Sparse CUDA: rename files .cu -> .cpp #63894

IvanYashchuk · 2021-08-24T22:06:24Z

Stack from ghstack:

Sparse CSR CUDA: add torch.add with all inputs sparse #63948 Sparse CSR CUDA: add torch.add with all inputs sparse
Sparse CSR CUDA: add torch.addmm with all inputs sparse #63511 Sparse CSR CUDA: add torch.addmm with all inputs sparse
Sparse CSR: Add tensor.resize_ and tensor.copy_ #63510 Sparse CSR: Add tensor.resize_ and tensor.copy_
Sparse CSR: Add torch.empty #63509 Sparse CSR: Add torch.empty
Sparse CSR CPU: add triangular_solve_out #62180 Sparse CSR CPU: add triangular_solve_out
Sparse CSR CUDA: add triangular_solve_out #61858 Sparse CSR CUDA: add triangular_solve_out
Port triangular_solve to structured kernel #61857 Port triangular_solve to structured kernel
Sparse CSR CPU: add addmv_out #61536 Sparse CSR CPU: add addmv_out
Sparse CSR: add test parametrization for enabling sparse_csr #61408 Sparse CSR: add test parametrization for enabling sparse_csr
Sparse CSR CUDA: add addmv_out #61407 Sparse CSR CUDA: add addmv_out
Add cuSPARSE descriptors and update CSR addmm #60838 Add cuSPARSE descriptors and update CSR addmm
**Sparse CUDA: rename files *.cu -> *.cpp #63894 Sparse CUDA: rename files .cu -> .cpp

This PR introduces a few code structure changes. There is no need to use
.cu extension for pure c++ code without cuda. Moved
s_addmm_out_csr_sparse_dense_cuda_worker to a separate cpp file from
cu file.

cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel

Differential Revision: D30548771

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. [ghstack-poisoned]

facebook-github-bot · 2021-08-24T22:06:30Z

🔗 Helpful links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/63894
📄 Preview docs built from this PR

💊 CI failures summary and remediations

As of commit 76f3ee0 (more details on the Dr. CI page):

1/1 failures introduced in this PR

🕵️ 1 new failure recognized by patterns

The following CI failures do not appear to be due to upstream breakages:

pytorch_linux_xenial_py3_6_gcc5_4_build (1/1)

Step: "Build" (full log | diagnosis details | 🔁 rerun)

Aug 26 22:17:17 CMake Error at CMakeLists.txt:1 (cmake_minimum_required):

Aug 26 22:17:17 invalid command 'bad_argument'
Aug 26 22:17:17 + retcode=1
Aug 26 22:17:17 + set -e
Aug 26 22:17:17 + return 1
Aug 26 22:17:17 + [[ pytorch-linux-xenial-py3.6-gcc5.4-build != *libtorch* ]]
Aug 26 22:17:17 + [[ pytorch-linux-xenial-py3.6-gcc5.4-build != *ppc64le* ]]
Aug 26 22:17:17 + [[ pytorch-linux-xenial-py3.6-gcc5.4-build != *clang* ]]
Aug 26 22:17:17 + WERROR=1
Aug 26 22:17:17 + python setup.py bdist_wheel
Aug 26 22:17:17 Building wheel torch-1.10.0a0+gitf99bfe2
Aug 26 22:17:17 CMake Error at CMakeLists.txt:1 (cmake_minimum_required):
Aug 26 22:17:17   CMake 3.10 or higher is required.  You are running version 3.5.1
Aug 26 22:17:17 
Aug 26 22:17:17 
Aug 26 22:17:17 -- Configuring incomplete, errors occurred!
Aug 26 22:17:17 -- Building version 1.10.0a0+gitf99bfe2
Aug 26 22:17:17 cmake -DBUILD_ENVIRONMENT=pytorch-linux-xenial-py3.6-gcc5.4-build -DBUILD_PYTHON=True -DBUILD_STATIC_RUNTIME_BENCHMARK=ON -DBUILD_TEST=True -DCMAKE_BUILD_TYPE=Release -DCMAKE_INSTALL_PREFIX=/var/lib/jenkins/workspace/torch -DCMAKE_PREFIX_PATH=/opt/conda/lib/python3.6/site-packages;/opt/conda -DNUMPY_INCLUDE_DIR=/opt/conda/lib/python3.6/site-packages/numpy/core/include -DOPENSSL_ROOT_DIR=/opt/openssl -DPYTHON_EXECUTABLE=/opt/conda/bin/python -DPYTHON_INCLUDE_DIR=/opt/conda/include/python3.6m -DPYTHON_LIBRARY=/opt/conda/lib/libpython3.6m.so.1.0 -DTORCH_BUILD_VERSION=1.10.0a0+gitf99bfe2 -DUSE_GLOO_WITH_OPENSSL=ON -DUSE_LLVM=/opt/llvm -DUSE_NUMPY=True -DWERROR=1 /var/lib/jenkins/workspace
Aug 26 22:17:17 =================== sccache compilation log ===================
Aug 26 22:17:17 + cleanup
Aug 26 22:17:17 + retcode=1
Aug 26 22:17:17 + set +x

This comment was automatically generated by Dr. CI (expand for details).

Follow this link to opt-out of these comments for your Pull Requests.

Please report bugs/suggestions to the (internal) Dr. CI Users group.

Click here to manually regenerate this comment.

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel [ghstack-poisoned]

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. ghstack-source-id: 9c933ff Pull Request resolved: pytorch#63894

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel [ghstack-poisoned]

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. ghstack-source-id: 92647d7 Pull Request resolved: pytorch#63894

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel [ghstack-poisoned]

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel [ghstack-poisoned]

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. ghstack-source-id: 36ab499 Pull Request resolved: pytorch#63894

aten/src/ATen/native/sparse/cuda/SparseBlasLegacy.cpp

cpuhrsch

Please address the comment, will merge after.

cpuhrsch · 2021-08-25T18:36:20Z

@cpuhrsch has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

cpuhrsch · 2021-08-25T18:39:23Z

cuda11.1 test failures are legit. https://app.circleci.com/pipelines/github/pytorch/pytorch/370358/workflows/e71238b3-183a-41c4-ac17-f233abcff26c/jobs/15625496

You'll need to add a decorator that restricts running those tests to older cuda versions.

ngimel · 2021-08-25T18:48:39Z

Wait, but it's not a question of decorators, we need some alternative csrmm2 implementation for cuda 11 (or restructure other functions so that they don't call csrmm2)?

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel Differential Revision: [D30548771](https://our.internmc.facebook.com/intern/diff/D30548771) [ghstack-poisoned]

IvanYashchuk · 2021-08-26T13:25:16Z

ATen/native/sparse/cuda/SparseCUDABlas.cu file had if defined(__CUDACC__) that after renaming to .cpp file became false enabling the codepath with deleted from CUDA 11 functions. Fixed by

 - #if defined(__CUDACC__)
 + #if defined(CUDART_VERSION)

@nikitaved

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. cc @nikitaved @pearu @cpuhrsch @IvanYashchuk @ngimel Differential Revision: [D30548771](https://our.internmc.facebook.com/intern/diff/D30548771) [ghstack-poisoned]

IvanYashchuk · 2021-08-26T13:49:45Z

@cpuhrsch, @ngimel, it should be safe now to import this PR.

cpuhrsch · 2021-08-26T17:08:38Z

@cpuhrsch has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. ghstack-source-id: db71374 Pull Request resolved: pytorch#63894

facebook-github-bot · 2021-08-27T20:24:08Z

@cpuhrsch merged this pull request in ff4569a.

Sparse CUDA: rename files *.cu -> *.cpp

6b1c1e6

This PR introduces a few code structure changes. There is no need to use .cu extension for pure c++ code without cuda. Moved `s_addmm_out_csr_sparse_dense_cuda_worker` to a separate cpp file from cu file. [ghstack-poisoned]

facebook-github-bot added the cla signed label Aug 24, 2021

IvanYashchuk added module: cuda Related to torch.cuda, and CUDA support in general module: sparse Related to torch.sparse labels Aug 24, 2021

IvanYashchuk marked this pull request as ready for review August 24, 2021 22:08

IvanYashchuk requested a review from cpuhrsch August 24, 2021 22:08

pytorchbot added the open source label Aug 24, 2021

IvanYashchuk mentioned this pull request Aug 25, 2021

Sparse CSR CUDA: add torch.add with all inputs sparse #63948

Closed

cpuhrsch reviewed Aug 25, 2021

View reviewed changes

aten/src/ATen/native/sparse/cuda/SparseBlasLegacy.cpp Show resolved Hide resolved

cpuhrsch approved these changes Aug 25, 2021

View reviewed changes

malfet approved these changes Aug 25, 2021

View reviewed changes

facebook-github-bot closed this in ff4569a Aug 27, 2021

facebook-github-bot added the Merged label Aug 27, 2021

facebook-github-bot deleted the gh/ivanyashchuk/36/head branch August 31, 2021 14:16

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Sparse CUDA: rename files .cu -> .cpp #63894

Sparse CUDA: rename files .cu -> .cpp #63894

Uh oh!

IvanYashchuk commented Aug 24, 2021 •

edited by cpuhrsch

Loading

Uh oh!

facebook-github-bot commented Aug 24, 2021 •

edited

Loading

Uh oh!

Uh oh!

cpuhrsch left a comment

Uh oh!

cpuhrsch commented Aug 25, 2021

Uh oh!

cpuhrsch commented Aug 25, 2021

Uh oh!

ngimel commented Aug 25, 2021

Uh oh!

IvanYashchuk commented Aug 26, 2021

Uh oh!

IvanYashchuk commented Aug 26, 2021

Uh oh!

cpuhrsch commented Aug 26, 2021

Uh oh!

facebook-github-bot commented Aug 27, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Sparse CUDA: rename files *.cu -> *.cpp #63894

Sparse CUDA: rename files *.cu -> *.cpp #63894

Uh oh!

Conversation

IvanYashchuk commented Aug 24, 2021 • edited by cpuhrsch Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

facebook-github-bot commented Aug 24, 2021 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful links

💊 CI failures summary and remediations

🕵️ 1 new failure recognized by patterns

pytorch_linux_xenial_py3_6_gcc5_4_build (1/1)

Uh oh!

Uh oh!

cpuhrsch left a comment

Choose a reason for hiding this comment

Uh oh!

cpuhrsch commented Aug 25, 2021

Uh oh!

cpuhrsch commented Aug 25, 2021

Uh oh!

ngimel commented Aug 25, 2021

Uh oh!

IvanYashchuk commented Aug 26, 2021

Uh oh!

IvanYashchuk commented Aug 26, 2021

Uh oh!

cpuhrsch commented Aug 26, 2021

Uh oh!

facebook-github-bot commented Aug 27, 2021

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

6 participants

Sparse CUDA: rename files .cu -> .cpp #63894

Sparse CUDA: rename files .cu -> .cpp #63894

IvanYashchuk commented Aug 24, 2021 •

edited by cpuhrsch

Loading

facebook-github-bot commented Aug 24, 2021 •

edited

Loading