Skip to content

[aarch64][caffe2] Fix FBGEMM detection on aarch64#169379

Closed
andrewjcg wants to merge 7 commits intopytorch:mainfrom
andrewjcg:export-D88096853
Closed

[aarch64][caffe2] Fix FBGEMM detection on aarch64#169379
andrewjcg wants to merge 7 commits intopytorch:mainfrom
andrewjcg:export-D88096853

Conversation

@andrewjcg
Copy link
Copy Markdown
Contributor

@andrewjcg andrewjcg commented Dec 2, 2025

Summary: Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853

submodule

cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @jerryzh168 @aditew01

@pytorch-bot
Copy link
Copy Markdown

pytorch-bot bot commented Dec 2, 2025

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/169379

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

⏳ No Failures, 1 Pending

As of commit 1fe48df with merge base afb173d (image):
💚 Looks good so far! There are no failures yet. 💚

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

@pytorch-bot pytorch-bot bot added the topic: not user facing topic category label Dec 2, 2025
@meta-codesync
Copy link
Copy Markdown

meta-codesync bot commented Dec 2, 2025

@andrewjcg has exported this pull request. If you are a Meta employee, you can view the originating Diff in D88096853.

Copy link
Copy Markdown
Contributor

@malfet malfet left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

IMO this is a misleading statement and will lead to compilation errors on RiscV, PowerPC, S390 hardware. Also, PyTorch is not really supported on 32-bit OSes, so one can just remove this check completely

CMakeLists.txt Outdated
if(USE_FBGEMM AND NOT CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|aarch64)$")
message(WARNING
"x64 operating system is required for FBGEMM. "
"64-bit operating system is required for FBGEMM. "
Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

This is a wrong statement. This will not work on RISCV or S390, will it? If one really wants to check for 64-bit OS, they need to modify the condition to CMAKE_SIZEOF_VOID_P EQUALS 8

Copy link
Copy Markdown
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Updated -- does this seem good?

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good to me, thank you for the update

andrewjcg added a commit to andrewjcg/pytorch that referenced this pull request Dec 2, 2025
Summary:

Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853
@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 3, 2025

IMO this is a misleading statement and will lead to compilation errors on RiscV, PowerPC, S390 hardware. Also, PyTorch is not really supported on 32-bit OSes, so one can just remove this check completely

@malfet Does this mean that we can just remove AND NOT CMAKE_SYSTEM_PROCESSOR MATCHES "^(x86_64|aarch64)$" and depend on USE_FBGEMM? If yes, it does make the logic simpler

@pytorch-bot pytorch-bot bot added the ciflow/trunk Trigger trunk jobs on your pull request label Dec 3, 2025
@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 4, 2025

The aarch64 build failure looks related https://github.com/pytorch/pytorch/actions/runs/19908456569/job/57070803485 also I have no idea yet how setting USE_FBGEMM on aarch64 leads to that

@huydhn huydhn added the no-runner-experiments Bypass Meta/LF runner determinator label Dec 5, 2025
huydhn added a commit to huydhn/pytorch that referenced this pull request Dec 5, 2025
Signed-off-by: Huy Do <[email protected]>
@malfet
Copy link
Copy Markdown
Contributor

malfet commented Dec 6, 2025

@pytorchbot rebase -b main

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

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

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

Successfully rebased export-D88096853 onto refs/remotes/origin/main, please pull locally before adding more changes (for example, via git checkout export-D88096853 && git pull --rebase)

pytorch-bot bot pushed a commit that referenced this pull request Dec 6, 2025
Summary:

Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853
@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 6, 2025

@pytorchbot rebase -b main

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

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

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

Successfully rebased export-D88096853 onto refs/remotes/origin/main, please pull locally before adding more changes (for example, via git checkout export-D88096853 && git pull --rebase)

pytorchmergebot pushed a commit to andrewjcg/pytorch that referenced this pull request Dec 6, 2025
Summary:

Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853
@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 8, 2025

@pytorchbot rebase -b main

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

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

Summary:

Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853
@pytorchmergebot
Copy link
Copy Markdown
Collaborator

Successfully rebased export-D88096853 onto refs/remotes/origin/main, please pull locally before adding more changes (for example, via git checkout export-D88096853 && git pull --rebase)

@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 10, 2025

With pytorch/FBGEMM#5207 landed, this should work now across all platforms

@cyyever
Copy link
Copy Markdown
Collaborator

cyyever commented Dec 10, 2025

@huydhn Also consider my fixes fe87087 and fd8f414

@huydhn
Copy link
Copy Markdown
Contributor

huydhn commented Dec 10, 2025

@huydhn Also consider my fixes fe87087 and fd8f414

Thank you for the context! I guess #163952 can be merged separately after this lands?

@cyyever
Copy link
Copy Markdown
Collaborator

cyyever commented Dec 10, 2025

Not sure, it enables more platforms.

@facebook-github-bot
Copy link
Copy Markdown
Contributor

@pytorchbot merge

(Initiating merge automatically since Phabricator Diff has merged)

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

This PR updates submodules third_party/fbgemm

If those updates are intentional, please add "submodule" keyword to PR title/description.

@izaitsevfb
Copy link
Copy Markdown
Contributor

@pytorchbot merge -i

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

Merge started

Your change will be merged while ignoring the following 0 checks:

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

@cyyever cyyever added the ciflow/linux-aarch64 linux aarch64 CI workflow label Dec 12, 2025
@cyyever
Copy link
Copy Markdown
Collaborator

cyyever commented Dec 12, 2025

@pytorchbot merge -f "Pending H100, but aarch64 CI is green"

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command
For more information see pytorch-bot wiki.

@pytorchmergebot
Copy link
Copy Markdown
Collaborator

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

vishalgoyal316 pushed a commit to vishalgoyal316/pytorch that referenced this pull request Dec 17, 2025
Summary: Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853

submodule

Pull Request resolved: pytorch#169379
Approved by: https://github.com/malfet

Co-authored-by: Huy Do <[email protected]>
krastogi-in pushed a commit to krastogi-in/pytorch that referenced this pull request Jan 9, 2026
Summary: Update as per-comment: FBGEMM works on aarch64, just not on x86_32.

Test Plan: CI

Differential Revision: D88096853

submodule

Pull Request resolved: pytorch#169379
Approved by: https://github.com/malfet

Co-authored-by: Huy Do <[email protected]>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

ciflow/b200 ciflow/h100 ciflow/linux-aarch64 linux aarch64 CI workflow ciflow/rocm Trigger "default" config CI on ROCm ciflow/trunk Trigger trunk jobs on your pull request fb-exported Merged meta-exported module: cpu CPU specific problem (e.g., perf, algorithm) no-runner-experiments Bypass Meta/LF runner determinator test-config/default topic: not user facing topic category

Projects

None yet

Development

Successfully merging this pull request may close these issues.

7 participants