[ATen][CPU][Sparse] Use Third-Party Eigen for sparse add and addmm by Aidyn-A · Pull Request #155357 · pytorch/pytorch

Aidyn-A · 2025-06-06T18:54:35Z

This pull request adds the following ops for sparse matrices using Eigen library:

    add(a_csr, b_csr)
    add(a_csc, b_csc)

    addmm(c_csr, a_csr, b_csr)
    addmm(c_csr, a_csr, b_csc)
    addmm(c_csr, a_csc, b_csc)
    addmm(c_csr, a_csc, b_csr)

    addmm(c_csc, a_csr, b_csr)
    addmm(c_csc, a_csr, b_csc)
    addmm(c_csc, a_csc, b_csc)
    addmm(c_csc, a_csc, b_csr)

Currently, the operations for sparse matrices on CPU are available through MKL only. The non-existence of MKL on aarch64 causes the unavailability of these ops on any machines with ARM based CPUs, including Apple Silicon, AWS Graviton and NVIDIA Grace. This PR addresses this issue by using Eigen as a backend for the above ops.

This is a re-factored version of my previous PR #101814. The main difference with the old one, this does not enable Eigen by default.

pytorch-bot · 2025-06-06T18:54:39Z

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/155357

📄 Preview Python docs built from this PR
📄 Preview C++ docs built from this PR
❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit c77592f with merge base 01bcf9a ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

Aidyn-A · 2025-06-07T06:18:28Z

@pytorchbot rebase

pytorchmergebot · 2025-06-07T06:20:08Z

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

pytorchmergebot · 2025-06-07T06:20:11Z

Successfully rebased eigen_for_sparse_addmm_v2 onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via git checkout eigen_for_sparse_addmm_v2 && git pull --rebase)

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp

pearu

Thanks for the PR, it looks good to me!

I have a number of nits and one question, see #155357 (comment)

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp

Aidyn-A · 2025-07-23T23:50:25Z

@pytorchbot rebase

pytorchmergebot · 2025-07-23T23:51:56Z

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

pytorchmergebot · 2025-07-23T23:52:01Z

Successfully rebased eigen_for_sparse_addmm_v2 onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via git checkout eigen_for_sparse_addmm_v2 && git pull --rebase)

pearu

Here is another review iteration that removes duplicated code.

Question: there exists multiple paths for addmm depending on the layout combinations of input and output tensors. Are all these paths covered by existing unittests?

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp

pearu

This looks good to me, however, see my comment below about the suspicion that there exists a bug in addmm_out_sparse_eigen logic that should be revealed in tests.

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp

seemethere · 2025-08-21T22:36:06Z

@pytorchbot revert -c ghfirst -m "This is causing buck builds to fail since we didn't add the definition of AT_USE_EIGEN_SPARSE in the buckbuild.bzl file, will follow-up and re-land this."

pytorch-bot · 2025-08-21T22:36:08Z

❌ 🤖 pytorchbot command failed:

Got EOF while in a quoted string```
Try `@pytorchbot --help` for more info.

seemethere · 2025-08-21T22:36:59Z

@pytorchbot revert -c ghfirst -m "This is causing buck builds to fail since we didn't add the definition of AT_USE_EIGEN_SPARSE in the buckbuild.bzl file, will follow-up and re-land this."

pytorchmergebot · 2025-08-21T22:38:30Z

@pytorchbot successfully started a revert job. Check the current status here.
Questions? Feedback? Please reach out to the PyTorch DevX Team

…addmm (#155357)" This reverts commit ce048de. Reverted #155357 on behalf of https://github.com/seemethere due to This is causing buck builds to fail since we didn't add the definition of AT_USE_EIGEN_SPARSE in the buckbuild.bzl file, will follow-up and re-land this. ([comment](#155357 (comment)))

pytorchmergebot · 2025-08-21T22:38:43Z

@Aidyn-A your PR has been successfully reverted.

Signed-off-by: Eli Uriegas <eliuriegas@meta.com>

facebook-github-bot · 2025-08-21T22:41:37Z

@seemethere has imported this pull request. If you are a Meta employee, you can view this in D80750986.

…addmm (pytorch#155357)" This reverts commit ce048de. Reverted pytorch#155357 on behalf of https://github.com/seemethere due to This is causing buck builds to fail since we didn't add the definition of AT_USE_EIGEN_SPARSE in the buckbuild.bzl file, will follow-up and re-land this. ([comment](pytorch#155357 (comment)))

facebook-github-bot · 2025-08-23T07:12:37Z

@pytorchbot merge

(Initiating merge automatically since Phabricator Diff has merged)

pytorchmergebot · 2025-08-23T07:14:30Z

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

pytorchmergebot · 2025-08-23T07:14:45Z

Merge failed

Reason: 1 jobs have failed, first few of them are: Meta Internal-Only Changes Check

Details for Dev Infra team

Raised by workflow job

seemethere · 2025-08-23T19:02:04Z

@pytorchbot merge -f 'merged internally already'

pytorchmergebot · 2025-08-23T19:03:39Z

Merge started

Your change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use -f as last resort and instead consider -i/--ignore-current to continue the merge ignoring current failures. This will allow currently pending tests to finish and report signal before the merge.

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here

…ytorch#155357) This pull request adds the following ops for sparse matrices using Eigen library: ```python add(a_csr, b_csr) add(a_csc, b_csc) addmm(c_csr, a_csr, b_csr) addmm(c_csr, a_csr, b_csc) addmm(c_csr, a_csc, b_csc) addmm(c_csr, a_csc, b_csr) addmm(c_csc, a_csr, b_csr) addmm(c_csc, a_csr, b_csc) addmm(c_csc, a_csc, b_csc) addmm(c_csc, a_csc, b_csr) ``` Currently, the operations for sparse matrices on CPU are available through MKL only. The non-existence of MKL on `aarch64` causes the unavailability of these ops on any machines with ARM based CPUs, including Apple Silicon, AWS Graviton and NVIDIA Grace. This PR addresses this issue by using Eigen as a backend for the above ops. This is a re-factored version of my previous PR pytorch#101814. The main difference with the old one, this does not enable Eigen by default. Pull Request resolved: pytorch#155357 Approved by: https://github.com/pearu, https://github.com/eqy

…addmm (pytorch#155357)" This reverts commit ce048de. Reverted pytorch#155357 on behalf of https://github.com/seemethere due to This is causing buck builds to fail since we didn't add the definition of AT_USE_EIGEN_SPARSE in the buckbuild.bzl file, will follow-up and re-land this. ([comment](pytorch#155357 (comment)))

…ytorch#155357) This pull request adds the following ops for sparse matrices using Eigen library: ```python add(a_csr, b_csr) add(a_csc, b_csc) addmm(c_csr, a_csr, b_csr) addmm(c_csr, a_csr, b_csc) addmm(c_csr, a_csc, b_csc) addmm(c_csr, a_csc, b_csr) addmm(c_csc, a_csr, b_csr) addmm(c_csc, a_csr, b_csc) addmm(c_csc, a_csc, b_csc) addmm(c_csc, a_csc, b_csr) ``` Currently, the operations for sparse matrices on CPU are available through MKL only. The non-existence of MKL on `aarch64` causes the unavailability of these ops on any machines with ARM based CPUs, including Apple Silicon, AWS Graviton and NVIDIA Grace. This PR addresses this issue by using Eigen as a backend for the above ops. This is a re-factored version of my previous PR pytorch#101814. The main difference with the old one, this does not enable Eigen by default. Pull Request resolved: pytorch#155357 Approved by: https://github.com/pearu, https://github.com/eqy Co-authored-by: Eli Uriegas <eliuriegas@meta.com>

pytorch-bot bot added the release notes: sparse release notes category label Jun 6, 2025

Aidyn-A changed the title ~~sparse add and addmm~~ [ATen][CPU][Sparse] Use Third-Party Eigen for sparse add and addmm Jun 6, 2025

Aidyn-A requested review from cpuhrsch and malfet June 6, 2025 18:55

pytorchbot added the open source label Jun 6, 2025

mikaylagawarecki requested a review from pearu June 6, 2025 22:02

mikaylagawarecki added the triaged This issue has been looked at a team member, and triaged and prioritized into an appropriate module label Jun 6, 2025

pytorchmergebot force-pushed the eigen_for_sparse_addmm_v2 branch from 71d7617 to 8db7338 Compare June 7, 2025 06:20

Skylion007 reviewed Jun 7, 2025

View reviewed changes

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp Outdated Show resolved Hide resolved

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp Outdated Show resolved Hide resolved

Aidyn-A force-pushed the eigen_for_sparse_addmm_v2 branch from 8db7338 to df77628 Compare June 7, 2025 18:03

pearu requested changes Jun 19, 2025

View reviewed changes

Aidyn-A force-pushed the eigen_for_sparse_addmm_v2 branch from df77628 to 36d5c43 Compare July 14, 2025 08:51

Aidyn-A requested a review from pearu July 14, 2025 16:53

pytorchmergebot force-pushed the eigen_for_sparse_addmm_v2 branch from 1afd750 to 2ea6b20 Compare July 23, 2025 23:52

pearu requested changes Aug 5, 2025

View reviewed changes

Aidyn-A force-pushed the eigen_for_sparse_addmm_v2 branch from 2ea6b20 to 2c69b71 Compare August 7, 2025 14:12

pearu reviewed Aug 8, 2025

View reviewed changes

aten/src/ATen/native/sparse/eigen/SparseBlasImpl.cpp Outdated Show resolved Hide resolved

Aidyn-A added 6 commits August 12, 2025 16:16

sparse add and addmm

f18af1d

fix lint and bazel

e0a0a15

clean up and enable fastpath

807eb71

apply suggestions

ff0ab1c

missed a couple of suggestions

9d1eb80

apply review suggestions

99f56da

pytorchmergebot added the Merged label Aug 20, 2025

pytorchmergebot closed this in ce048de Aug 20, 2025

pytorchmergebot removed the merging label Aug 20, 2025

pytorchmergebot added Reverted ci-no-td Do not run TD on this PR labels Aug 21, 2025

pytorchmergebot reopened this Aug 21, 2025

add AT_USE_EIGEN_SPARSE to buckbuild.bzl

c77592f

Signed-off-by: Eli Uriegas <eliuriegas@meta.com>

pytorchmergebot added the merging label Aug 23, 2025

pytorchmergebot removed the merging label Aug 23, 2025

pytorchmergebot added the merging label Aug 23, 2025

pytorchmergebot closed this in 3e5b021 Aug 23, 2025

pytorchmergebot removed the merging label Aug 23, 2025

Conversation

Aidyn-A commented Jun 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Jun 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/155357

✅ No Failures

Uh oh!

Aidyn-A commented Jun 7, 2025

Uh oh!

pytorchmergebot commented Jun 7, 2025

Uh oh!

pytorchmergebot commented Jun 7, 2025

Uh oh!

Uh oh!

Uh oh!

pearu left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Aidyn-A commented Jul 23, 2025

Uh oh!

pytorchmergebot commented Jul 23, 2025

Uh oh!

pytorchmergebot commented Jul 23, 2025

Uh oh!

pearu left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

pearu left a comment

Choose a reason for hiding this comment

Uh oh!

Uh oh!

seemethere commented Aug 21, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

pytorch-bot bot commented Aug 21, 2025

Uh oh!

seemethere commented Aug 21, 2025

Uh oh!

pytorchmergebot commented Aug 21, 2025

Uh oh!

pytorchmergebot commented Aug 21, 2025

Uh oh!

facebook-github-bot commented Aug 21, 2025

Uh oh!

facebook-github-bot commented Aug 23, 2025

Uh oh!

pytorchmergebot commented Aug 23, 2025

Merge started

Uh oh!

pytorchmergebot commented Aug 23, 2025

Merge failed

Uh oh!

seemethere commented Aug 23, 2025

Uh oh!

pytorchmergebot commented Aug 23, 2025

Merge started

Uh oh!

Reviewers

Aidyn-A commented Jun 6, 2025 •

edited

Loading

pytorch-bot bot commented Jun 6, 2025 •

edited

Loading

pearu left a comment •

edited

Loading

seemethere commented Aug 21, 2025 •

edited

Loading