[PyTorch] Use 128-bit vectors for ARM64 #137426

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

Closed

swolchok wants to merge 23 commits into gh/swolchok/650/base from gh/swolchok/650/head

Contributor

swolchok commented Oct 7, 2024 •

edited

Loading

Stack from ghstack (oldest at bottom):

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: D63984039

cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @malfet @snadampal @milpuz01 @voznesenskym @penguinwu @EikanWang @Guobing-Chen @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang @aakhundov


          [PyTorch] Use 128-bit vectors for ARM64

f89be78

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

[ghstack-poisoned]

pytorch-bot bot commented Oct 7, 2024 •

edited

Loading

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/137426

📄 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 7370baf with merge base fb0da32 ():
💚 Looks good so far! There are no failures yet. 💚

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

pytorch-bot bot added ciflow/inductor module: cpu module: inductor labels

Contributor

facebook-github-bot commented Oct 7, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

f828080

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

ghstack-source-id: 246599433
Pull Request resolved: #137426

facebook-github-bot added the fb-exported label

Collaborator

abhishek-iitmadras commented Oct 7, 2024

@pytorchbot label "ciflow/linux-aarch64"

pytorch-bot bot added the ciflow/linux-aarch64 label


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

c383c17

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

swolchok mentioned this pull request

[PyTorch] Port ExecuTorch bfdot improvement back to ATen BlasKernel, Try #2 #137377

Closed

Contributor

facebook-github-bot commented Oct 7, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

pytorch-bot bot added the release notes: jit label


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

c27ed1b

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

Contributor

facebook-github-bot commented Oct 7, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

02ca0f9

Pull Request resolved: #137426

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.
ghstack-source-id: 246714926

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

swolchok commented

View reviewed changes

aten/src/ATen/CMakeLists.txt Show resolved Hide resolved


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

cba3897

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

Contributor

facebook-github-bot commented Oct 8, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

50c0bef

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

1af48f4

Pull Request resolved: #137426

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.
ghstack-source-id: 246998171

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

Contributor

facebook-github-bot commented Oct 9, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039


          another attempted linux aarch64 fix on "[PyTorch] Use 128-bit vectors…

c5f85e5

… for ARM64"

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

f57efa8

Pull Request resolved: #137426

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.
ghstack-source-id: 247065546

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

Contributor

facebook-github-bot commented Oct 9, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok requested review from jgong5 and malfet

October 9, 2024 19:22


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

f3c4a3c

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 malfet snadampal milpuz01 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

Contributor

facebook-github-bot commented Oct 23, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok mentioned this pull request

[PyTorch] Fix ASAN failures for vec_test_all_types Cast test #138716

Closed

swolchok added a commit that referenced this pull request


          [PyTorch] Fix out-of-bounds array access in atomic_add_vec

b0e8635

There is no guarantee that `len` here is enough for a full vector. This was causing at least one test failure on #137426.

Differential Revision: [D64857786](https://our.internmc.facebook.com/intern/diff/D64857786/)

[ghstack-poisoned]

swolchok mentioned this pull request

[PyTorch] Fix out-of-bounds array access in atomic_add_vec #138744

Closed


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

83f4b9a

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 malfet snadampal milpuz01 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

Contributor

facebook-github-bot commented Oct 23, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

16d08b8

Pull Request resolved: #137426

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.
ghstack-source-id: 249693590

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)


          Update on "[PyTorch] Use 128-bit vectors for ARM64"

7370baf

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

cc jgong5 mingfeima XiaobingSuper sanchitintel ashokei jingxu10 malfet snadampal milpuz01 voznesenskym penguinwu EikanWang Guobing-Chen zhuhaozhe blzheng wenzhe-nrv jiayisunx ipiszy yf225 chenyang78 kadeng muchulee8 ColinPeppler amjames desertfire chauhang

[ghstack-poisoned]

Contributor

facebook-github-bot commented Oct 24, 2024

This pull request was exported from Phabricator. Differential Revision: D63984039

swolchok requested a review from kulinseth as a code owner

October 24, 2024 16:54

swolchok added a commit that referenced this pull request


          [PyTorch] Use 128-bit vectors for ARM64

275a703

Pull Request resolved: #137426

The correct vector length for ARM64 is 128 bits (16
bytes). We were previously using double this, apparently just because
that would be the same length as AVX2.
ghstack-source-id: 249869364

Differential Revision: [D63984039](https://our.internmc.facebook.com/intern/diff/D63984039/)

pytorch-bot bot added the ciflow/mps label

pytorchmergebot pushed a commit that referenced this pull request


          [PyTorch] Fix inductor bug with unrolled vectorized prod (#138542)

ead5738

This issue is one of two inductor bugs blocking land of #137426. Turned out to be simple

Differential Revision: [D64734116](https://our.internmc.facebook.com/intern/diff/D64734116/)

Pull Request resolved: #138542
Approved by: https://github.com/jgong5, https://github.com/malfet
ghstack dependencies: #138486

Co-authored-by: leslie-fang-intel <leslie.fang@intel.com>

pytorchmergebot pushed a commit that referenced this pull request


          [PyTorch] Fix out-of-bounds array access in atomic_add_vec (#138744)

10a34dc

There is no guarantee that `len` here is enough for a full vector. This was causing at least one test failure on #137426.

Differential Revision: [D64857786](https://our.internmc.facebook.com/intern/diff/D64857786/)

Pull Request resolved: #138744
Approved by: https://github.com/jgong5, https://github.com/malfet
ghstack dependencies: #138486, #138542, #138655, #138716

Contributor Author

swolchok commented Oct 25, 2024

@pytorchbot merge

pytorchmergebot added the merging label

Collaborator

pytorchmergebot commented Oct 25, 2024

Merge failed

Reason: Approvers from one of the following sets are needed:

superuser (pytorch/metamates)
Core Reviewers (mruberry, lezcano, Skylion007, ngimel, peterbell10, ...)
Core Maintainers (soumith, gchanan, ezyang, dzhulgakov, malfet, ...)

Details for Dev Infra team

Raised by workflow job

Failing merge rule: Core Maintainers

pytorchmergebot removed the merging label

malfet approved these changes

View reviewed changes

Contributor

malfet commented Oct 26, 2024

@pytorchbot merge

pytorchmergebot added the merging label

Collaborator

pytorchmergebot commented Oct 26, 2024

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 added the Merged label

pytorchmergebot closed this in

a3de067

pytorchmergebot removed the merging label

maajidkhann mentioned this pull request

Extending SVE VEC Backend Support in PyTorch to SVE128 and SVE512. #138388

Open

github-actions bot deleted the gh/swolchok/650/head branch

November 25, 2024 02:11

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/inductor ciflow/linux-aarch64 ciflow/mps ciflow/trunk fb-exported Merged module: arm module: cpu module: inductor release notes: jit