,-343.0656 1653.1129,-341.9544 1657.7139,-340.7623"/>
+<polygon fill="#191970" stroke="#191970" points="1658.9089,-344.0645 1667.6386,-338.0609 1657.0703,-337.3102 1658.9089,-344.0645"/>
</g>
<!-- Node72->Node16 -->
<g id="edge159" class="edge">
<title>Node72->Node16</title>
-<path fill="none" stroke="#191970" d="M1338.3316,-623.6215C1088.5239,-616.4995 0,-581.5338 0,-513 0,-513 0,-513 0,-194.5 0,-151.5536 27.26,-143.4961 65,-123 113.5181,-96.6505 277.364,-80.1177 347.5978,-74.0843"/>
-<polygon fill="#191970" stroke="#191970" points="347.9747,-77.565 357.6457,-73.2382 347.3873,-70.5897 347.9747,-77.565"/>
+<path fill="none" stroke="#191970" d="M1133.1473,-620.536C915.8129,-601.3757 76,-526.4659 76,-513 76,-513 76,-513 76,-457 76,-363.7765 0,-354.7235 0,-261.5 0,-261.5 0,-261.5 0,-194.5 0,-108.8325 95.5534,-142.38 179,-123 303.2478,-94.1441 702.8794,-77.1224 819.6649,-72.6754"/>
+<polygon fill="#191970" stroke="#191970" points="819.8972,-76.1693 829.7586,-72.2957 819.634,-69.1742 819.8972,-76.1693"/>
</g>
<!-- Node72->Node49 -->
<g id="edge152" class="edge">
<title>Node72->Node49</title>
-<path fill="none" stroke="#191970" d="M1389.9379,-615.4509C1391.7319,-597.184 1395.683,-556.9553 1398.0654,-532.6976"/>
-<polygon fill="#191970" stroke="#191970" points="1401.5702,-532.8191 1399.0645,-522.5249 1394.6038,-532.1349 1401.5702,-532.8191"/>
+<path fill="none" stroke="#191970" d="M1183.7442,-615.4509C1183.2549,-597.184 1182.1774,-556.9553 1181.5276,-532.6976"/>
+<polygon fill="#191970" stroke="#191970" points="1185.0217,-532.4276 1181.2551,-522.5249 1178.0242,-532.6151 1185.0217,-532.4276"/>
</g>
<!-- Node72->Node50 -->
<g id="edge151" class="edge">
<title>Node72->Node50</title>
-<path fill="none" stroke="#191970" d="M1379.7694,-615.4862C1353.3336,-588.2396 1277.0512,-509.6172 1242.8915,-474.4096"/>
-<polygon fill="#191970" stroke="#191970" points="1245.0109,-471.5678 1235.5354,-466.8279 1239.9869,-476.4422 1245.0109,-471.5678"/>
+<path fill="none" stroke="#191970" d="M1177.7289,-615.2552C1168.2431,-601.178 1149.0011,-575.0819 1127,-559 1075.8815,-521.6344 1056.7459,-520.8667 996,-503 995.6748,-502.9044 859.0863,-479.7366 779.9733,-466.3211"/>
+<polygon fill="#191970" stroke="#191970" points="780.1268,-462.7972 769.6824,-464.576 778.9565,-469.6987 780.1268,-462.7972"/>
</g>
<!-- Node72->Node30 -->
<g id="edge150" class="edge">
<title>Node72->Node30</title>
-<path fill="none" stroke="#191970" d="M1439.6223,-620.4771C1459.1181,-618.7465 1481.5937,-616.7652 1502,-615 1596.6506,-606.8126 1846.1098,-627.164 1928,-579 1984.3137,-545.8789 2010,-522.3318 2010,-457 2010,-457 2010,-457 2010,-395.5 2010,-338.6406 2071.0268,-296.6054 2109.1313,-275.9009"/>
-<polygon fill="#191970" stroke="#191970" points="2111.0472,-278.8481 2118.2739,-271.1006 2107.7931,-272.6505 2111.0472,-278.8481"/>
+<path fill="none" stroke="#191970" d="M1234.8545,-624.0092C1462.7732,-619.4573 2381.597,-599.8783 2439,-579 2615.9847,-514.6281 2773.1542,-335.935 2819.9128,-279.0981"/>
+<polygon fill="#191970" stroke="#191970" points="2822.858,-281.024 2826.4553,-271.0585 2817.4286,-276.6056 2822.858,-281.024"/>
</g>
<!-- Node64 -->
<g id="node44" class="node">
<title>Node64</title>
<g id="a_node44"><a xlink:href="ir_2function_8h.html" target="_top" xlink:title="Function nodes. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1027,-447.5 1027,-466.5 1125,-466.5 1125,-447.5 1027,-447.5"/>
-<text text-anchor="middle" x="1076" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/function.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="520,-447.5 520,-466.5 618,-466.5 618,-447.5 520,-447.5"/>
+<text text-anchor="middle" x="569" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/function.h</text>
</a>
</g>
</g>
<!-- Node72->Node64 -->
<g id="edge143" class="edge">
<title>Node72->Node64</title>
-<path fill="none" stroke="#191970" d="M1352.7223,-615.4461C1301.8868,-600.9395 1207.9591,-569.977 1139,-523 1119.4855,-509.7062 1101.3368,-489.5167 1089.5797,-474.9692"/>
-<polygon fill="#191970" stroke="#191970" points="1092.0577,-472.4576 1083.1301,-466.7524 1086.5514,-476.7797 1092.0577,-472.4576"/>
+<path fill="none" stroke="#191970" d="M1144.8678,-615.4344C1107.9573,-606.335 1051.1185,-592.1244 1002,-579 858.4801,-540.6517 689.1908,-491.9382 611.6575,-469.4329"/>
+<polygon fill="#191970" stroke="#191970" points="612.3592,-465.9921 601.7798,-466.5634 610.4064,-472.7142 612.3592,-465.9921"/>
</g>
<!-- Node73 -->
<g id="node45" class="node">
<title>Node73</title>
<g id="a_node45"><a xlink:href="stmt_8h.html" target="_top" xlink:title="TIR statements. ">
-<polygon fill="#ffffff" stroke="#000000" points="1417,-559.5 1417,-578.5 1501,-578.5 1501,-559.5 1417,-559.5"/>
-<text text-anchor="middle" x="1459" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/stmt.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="1987,-559.5 1987,-578.5 2071,-578.5 2071,-559.5 1987,-559.5"/>
+<text text-anchor="middle" x="2029" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/stmt.h</text>
</a>
</g>
</g>
<!-- Node72->Node73 -->
<g id="edge153" class="edge">
<title>Node72->Node73</title>
-<path fill="none" stroke="#191970" d="M1401.1931,-615.2455C1411.6321,-606.8943 1426.8655,-594.7076 1439.074,-584.9408"/>
-<polygon fill="#191970" stroke="#191970" points="1441.3244,-587.6227 1446.9467,-578.6427 1436.9515,-582.1566 1441.3244,-587.6227"/>
+<path fill="none" stroke="#191970" d="M1234.9513,-624.1187C1360.6933,-621.4902 1695.6985,-611.7259 1973,-579 1974.2689,-578.8502 1975.5526,-578.6901 1976.8463,-578.521"/>
+<polygon fill="#191970" stroke="#191970" points="1977.4849,-581.9652 1986.8824,-577.0725 1976.4849,-575.0369 1977.4849,-581.9652"/>
</g>
<!-- Node64->Node3 -->
<g id="edge144" class="edge">
<title>Node64->Node3</title>
-<path fill="none" stroke="#191970" d="M1063.9688,-447.3906C1051.9893,-437.8227 1033.3834,-422.9621 1019.1475,-411.5919"/>
-<polygon fill="#191970" stroke="#191970" points="1020.9447,-408.5479 1010.9468,-405.0419 1016.5761,-414.0175 1020.9447,-408.5479"/>
+<path fill="none" stroke="#191970" d="M618.1056,-451.2695C718.3021,-439.5767 945.3334,-413.0825 1046.1525,-401.3171"/>
+<polygon fill="#191970" stroke="#191970" points="1046.8439,-404.7603 1056.3708,-400.1247 1046.0325,-397.8075 1046.8439,-404.7603"/>
</g>
<!-- Node64->Node16 -->
<g id="edge148" class="edge">
<title>Node64->Node16</title>
-<path fill="none" stroke="#191970" d="M1026.8012,-455.4555C881.5217,-450.641 460.9926,-434.7906 404,-411 341.6685,-384.9807 203.7961,-271.5597 176,-210 160.0877,-174.7593 152.2392,-153.5047 176,-123 196.9472,-96.1074 295.7849,-81.0536 347.8038,-74.8704"/>
-<polygon fill="#191970" stroke="#191970" points="348.3119,-78.3352 357.8476,-73.7178 347.5138,-71.3808 348.3119,-78.3352"/>
+<path fill="none" stroke="#191970" d="M519.7799,-451.6045C425.3465,-438.2356 228,-394.3988 228,-261.5 228,-261.5 228,-261.5 228,-194.5 228,-161.6194 221.809,-144.1317 247,-123 291.1021,-86.0044 700.4015,-74.648 819.5845,-72.1116"/>
+<polygon fill="#191970" stroke="#191970" points="819.9528,-75.6048 829.8785,-71.8993 819.8084,-68.6063 819.9528,-75.6048"/>
</g>
<!-- Node64->Node17 -->
<g id="edge149" class="edge">
<title>Node64->Node17</title>
-<path fill="none" stroke="#191970" d="M1026.8314,-452.0254C880.1882,-436.5337 456,-386.6923 456,-328.5 456,-328.5 456,-328.5 456,-261.5 456,-170.4342 525.6182,-163.8655 607,-123 645.8607,-103.4863 766.2323,-85.6554 832.3618,-77.0013"/>
-<polygon fill="#191970" stroke="#191970" points="833.0199,-80.4455 842.4887,-75.6923 832.1225,-73.5032 833.0199,-80.4455"/>
+<path fill="none" stroke="#191970" d="M566.152,-447.3404C553.4969,-402.1878 507.1669,-210.4559 605,-123 632.9146,-98.0463 1186.4893,-78.2153 1349.3196,-72.9041"/>
+<polygon fill="#191970" stroke="#191970" points="1349.5888,-76.3974 1359.4702,-72.5754 1349.3621,-69.401 1349.5888,-76.3974"/>
</g>
<!-- Node64->Node21 -->
<g id="edge145" class="edge">
<title>Node64->Node21</title>
-<path fill="none" stroke="#191970" d="M1095.5189,-447.4774C1113.5484,-438.5224 1140.9454,-424.5086 1164,-411 1237.2001,-368.1093 1319.7835,-312.2563 1362.9871,-282.4334"/>
-<polygon fill="#191970" stroke="#191970" points="1365.1086,-285.2217 1371.3398,-276.653 1361.1251,-279.4656 1365.1086,-285.2217"/>
+<path fill="none" stroke="#191970" d="M596.5847,-447.4383C685.1036,-416.7549 963.0341,-320.4155 1080.1423,-279.8221"/>
+<polygon fill="#191970" stroke="#191970" points="1081.3277,-283.1156 1089.6298,-276.5335 1079.035,-276.5017 1081.3277,-283.1156"/>
</g>
<!-- Node64->Node33 -->
<g id="edge147" class="edge">
<title>Node64->Node33</title>
-<path fill="none" stroke="#191970" d="M1083.245,-447.3462C1111.3808,-409.8556 1213.8865,-273.268 1255.3824,-217.9752"/>
-<polygon fill="#191970" stroke="#191970" points="1258.2946,-219.9258 1261.4977,-209.8267 1252.6958,-215.724 1258.2946,-219.9258"/>
+<path fill="none" stroke="#191970" d="M568.1604,-447.4046C565.7743,-413.3774 562.8312,-297.756 627,-246 651.294,-226.4055 838.9415,-208.6577 945.5607,-200.0235"/>
+<polygon fill="#191970" stroke="#191970" points="945.9295,-203.5053 955.6175,-199.2166 945.3696,-196.5277 945.9295,-203.5053"/>
</g>
<!-- Node64->Node42 -->
<g id="edge146" class="edge">
<title>Node64->Node42</title>
-<path fill="none" stroke="#191970" d="M1125.3436,-451.8854C1140.619,-450.3095 1157.5047,-448.575 1173,-447 1230.1028,-441.1959 2147.3279,-364.3402 2201,-344 2223.0308,-335.651 2221.9584,-321.3203 2244,-313 2398.1609,-254.8071 2449.101,-294.0031 2613,-277 2634.9921,-274.7185 2658.901,-272.1309 2680.6773,-269.7324"/>
-<polygon fill="#191970" stroke="#191970" points="2681.2918,-273.1859 2690.847,-268.6091 2680.5232,-266.2282 2681.2918,-273.1859"/>
+<path fill="none" stroke="#191970" d="M571.4005,-447.3775C577.8665,-422.894 597.5285,-357.057 632,-313 641.1664,-301.2847 653.5817,-290.7709 665.2345,-282.3446"/>
+<polygon fill="#191970" stroke="#191970" points="667.3067,-285.1668 673.5274,-276.5905 663.3162,-279.4156 667.3067,-285.1668"/>
</g>
<!-- Node73->Node16 -->
<g id="edge155" class="edge">
<title>Node73->Node16</title>
-<path fill="none" stroke="#191970" d="M1416.8578,-567.9403C1181.2797,-561.7469 38,-528.0306 38,-457 38,-457 38,-457 38,-194.5 38,-150.6456 67.475,-143.9531 106,-123 147.4008,-100.4828 284.3622,-82.4227 347.5459,-75.068"/>
-<polygon fill="#191970" stroke="#191970" points="348.2016,-78.5159 357.7382,-73.9003 347.4047,-71.5614 348.2016,-78.5159"/>
+<path fill="none" stroke="#191970" d="M1986.9045,-567.1358C1858.5104,-561.3872 1460.554,-543.1044 1131,-523 779.2185,-501.5396 674.1054,-579.1801 340,-467 296.7609,-452.4819 284.8671,-444.58 254,-411 166.1323,-315.4095 100.6974,-200.3255 205,-123 254.7001,-86.1545 695.3086,-74.6249 819.5288,-72.0922"/>
+<polygon fill="#191970" stroke="#191970" points="819.9633,-75.5844 829.892,-71.887 819.8247,-68.5858 819.9633,-75.5844"/>
</g>
<!-- Node73->Node17 -->
<g id="edge156" class="edge">
<title>Node73->Node17</title>
-<path fill="none" stroke="#191970" d="M1416.6835,-567.2063C1212.6174,-558.3288 335.463,-517.4929 225,-467 183.6961,-448.1199 152,-440.9144 152,-395.5 152,-395.5 152,-395.5 152,-328.5 152,-268.8016 207.7978,-278.3054 258,-246 379.8742,-167.5734 419.0967,-156.9162 560,-123 655.8253,-99.9343 770.4636,-84.1671 832.242,-76.617"/>
-<polygon fill="#191970" stroke="#191970" points="832.9104,-80.0619 842.4192,-75.3894 832.072,-73.1123 832.9104,-80.0619"/>
+<path fill="none" stroke="#191970" d="M1991.5549,-559.4287C1895.3605,-533.8652 1633.5564,-458.082 1440,-344 1360.3844,-297.0746 1319.9665,-293.7989 1281,-210 1256.9643,-158.3104 1325.6023,-109.4905 1366.2943,-86.0354"/>
+<polygon fill="#191970" stroke="#191970" points="1368.223,-88.9668 1375.2376,-81.0267 1364.8025,-82.8594 1368.223,-88.9668"/>
</g>
<!-- Node73->Node18 -->
<g id="edge157" class="edge">
<title>Node73->Node18</title>
-<path fill="none" stroke="#191970" d="M1501.249,-564.3516C1539.2162,-558.7881 1595.5229,-547.021 1639,-523 1827.0972,-419.0767 1942.7385,-163.0479 1972.3959,-90.7765"/>
-<polygon fill="#191970" stroke="#191970" points="1975.7862,-91.7279 1976.2817,-81.1446 1969.2945,-89.1089 1975.7862,-91.7279"/>
+<path fill="none" stroke="#191970" d="M2071.1505,-566.739C2139.3117,-562.3178 2269.7646,-550.4491 2306,-523 2444.9204,-417.7649 2410.1687,-163.6497 2396.9292,-91.1249"/>
+<polygon fill="#191970" stroke="#191970" points="2400.3281,-90.2635 2395.0189,-81.0949 2393.4517,-91.5732 2400.3281,-90.2635"/>
</g>
<!-- Node73->Node20 -->
<g id="edge158" class="edge">
<title>Node73->Node20</title>
-<path fill="none" stroke="#191970" d="M1462.207,-559.488C1480.6957,-504.6509 1573.7429,-228.6741 1599.4253,-152.5004"/>
-<polygon fill="#191970" stroke="#191970" points="1602.8305,-153.3556 1602.7089,-142.7615 1596.1974,-151.1192 1602.8305,-153.3556"/>
+<path fill="none" stroke="#191970" d="M2045.0523,-559.3084C2068.8077,-545.2974 2115.0446,-519.2829 2157,-503 2214.0895,-480.8435 2252.7823,-515.6309 2290,-467 2350.5708,-387.8544 2311.093,-326.255 2252,-246 2224.5855,-208.7681 2186.1966,-170.5896 2163.9303,-149.5214"/>
+<polygon fill="#191970" stroke="#191970" points="2166.1974,-146.8493 2156.5066,-142.567 2161.4118,-151.9579 2166.1974,-146.8493"/>
</g>
<!-- Node73->Node49 -->
<g id="edge154" class="edge">
<title>Node73->Node49</title>
-<path fill="none" stroke="#191970" d="M1448.723,-559.2455C1440.1782,-551.1352 1427.8222,-539.4075 1417.6915,-529.7919"/>
-<polygon fill="#191970" stroke="#191970" points="1419.8218,-526.9884 1410.1592,-522.6427 1415.0028,-532.0655 1419.8218,-526.9884"/>
+<path fill="none" stroke="#191970" d="M1986.9475,-566.223C1845.6539,-556.8922 1388.1575,-526.6802 1233.0129,-516.4348"/>
+<polygon fill="#191970" stroke="#191970" points="1233.0139,-512.9273 1222.805,-515.7607 1232.5526,-519.9121 1233.0139,-512.9273"/>
</g>
<!-- Node74->Node18 -->
-<g id="edge169" class="edge">
+<g id="edge170" class="edge">
<title>Node74->Node18</title>
-<path fill="none" stroke="#191970" d="M2008.8865,-622.046C2113.3493,-616.598 2331.1649,-603.2482 2514,-579 2813.206,-539.3183 2892.0897,-523.912 3172,-411 3248.4048,-380.1794 3338,-410.8869 3338,-328.5 3338,-328.5 3338,-328.5 3338,-194.5 3338,-153.5815 3316.5842,-141.328 3280,-123 3250.7821,-108.3624 2211.4108,-78.0863 2012.6621,-72.4232"/>
-<polygon fill="#191970" stroke="#191970" points="2012.6526,-68.9217 2002.5571,-72.1359 2012.4536,-75.9188 2012.6526,-68.9217"/>
+<path fill="none" stroke="#191970" d="M1841.7057,-678.3046C1964.908,-672.5156 2237.32,-657.562 2328,-635 2641.0169,-557.1185 2986.8565,-659.1821 3277,-210 3327.6122,-131.6454 3269.054,-162.4188 2986,-123 2773.5573,-93.4147 2515.7803,-77.9492 2425.8383,-73.156"/>
+<polygon fill="#191970" stroke="#191970" points="2425.886,-69.6537 2415.7161,-72.6241 2425.5186,-76.6441 2425.886,-69.6537"/>
</g>
<!-- Node74->Node36 -->
-<g id="edge168" class="edge">
+<g id="edge169" class="edge">
<title>Node74->Node36</title>
-<path fill="none" stroke="#191970" d="M2008.5733,-623.4728C2226.1839,-617.9225 2948.4481,-598.1401 2994,-579 3031.5779,-563.2104 3262,-369.2604 3262,-328.5 3262,-328.5 3262,-328.5 3262,-261.5 3262,-221.195 3244.596,-176.5903 3233.3445,-151.9654"/>
-<polygon fill="#191970" stroke="#191970" points="3236.4056,-150.2511 3228.9678,-142.7058 3230.0769,-153.2425 3236.4056,-150.2511"/>
+<path fill="none" stroke="#191970" d="M1718.2574,-680.3275C1408.4369,-676.5566 38,-655.0155 38,-569 38,-569 38,-569 38,-457 38,-306.7403 210.8859,-187.8364 277.3735,-147.6648"/>
+<polygon fill="#191970" stroke="#191970" points="279.2132,-150.6429 286.0199,-142.5238 275.6357,-144.6261 279.2132,-150.6429"/>
</g>
<!-- Node74->Node49 -->
<g id="edge162" class="edge">
<title>Node74->Node49</title>
-<path fill="none" stroke="#191970" d="M1885.2293,-616.5459C1827.5126,-608.2803 1739.071,-594.6759 1663,-579 1585.411,-563.0112 1496.1839,-539.5238 1444.146,-525.305"/>
-<polygon fill="#191970" stroke="#191970" points="1444.8184,-521.8602 1434.2487,-522.5898 1442.9664,-528.6108 1444.8184,-521.8602"/>
+<path fill="none" stroke="#191970" d="M1718.4386,-675.8812C1622.1752,-666.246 1431.5544,-640.5479 1282,-579 1251.4281,-566.4184 1220.163,-544.2642 1200.6796,-529.123"/>
+<polygon fill="#191970" stroke="#191970" points="1202.6123,-526.1887 1192.6017,-522.7192 1198.2638,-531.6742 1202.6123,-526.1887"/>
</g>
<!-- Node74->Node8 -->
<g id="edge161" class="edge">
<title>Node74->Node8</title>
-<path fill="none" stroke="#191970" d="M1945.368,-615.1788C1942.8998,-601.8997 1937.471,-577.727 1928,-559 1848.7642,-402.3271 1833.6517,-342.5297 1687,-246 1645.9465,-218.9776 1591.6046,-206.1323 1549.6255,-200.0272"/>
-<polygon fill="#191970" stroke="#191970" points="1549.8889,-196.5308 1539.5082,-198.6503 1548.9449,-203.4669 1549.8889,-196.5308"/>
+<path fill="none" stroke="#191970" d="M1841.7214,-677.8653C1929.0046,-672.7177 2093.2274,-660.4007 2231,-635 2325.1495,-617.642 2380.7649,-654.9875 2439,-579 2531.7808,-457.9359 2452.5218,-364.1031 2356,-246 2344.526,-231.9605 2328.6106,-219.2709 2315.4004,-210.0006"/>
+<polygon fill="#191970" stroke="#191970" points="2317.0164,-206.8678 2306.7749,-204.1545 2313.0891,-212.6623 2317.0164,-206.8678"/>
</g>
-<!-- Node74->Node73 -->
+<!-- Node74->Node72 -->
<g id="edge167" class="edge">
+<title>Node74->Node72</title>
+<path fill="none" stroke="#191970" d="M1718.1374,-676.9763C1618.948,-670.2245 1418.2751,-655.3306 1249,-635 1247.6928,-634.843 1246.3715,-634.6799 1245.04,-634.5114"/>
+<polygon fill="#191970" stroke="#191970" points="1245.0726,-630.9849 1234.6976,-633.1315 1244.1468,-637.9234 1245.0726,-630.9849"/>
+</g>
+<!-- Node74->Node73 -->
+<g id="edge168" class="edge">
<title>Node74->Node73</title>
-<path fill="none" stroke="#191970" d="M1885.1945,-617.9076C1788.1411,-606.7703 1601.738,-585.3798 1511.0772,-574.9761"/>
-<polygon fill="#191970" stroke="#191970" points="1511.3661,-571.4863 1501.0323,-573.8234 1510.568,-578.4407 1511.3661,-571.4863"/>
+<path fill="none" stroke="#191970" d="M1801.2296,-671.4509C1845.4244,-651.5722 1947.441,-605.6852 1998.5963,-582.6756"/>
+<polygon fill="#191970" stroke="#191970" points="2000.14,-585.819 2007.8241,-578.5249 1997.2685,-579.4351 2000.14,-585.819"/>
</g>
<!-- Node75 -->
<g id="node47" class="node">
<title>Node75</title>
<g id="a_node47"><a xlink:href="tir_2expr__functor_8h.html" target="_top" xlink:title="Functors for tir expressions. ">
-<polygon fill="#ffffff" stroke="#000000" points="1797,-559.5 1797,-578.5 1919,-578.5 1919,-559.5 1797,-559.5"/>
-<text text-anchor="middle" x="1858" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/expr_functor.h</text>
+<polygon fill="#ffffff" stroke="#000000" points="2308,-559.5 2308,-578.5 2430,-578.5 2430,-559.5 2308,-559.5"/>
+<text text-anchor="middle" x="2369" y="-566.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/tir/expr_functor.h</text>
</a>
</g>
</g>
<!-- Node74->Node75 -->
<g id="edge163" class="edge">
<title>Node74->Node75</title>
-<path fill="none" stroke="#191970" d="M1931.4974,-615.2455C1917.842,-606.6534 1897.7343,-594.0014 1882.0011,-584.1018"/>
-<polygon fill="#191970" stroke="#191970" points="1883.6529,-581.006 1873.325,-578.6427 1879.9249,-586.9307 1883.6529,-581.006"/>
+<path fill="none" stroke="#191970" d="M1841.6576,-678.3734C1921.5362,-673.9739 2064.6932,-662.6868 2184,-635 2239.565,-622.1054 2301.5015,-597.8833 2337.7598,-582.642"/>
+<polygon fill="#191970" stroke="#191970" points="2339.4435,-585.7295 2347.279,-578.5982 2336.7065,-579.2868 2339.4435,-585.7295"/>
</g>
<!-- Node75->Node18 -->
<g id="edge166" class="edge">
<title>Node75->Node18</title>
-<path fill="none" stroke="#191970" d="M1907.8752,-559.4886C2085.0584,-524.8787 2681.287,-401.1577 2826,-277 2851.0452,-255.5122 2850.7087,-242.1842 2858,-210 2875.1133,-134.461 2782.7041,-139.3677 2707,-123 2571.6428,-93.7349 2136.2397,-76.8638 2012.8628,-72.5878"/>
-<polygon fill="#191970" stroke="#191970" points="2012.6753,-69.0794 2002.5613,-72.235 2012.4357,-76.0753 2012.6753,-69.0794"/>
+<path fill="none" stroke="#191970" d="M2379.5968,-559.2239C2403.59,-536.2333 2461.4241,-475.7704 2480,-411 2515.9023,-285.816 2487.4939,-237.7976 2426,-123 2419.8442,-111.5083 2412.0035,-99.1949 2405.5094,-89.4906"/>
+<polygon fill="#191970" stroke="#191970" points="2408.3161,-87.3938 2399.7948,-81.0979 2402.53,-91.3336 2408.3161,-87.3938"/>
</g>
<!-- Node75->Node49 -->
<g id="edge165" class="edge">
<title>Node75->Node49</title>
-<path fill="none" stroke="#191970" d="M1796.9284,-560.1879C1793.9103,-559.7799 1790.9214,-559.3818 1788,-559 1667.7117,-543.2798 1526.1337,-527.0743 1451.7151,-518.7335"/>
-<polygon fill="#191970" stroke="#191970" points="1451.9964,-515.2432 1441.6692,-517.6096 1451.2181,-522.1998 1451.9964,-515.2432"/>
+<path fill="none" stroke="#191970" d="M2307.6296,-566.1071C2099.8977,-556.315 1425.4269,-524.5218 1232.8035,-515.4419"/>
+<polygon fill="#191970" stroke="#191970" points="1232.8204,-511.9389 1222.6667,-514.9641 1232.4907,-518.9312 1232.8204,-511.9389"/>
</g>
<!-- Node75->Node8 -->
<g id="edge164" class="edge">
<title>Node75->Node8</title>
-<path fill="none" stroke="#191970" d="M1848.4049,-559.3663C1798.3215,-509.0815 1566.7386,-276.5673 1501.9077,-211.4757"/>
-<polygon fill="#191970" stroke="#191970" points="1504.0684,-208.6854 1494.5317,-204.07 1499.1087,-213.6252 1504.0684,-208.6854"/>
+<path fill="none" stroke="#191970" d="M2372.0986,-559.2275C2377.9401,-539.9094 2390,-495.3796 2390,-457 2390,-457 2390,-457 2390,-395.5 2390,-320.8702 2334.3963,-245.2085 2306.5317,-211.9777"/>
+<polygon fill="#191970" stroke="#191970" points="2308.954,-209.4259 2299.7856,-204.1165 2303.6419,-213.9846 2308.954,-209.4259"/>
</g>
<!-- Node76->Node3 -->
-<g id="edge171" class="edge">
+<g id="edge172" class="edge">
<title>Node76->Node3</title>
-<path fill="none" stroke="#191970" d="M1510.2581,-617.4724C1420.4608,-604.5103 1240.923,-574.3064 1096,-523 1043.9118,-504.5595 1013.9653,-514.0558 985,-467 975.1008,-450.9181 981.5094,-429.2759 988.5715,-414.0284"/>
-<polygon fill="#191970" stroke="#191970" points="991.7222,-415.5541 993.1419,-405.0548 985.4846,-412.3772 991.7222,-415.5541"/>
+<path fill="none" stroke="#191970" d="M2115.4975,-615.4457C2111.0661,-600.6125 2100.3294,-572.3633 2080,-559 1930.6488,-460.8259 1454.5326,-509.783 1281,-467 1224.0071,-452.9489 1160.8829,-425.9057 1125.0659,-409.4043"/>
+<polygon fill="#191970" stroke="#191970" points="1126.2583,-406.0987 1115.7155,-405.0492 1123.3027,-412.4441 1126.2583,-406.0987"/>
</g>
<!-- Node76->Node54 -->
-<g id="edge179" class="edge">
+<g id="edge180" class="edge">
<title>Node76->Node54</title>
-<path fill="none" stroke="#191970" d="M1623.7932,-622.793C1850.6173,-613.6364 2684.3713,-576.5922 2789,-523 2833.9055,-499.9988 2867.9528,-449.735 2885.2719,-419.6434"/>
-<polygon fill="#191970" stroke="#191970" points="2888.4787,-421.0796 2890.2998,-410.6424 2882.3675,-417.6659 2888.4787,-421.0796"/>
+<path fill="none" stroke="#191970" d="M2174.9103,-622.2246C2282.8289,-616.4608 2508.9091,-601.8004 2536,-579 2583.0602,-539.3929 2589.7484,-460.8418 2589.8628,-420.5776"/>
+<polygon fill="#191970" stroke="#191970" points="2593.3622,-420.4895 2589.7546,-410.5278 2586.3626,-420.565 2593.3622,-420.4895"/>
</g>
<!-- Node76->Node73 -->
-<g id="edge181" class="edge">
+<g id="edge182" class="edge">
<title>Node76->Node73</title>
-<path fill="none" stroke="#191970" d="M1548.1878,-615.2455C1531.1527,-606.4125 1505.8425,-593.2887 1486.5319,-583.2758"/>
-<polygon fill="#191970" stroke="#191970" points="1488.0853,-580.1388 1477.5966,-578.6427 1484.863,-586.353 1488.0853,-580.1388"/>
+<path fill="none" stroke="#191970" d="M2102.4974,-615.2455C2088.842,-606.6534 2068.7343,-594.0014 2053.0011,-584.1018"/>
+<polygon fill="#191970" stroke="#191970" points="2054.6529,-581.006 2044.325,-578.6427 2050.9249,-586.9307 2054.6529,-581.006"/>
</g>
<!-- Node77 -->
<g id="node49" class="node">
<title>Node77</title>
<g id="a_node49"><a xlink:href="memory__pools_8h.html" target="_top" xlink:title="The object definition for relay.build argument type of memory pools. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1498.5,-503.5 1498.5,-522.5 1629.5,-522.5 1629.5,-503.5 1498.5,-503.5"/>
-<text text-anchor="middle" x="1564" y="-510.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/memory_pools.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="2165.5,-503.5 2165.5,-522.5 2296.5,-522.5 2296.5,-503.5 2165.5,-503.5"/>
+<text text-anchor="middle" x="2231" y="-510.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/ir/memory_pools.h</text>
</a>
</g>
</g>
<!-- Node76->Node77 -->
-<g id="edge172" class="edge">
+<g id="edge173" class="edge">
<title>Node76->Node77</title>
-<path fill="none" stroke="#191970" d="M1566.7442,-615.4509C1566.2549,-597.184 1565.1774,-556.9553 1564.5276,-532.6976"/>
-<polygon fill="#191970" stroke="#191970" points="1568.0217,-532.4276 1564.2551,-522.5249 1561.0242,-532.6151 1568.0217,-532.4276"/>
+<path fill="none" stroke="#191970" d="M2127.6343,-615.4509C2146.8149,-596.4401 2189.9948,-553.6423 2214.0338,-529.816"/>
+<polygon fill="#191970" stroke="#191970" points="2216.7515,-532.0503 2221.3901,-522.5249 2211.8238,-527.0786 2216.7515,-532.0503"/>
</g>
<!-- Node78 -->
<g id="node50" class="node">
<title>Node78</title>
<g id="a_node50"><a xlink:href="target_8h.html" target="_top" xlink:title="Compilation target object. ">
-<polygon fill="#ffffff" stroke="#ff0000" points="1520,-447.5 1520,-466.5 1630,-466.5 1630,-447.5 1520,-447.5"/>
-<text text-anchor="middle" x="1575" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/target/target.h</text>
+<polygon fill="#ffffff" stroke="#ff0000" points="2055,-447.5 2055,-466.5 2165,-466.5 2165,-447.5 2055,-447.5"/>
+<text text-anchor="middle" x="2110" y="-454.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">tvm/target/target.h</text>
</a>
</g>
</g>
<!-- Node76->Node78 -->
-<g id="edge180" class="edge">
+<g id="edge181" class="edge">
<title>Node76->Node78</title>
-<path fill="none" stroke="#191970" d="M1576.9321,-615.4544C1593.6209,-598.6435 1626.6724,-561.8409 1639,-523 1641.689,-514.5276 1643.0557,-510.9097 1639,-503 1632.067,-489.4789 1619.1064,-479.0326 1606.6519,-471.5377"/>
-<polygon fill="#191970" stroke="#191970" points="1608.2072,-468.3979 1597.766,-466.6001 1604.8072,-474.5168 1608.2072,-468.3979"/>
+<path fill="none" stroke="#191970" d="M2117.547,-615.4862C2116.2823,-588.9293 2112.6936,-513.5651 2110.9603,-477.1655"/>
+<polygon fill="#191970" stroke="#191970" points="2114.4398,-476.6501 2110.468,-466.8279 2107.4477,-476.9831 2114.4398,-476.6501"/>
</g>
<!-- Node77->Node78 -->
-<g id="edge173" class="edge">
+<g id="edge174" class="edge">
<title>Node77->Node78</title>
-<path fill="none" stroke="#191970" d="M1565.9161,-503.2455C1567.3514,-495.9382 1569.3636,-485.6944 1571.1295,-476.7046"/>
-<polygon fill="#191970" stroke="#191970" points="1574.6127,-477.1298 1573.1059,-466.6427 1567.744,-475.7805 1574.6127,-477.1298"/>
+<path fill="none" stroke="#191970" d="M2210.1994,-503.3733C2190.7757,-494.3838 2161.5776,-480.8706 2139.7036,-470.7471"/>
+<polygon fill="#191970" stroke="#191970" points="2141.0745,-467.5249 2130.5292,-466.5011 2138.1344,-473.8776 2141.0745,-467.5249"/>
</g>
<!-- Node78->Node3 -->
-<g id="edge174" class="edge">
+<g id="edge175" class="edge">
<title>Node78->Node3</title>
-<path fill="none" stroke="#191970" d="M1519.7707,-451.1031C1407.9078,-439.1594 1156.1073,-412.2745 1048.8772,-400.8254"/>
-<polygon fill="#191970" stroke="#191970" points="1048.9715,-397.3157 1038.6564,-399.7342 1048.2283,-404.2761 1048.9715,-397.3157"/>
+<path fill="none" stroke="#191970" d="M2054.7629,-453.6498C1876.2443,-442.8225 1316.5273,-408.8752 1145.8828,-398.5254"/>
+<polygon fill="#191970" stroke="#191970" points="1145.8766,-395.0187 1135.683,-397.9068 1145.4528,-402.0059 1145.8766,-395.0187"/>
</g>
<!-- Node78->Node5 -->
-<g id="edge175" class="edge">
+<g id="edge176" class="edge">
<title>Node78->Node5</title>
-<path fill="none" stroke="#191970" d="M1554.8833,-447.3462C1471.7821,-407.4667 1155.027,-255.459 1057.0533,-208.4424"/>
-<polygon fill="#191970" stroke="#191970" points="1058.5402,-205.2738 1048.0103,-204.1027 1055.5116,-211.5848 1058.5402,-205.2738"/>
+<path fill="none" stroke="#191970" d="M2105.8869,-447.2323C2097.935,-427.9032 2080.4899,-383.2456 2072,-344 2062.7429,-301.2081 2088.024,-281.2078 2062,-246 2047.0533,-225.7786 2022.6087,-213.4355 1999.5103,-205.9299"/>
+<polygon fill="#191970" stroke="#191970" points="2000.3088,-202.5155 1989.7251,-203.0005 1998.3013,-209.2215 2000.3088,-202.5155"/>
</g>
<!-- Node78->Node16 -->
-<g id="edge177" class="edge">
+<g id="edge178" class="edge">
<title>Node78->Node16</title>
-<path fill="none" stroke="#191970" d="M1519.9652,-454.1792C1331.6561,-444.0448 714.4246,-406.7152 527,-344 434.1024,-312.915 382.4174,-303.1183 352,-210 349.0319,-200.9135 349.2167,-137.7942 353,-123 356.0014,-111.2634 362.1216,-99.214 367.7773,-89.7238"/>
-<polygon fill="#191970" stroke="#191970" points="370.7817,-91.5209 373.1453,-81.1931 364.8571,-87.7927 370.7817,-91.5209"/>
+<path fill="none" stroke="#191970" d="M2054.833,-453.4712C1912.1589,-443.6306 1519.8015,-412.0237 1200,-344 1100.025,-322.7347 1072.5455,-320.3882 980,-277 930.8907,-253.976 910.6806,-253.2874 878,-210 851.3221,-174.6635 849.3697,-120.0713 850.5165,-91.1085"/>
+<polygon fill="#191970" stroke="#191970" points="854.0159,-91.2162 851.0865,-81.0345 847.0271,-90.8208 854.0159,-91.2162"/>
</g>
<!-- Node78->Node20 -->
-<g id="edge178" class="edge">
+<g id="edge179" class="edge">
<title>Node78->Node20</title>
-<path fill="none" stroke="#191970" d="M1579.0527,-447.1981C1587.4556,-425.959 1606,-373.9391 1606,-328.5 1606,-328.5 1606,-328.5 1606,-261.5 1606,-223.0012 1606,-178.0145 1606,-152.7812"/>
-<polygon fill="#191970" stroke="#191970" points="1609.5001,-152.6718 1606,-142.6719 1602.5001,-152.6719 1609.5001,-152.6718"/>
+<path fill="none" stroke="#191970" d="M2110.6476,-447.345C2112.5932,-423.3315 2119.8364,-359.4986 2144,-313 2153.8713,-294.0045 2168.3923,-297.0098 2176,-277 2180.8964,-264.1216 2179.456,-259.3373 2176,-246 2171.4088,-228.2818 2161.725,-227.385 2156,-210 2149.8249,-191.2482 2147.4424,-168.7512 2146.534,-152.9311"/>
+<polygon fill="#191970" stroke="#191970" points="2150.0225,-152.5826 2146.1,-142.7407 2143.0288,-152.8806 2150.0225,-152.5826"/>
</g>
<!-- Node78->Node52 -->
-<g id="edge176" class="edge">
+<g id="edge177" class="edge">
<title>Node78->Node52</title>
-<path fill="none" stroke="#191970" d="M1630.2168,-456.7049C1854.7159,-455.2681 2699.5934,-447.4986 2964,-411 2970.4437,-410.1105 2977.1861,-408.8336 2983.7647,-407.3892"/>
-<polygon fill="#191970" stroke="#191970" points="2984.9636,-410.7032 2993.901,-405.0134 2983.3661,-403.8879 2984.9636,-410.7032"/>
+<path fill="none" stroke="#191970" d="M2122.9688,-447.3906C2136.0034,-437.7324 2156.3164,-422.6812 2171.7155,-411.271"/>
+<polygon fill="#191970" stroke="#191970" points="2174.1713,-413.8075 2180.1223,-405.0419 2170.0039,-408.1832 2174.1713,-413.8075"/>
</g>
</g>
</svg>
diff --git a/docs/reference/api/doxygen/ir_2attrs_8h__dep__incl.svg b/docs/reference/api/doxygen/ir_2attrs_8h__dep__incl.svg
index 81b7ce069..577b0c086 100644
--- a/docs/reference/api/doxygen/ir_2attrs_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/ir_2attrs_8h__dep__incl.svg
@@ -184,9 +184,9 @@
<path fill="none" stroke="#191970" d="M1870.9795,-519.1184C1796.5025,-512.8656 1684.4893,-500.1037 1648,-478 1611.582,-455.9395 1591,-443.5786 1591,-401 1591,-401 1591,-401 1591,-339.5 1591,-291.5821 1570.8584,-236.7488 1562.0464,-215.0522"/>
<polygon fill="#191970" stroke="#191970" points="1870.9753,-522.6297 1881.2277,-519.9578 1871.5468,-515.6531 1870.9753,-522.6297"/>
</g>
-<!-- Node181 -->
+<!-- Node182 -->
<g id="node38" class="node">
-<title>Node181</title>
+<title>Node182</title>
<g id="a_node38"><a xlink:href="annotation_8h.html" target="_top" xlink:title="Attribute for annotation operators. ">
<polygon fill="#ffffff" stroke="#000000" points="1804,-447.5 1804,-477.5 1932,-477.5 1932,-447.5 1804,-447.5"/>
<text text-anchor="start" x="1812" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -194,15 +194,15 @@
</a>
</g>
</g>
-<!-- Node48->Node181 -->
+<!-- Node48->Node182 -->
<g id="edge77" class="edge">
-<title>Node48->Node181</title>
+<title>Node48->Node182</title>
<path fill="none" stroke="#191970" d="M1921.7276,-507.7637C1910.6106,-498.398 1896.7268,-486.7014 1885.8578,-477.5446"/>
<polygon fill="#191970" stroke="#191970" points="1919.6909,-510.6243 1929.5938,-514.3906 1924.201,-505.2709 1919.6909,-510.6243"/>
</g>
-<!-- Node182 -->
+<!-- Node183 -->
<g id="node39" class="node">
-<title>Node182</title>
+<title>Node183</title>
<g id="a_node39"><a xlink:href="bitserial_8h.html" target="_top" xlink:title="Auxiliary attributes for bitserial operators. ">
<polygon fill="#ffffff" stroke="#000000" points="1950,-447.5 1950,-477.5 2078,-477.5 2078,-447.5 1950,-447.5"/>
<text text-anchor="start" x="1958" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -210,15 +210,15 @@
</a>
</g>
</g>
-<!-- Node48->Node182 -->
+<!-- Node48->Node183 -->
<g id="edge78" class="edge">
-<title>Node48->Node182</title>
+<title>Node48->Node183</title>
<path fill="none" stroke="#191970" d="M1960.2724,-507.7637C1971.3894,-498.398 1985.2732,-486.7014 1996.1422,-477.5446"/>
<polygon fill="#191970" stroke="#191970" points="1957.799,-505.2709 1952.4063,-514.3906 1962.3091,-510.6243 1957.799,-505.2709"/>
</g>
-<!-- Node183 -->
+<!-- Node184 -->
<g id="node40" class="node">
-<title>Node183</title>
+<title>Node184</title>
<g id="a_node40"><a xlink:href="call_8h.html" target="_top" xlink:title="Attribute for call_lowered operator. ">
<polygon fill="#ffffff" stroke="#000000" points="2096,-447.5 2096,-477.5 2224,-477.5 2224,-447.5 2096,-447.5"/>
<text text-anchor="start" x="2104" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -226,15 +226,15 @@
</a>
</g>
</g>
-<!-- Node48->Node183 -->
+<!-- Node48->Node184 -->
<g id="edge79" class="edge">
-<title>Node48->Node183</title>
+<title>Node48->Node184</title>
<path fill="none" stroke="#191970" d="M1985.147,-511.6025C2020.1433,-501.7748 2069.1842,-488.0031 2106.4266,-477.5446"/>
<polygon fill="#191970" stroke="#191970" points="1983.9,-508.3173 1975.2188,-514.3906 1985.7926,-515.0566 1983.9,-508.3173"/>
</g>
-<!-- Node184 -->
+<!-- Node185 -->
<g id="node41" class="node">
-<title>Node184</title>
+<title>Node185</title>
<g id="a_node41"><a xlink:href="relay_2attrs_2debug_8h.html" target="_top" xlink:title="Auxiliary attributes for debug operators. ">
<polygon fill="#ffffff" stroke="#000000" points="2242,-447.5 2242,-477.5 2370,-477.5 2370,-447.5 2242,-447.5"/>
<text text-anchor="start" x="2250" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -242,9 +242,9 @@
</a>
</g>
</g>
-<!-- Node48->Node184 -->
+<!-- Node48->Node185 -->
<g id="edge80" class="edge">
-<title>Node48->Node184</title>
+<title>Node48->Node185</title>
<path fill="none" stroke="#191970" d="M2010.7144,-514.0782C2069.983,-505.3895 2157.2677,-491.9951 2233,-478 2235.9303,-477.4585 2238.9233,-476.8892 2241.9441,-476.3014"/>
<polygon fill="#191970" stroke="#191970" points="2010.1063,-510.6298 2000.717,-515.5381 2011.1179,-517.5563 2010.1063,-510.6298"/>
</g>
@@ -264,9 +264,9 @@
<path fill="none" stroke="#191970" d="M2010.7495,-518.9384C2097.4228,-512.1291 2249.6936,-498.4603 2379,-478 2381.9433,-477.5343 2384.947,-477.0271 2387.9764,-476.4897"/>
<polygon fill="#191970" stroke="#191970" points="2010.4666,-515.4498 2000.7683,-519.715 2011.0096,-522.4287 2010.4666,-515.4498"/>
</g>
-<!-- Node185 -->
+<!-- Node186 -->
<g id="node43" class="node">
-<title>Node185</title>
+<title>Node186</title>
<g id="a_node43"><a xlink:href="image_8h.html" target="_top" xlink:title="Auxiliary attributes for image operators. ">
<polygon fill="#ffffff" stroke="#000000" points="2534,-447.5 2534,-477.5 2662,-477.5 2662,-447.5 2534,-447.5"/>
<text text-anchor="start" x="2542" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -274,24 +274,24 @@
</a>
</g>
</g>
-<!-- Node48->Node185 -->
+<!-- Node48->Node186 -->
<g id="edge82" class="edge">
-<title>Node48->Node185</title>
+<title>Node48->Node186</title>
<path fill="none" stroke="#191970" d="M2010.8422,-521.449C2120.7227,-516.7731 2340.4695,-504.7412 2525,-478 2527.9491,-477.5726 2530.9575,-477.0969 2533.9909,-476.5851"/>
<polygon fill="#191970" stroke="#191970" points="2010.4016,-517.9643 2000.5564,-521.8788 2010.6939,-524.9582 2010.4016,-517.9643"/>
</g>
-<!-- Node186 -->
+<!-- Node187 -->
<g id="node44" class="node">
-<title>Node186</title>
+<title>Node187</title>
<g id="a_node44"><a xlink:href="relay_2attrs_2nn_8h.html" target="_top" xlink:title="Auxiliary attributes for nn operators. ">
<polygon fill="#ffffff" stroke="#000000" points="2680,-453 2680,-472 2832,-472 2832,-453 2680,-453"/>
<text text-anchor="middle" x="2756" y="-460" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs/nn.h</text>
</a>
</g>
</g>
-<!-- Node48->Node186 -->
+<!-- Node48->Node187 -->
<g id="edge84" class="edge">
-<title>Node48->Node186</title>
+<title>Node48->Node187</title>
<path fill="none" stroke="#191970" d="M2011.0711,-522.3206C2141.4223,-518.5898 2429.8445,-507.4823 2671,-478 2683.5071,-476.4709 2696.9581,-474.2765 2709.3868,-472.018"/>
<polygon fill="#191970" stroke="#191970" points="2010.7307,-518.8286 2000.8328,-522.6076 2010.927,-525.8259 2010.7307,-518.8286"/>
</g>
@@ -311,9 +311,9 @@
<path fill="none" stroke="#191970" d="M2010.7193,-521.8649C2193.0112,-516.0535 2680.7543,-498.9736 2841,-478 2843.8262,-477.6301 2846.7059,-477.2093 2849.6088,-476.7491"/>
<polygon fill="#191970" stroke="#191970" points="2010.5961,-518.367 2000.7121,-522.1824 2010.8181,-525.3635 2010.5961,-518.367"/>
</g>
-<!-- Node187 -->
+<!-- Node188 -->
<g id="node46" class="node">
-<title>Node187</title>
+<title>Node188</title>
<g id="a_node46"><a xlink:href="random_8h.html" target="_top" xlink:title="include/tvm/relay/attrs\l/random.h">
<polygon fill="#ffffff" stroke="#000000" points="2996,-447.5 2996,-477.5 3124,-477.5 3124,-447.5 2996,-447.5"/>
<text text-anchor="start" x="3004" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -321,15 +321,15 @@
</a>
</g>
</g>
-<!-- Node48->Node187 -->
+<!-- Node48->Node188 -->
<g id="edge86" class="edge">
-<title>Node48->Node187</title>
+<title>Node48->Node188</title>
<path fill="none" stroke="#191970" d="M2010.8497,-522.4355C2213.1886,-517.664 2797.3567,-502.1351 2987,-478 2989.8275,-477.6402 2992.7082,-477.2277 2995.612,-476.7743"/>
<polygon fill="#191970" stroke="#191970" points="2010.5611,-518.9412 2000.6459,-522.6746 2010.7252,-525.9392 2010.5611,-518.9412"/>
</g>
-<!-- Node188 -->
+<!-- Node189 -->
<g id="node47" class="node">
-<title>Node188</title>
+<title>Node189</title>
<g id="a_node47"><a xlink:href="reduce_8h.html" target="_top" xlink:title="Auxiliary attributes for reduce operators. ">
<polygon fill="#ffffff" stroke="#000000" points="3142,-447.5 3142,-477.5 3270,-477.5 3270,-447.5 3142,-447.5"/>
<text text-anchor="start" x="3150" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -337,15 +337,15 @@
</a>
</g>
</g>
-<!-- Node48->Node188 -->
+<!-- Node48->Node189 -->
<g id="edge87" class="edge">
-<title>Node48->Node188</title>
+<title>Node48->Node189</title>
<path fill="none" stroke="#191970" d="M2010.9356,-522.8692C2231.9603,-519.0481 2913.7575,-505.3032 3133,-478 3135.8285,-477.6478 3138.71,-477.2416 3141.6144,-476.7933"/>
<polygon fill="#191970" stroke="#191970" points="2010.701,-519.3726 2000.7625,-523.0436 2010.8211,-526.3715 2010.701,-519.3726"/>
</g>
-<!-- Node189 -->
+<!-- Node190 -->
<g id="node48" class="node">
-<title>Node189</title>
+<title>Node190</title>
<g id="a_node48"><a xlink:href="vision_8h.html" target="_top" xlink:title="Auxiliary attributes for vision operators. ">
<polygon fill="#ffffff" stroke="#000000" points="3288,-447.5 3288,-477.5 3416,-477.5 3416,-447.5 3288,-447.5"/>
<text text-anchor="start" x="3296" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs</text>
@@ -353,30 +353,30 @@
</a>
</g>
</g>
-<!-- Node48->Node189 -->
+<!-- Node48->Node190 -->
<g id="edge89" class="edge">
-<title>Node48->Node189</title>
+<title>Node48->Node190</title>
<path fill="none" stroke="#191970" d="M2010.8214,-523.2121C2249.2089,-520.2693 3029.9342,-508.4856 3279,-478 3281.8292,-477.6537 3284.7113,-477.2525 3287.6163,-476.8083"/>
<polygon fill="#191970" stroke="#191970" points="2010.5092,-519.7155 2000.5527,-523.3375 2010.5948,-526.715 2010.5092,-519.7155"/>
</g>
-<!-- Node190 -->
+<!-- Node191 -->
<g id="node49" class="node">
-<title>Node190</title>
+<title>Node191</title>
<g id="a_node49"><a xlink:href="relay_2attrs_2vm_8h.html" target="_top" xlink:title="Attributes for Relay vm operators. ">
<polygon fill="#ffffff" stroke="#000000" points="3434.5,-453 3434.5,-472 3589.5,-472 3589.5,-453 3434.5,-453"/>
<text text-anchor="middle" x="3512" y="-460" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/attrs/vm.h</text>
</a>
</g>
</g>
-<!-- Node48->Node190 -->
+<!-- Node48->Node191 -->
<g id="edge90" class="edge">
-<title>Node48->Node190</title>
+<title>Node48->Node191</title>
<path fill="none" stroke="#191970" d="M2010.99,-523.1808C2266.0751,-519.9847 3145.5741,-507.0664 3425,-478 3438.2759,-476.619 3452.5763,-474.3865 3465.6716,-472.0408"/>
<polygon fill="#191970" stroke="#191970" points="2010.7014,-519.684 2000.7457,-523.3081 2010.7885,-526.6835 2010.7014,-519.684"/>
</g>
-<!-- Node191 -->
+<!-- Node192 -->
<g id="node50" class="node">
-<title>Node191</title>
+<title>Node192</title>
<g id="a_node50"><a xlink:href="relay_2qnn_2attrs_8h.html" target="_top" xlink:title="Auxiliary attributes for qnn operators. ">
<polygon fill="#ffffff" stroke="#000000" points="3607.5,-447.5 3607.5,-477.5 3730.5,-477.5 3730.5,-447.5 3607.5,-447.5"/>
<text text-anchor="start" x="3615.5" y="-465.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/relay/qnn</text>
@@ -384,9 +384,9 @@
</a>
</g>
</g>
-<!-- Node48->Node191 -->
+<!-- Node48->Node192 -->
<g id="edge94" class="edge">
-<title>Node48->Node191</title>
+<title>Node48->Node192</title>
<path fill="none" stroke="#191970" d="M2010.8597,-523.8313C2284.575,-522.8961 3284.3911,-516.8373 3599,-478 3601.7177,-477.6645 3604.4853,-477.2722 3607.2742,-476.8353"/>
<polygon fill="#191970" stroke="#191970" points="2010.6525,-520.3319 2000.6641,-523.8649 2010.6756,-527.3318 2010.6525,-520.3319"/>
</g>
diff --git a/docs/reference/api/doxygen/ir_2expr_8h__dep__incl.svg b/docs/reference/api/doxygen/ir_2expr_8h__dep__incl.svg
index f4ad5eae3..5297f4532 100644
--- a/docs/reference/api/doxygen/ir_2expr_8h__dep__incl.svg
+++ b/docs/reference/api/doxygen/ir_2expr_8h__dep__incl.svg
@@ -291,9 +291,9 @@
<path fill="none" stroke="#191970" d="M1013.5882,-718.9503C1054.1257,-713.0913 1105.3151,-701.582 1146,-679 1179.7246,-660.2813 1208.7716,-623.5703 1221.3478,-606.1348"/>
<polygon fill="#191970" stroke="#191970" points="1012.9604,-715.5034 1003.5244,-720.3214 1013.9054,-722.4393 1012.9604,-715.5034"/>
</g>
-<!-- Node197 -->
+<!-- Node198 -->
<g id="node41" class="node">
-<title>Node197</title>
+<title>Node198</title>
<g id="a_node41"><a xlink:href="tensor__type_8h.html" target="_top" xlink:title="Polymorphic tensor types. ">
<polygon fill="#ffffff" stroke="#ff0000" points="1964,-648.5 1964,-678.5 2082,-678.5 2082,-648.5 1964,-648.5"/>
<text text-anchor="start" x="1972" y="-666.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/ir/tensor</text>
@@ -301,9 +301,9 @@
</a>
</g>
</g>
-<!-- Node46->Node197 -->
+<!-- Node46->Node198 -->
<g id="edge71" class="edge">
-<title>Node46->Node197</title>
+<title>Node46->Node198</title>
<path fill="none" stroke="#191970" d="M1013.787,-723.4689C1210.4216,-718.8375 1772.48,-703.7562 1955,-679 1957.8796,-678.6094 1960.817,-678.1558 1963.776,-677.6545"/>
<polygon fill="#191970" stroke="#191970" points="1013.5027,-719.9745 1003.5873,-723.7075 1013.6664,-726.9726 1013.5027,-719.9745"/>
</g>
@@ -323,9 +323,9 @@
<path fill="none" stroke="#191970" d="M1014.0032,-723.5013C1228.283,-718.6465 1880.4582,-702.3056 2091,-679 2093.8328,-678.6864 2096.7132,-678.334 2099.6188,-677.9503"/>
<polygon fill="#191970" stroke="#191970" points="1013.75,-720.006 1003.8315,-723.7305 1013.9078,-727.0042 1013.75,-720.006"/>
</g>
-<!-- Node198 -->
+<!-- Node199 -->
<g id="node43" class="node">
-<title>Node198</title>
+<title>Node199</title>
<g id="a_node43"><a xlink:href="metadata__base_8h.html" target="_top" xlink:title="Defines types which can be used in Metadata. ">
<polygon fill="#ffffff" stroke="#000000" points="2270,-648.5 2270,-678.5 2386,-678.5 2386,-648.5 2270,-648.5"/>
<text text-anchor="start" x="2278" y="-666.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/runtime</text>
@@ -333,15 +333,15 @@
</a>
</g>
</g>
-<!-- Node46->Node198 -->
+<!-- Node46->Node199 -->
<g id="edge76" class="edge">
-<title>Node46->Node198</title>
+<title>Node46->Node199</title>
<path fill="none" stroke="#191970" d="M1013.6944,-724.3436C1248.2308,-721.8292 2016.3155,-711.2736 2261,-679 2263.8408,-678.6253 2266.7378,-678.1846 2269.6555,-677.6936"/>
<polygon fill="#191970" stroke="#191970" points="1013.554,-720.8448 1003.5916,-724.4504 1013.6281,-727.8444 1013.554,-720.8448"/>
</g>
-<!-- Node199 -->
+<!-- Node200 -->
<g id="node44" class="node">
-<title>Node199</title>
+<title>Node200</title>
<g id="a_node44"><a xlink:href="target__info_8h.html" target="_top" xlink:title="Various information about target. ">
<polygon fill="#ffffff" stroke="#000000" points="2404.5,-648.5 2404.5,-678.5 2511.5,-678.5 2511.5,-648.5 2404.5,-648.5"/>
<text text-anchor="start" x="2412.5" y="-666.5" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/target</text>
@@ -349,39 +349,39 @@
</a>
</g>
</g>
-<!-- Node46->Node199 -->
+<!-- Node46->Node200 -->
<g id="edge78" class="edge">
-<title>Node46->Node199</title>
+<title>Node46->Node200</title>
<path fill="none" stroke="#191970" d="M1013.8814,-724.7266C1263.9189,-723.4364 2123.2029,-716.3261 2395,-679 2398.0129,-678.5862 2401.0911,-678.0873 2404.1872,-677.5259"/>
<polygon fill="#191970" stroke="#191970" points="1013.8205,-721.2268 1003.8382,-724.777 1013.8556,-728.2267 1013.8205,-721.2268"/>
</g>
-<!-- Node200 -->
+<!-- Node201 -->
<g id="node45" class="node">
-<title>Node200</title>
+<title>Node201</title>
<g id="a_node45"><a xlink:href="buffer_8h.html" target="_top" xlink:title="Symbolic n-dimensional array, to represent a memory buffer. ">
<polygon fill="#ffffff" stroke="#ff0000" points="280,-520 280,-539 408,-539 408,-520 280,-520"/>
<text text-anchor="middle" x="344" y="-527" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/buffer.h</text>
</a>
</g>
</g>
-<!-- Node46->Node200 -->
+<!-- Node46->Node201 -->
<g id="edge79" class="edge">
-<title>Node46->Node200</title>
+<title>Node46->Node201</title>
<path fill="none" stroke="#191970" d="M876.2008,-722.6067C742.8351,-717.4397 458.9374,-703.6685 422,-679 371.1187,-645.0192 351.3285,-566.878 345.7769,-539.331"/>
<polygon fill="#191970" stroke="#191970" points="876.1842,-726.1085 886.3106,-722.993 876.4516,-719.1136 876.1842,-726.1085"/>
</g>
-<!-- Node201 -->
+<!-- Node202 -->
<g id="node46" class="node">
-<title>Node201</title>
+<title>Node202</title>
<g id="a_node46"><a xlink:href="tir_2expr_8h.html" target="_top" xlink:title="TIR expressions. ">
<polygon fill="#ffffff" stroke="#ff0000" points="272.5,-453 272.5,-472 393.5,-472 393.5,-453 272.5,-453"/>
<text text-anchor="middle" x="333" y="-460" font-family="Helvetica,sans-Serif" font-size="10.00" fill="#000000">include/tvm/tir/expr.h</text>
</a>
</g>
</g>
-<!-- Node46->Node201 -->
+<!-- Node46->Node202 -->
<g id="edge91" class="edge">
-<title>Node46->Node201</title>
+<title>Node46->Node202</title>
<path fill="none" stroke="#191970" d="M876.3051,-722.2131C736.9581,-716.1431 430.6173,-700.5256 388,-679 317.4287,-643.3552 298.7692,-619.0252 271,-545 266.1608,-532.1 265.1462,-526.4724 271,-514 279.8584,-495.126 299.4082,-480.7763 314.2031,-472.0782"/>
<polygon fill="#191970" stroke="#191970" points="876.3044,-725.7163 886.4461,-722.6508 876.6063,-718.7228 876.3044,-725.7163"/>
</g>
... 23060 lines suppressed ...