Remove unused sleef.h which breaks cross-compilation on windows

[ykhrustalev]

Summary

The sleef.h header contains

#if (defined(__MINGW32__) || defined(__MINGW64__) || defined(__CYGWIN__) || defined(_MSC_VER)) && !defined(SLEEF_STATIC_LIBS)
#ifdef SLEEF_IMPORT_IS_EXPORT
#define SLEEF_IMPORT __declspec(dllexport)
#else // #ifdef SLEEF_IMPORT_IS_EXPORT
#define SLEEF_IMPORT __declspec(dllimport)
#if (defined(_MSC_VER))
#pragma comment(lib,"sleef.lib")
#endif // #if (defined(_MSC_VER))
#endif // #ifdef SLEEF_IMPORT_IS_EXPORT
#else // #if (defined(__MINGW32__) || defined(__MINGW64__) || defined(__CYGWIN__) || defined(_MSC_VER)) && !defined(SLEEF_STATIC_LIBS)
#define SLEEF_IMPORT
#endif // #if (defined(__MINGW32__) || defined(__MINGW64__) || defined(__CYGWIN__) || defined(_MSC_VER)) && !defined(SLEEF_STATIC_LIBS)

#pragma comment(lib,"sleef.lib")

which forces the sleef.lib to be linked for the static windows lib.
The 2 particular optimized kernels are not using any sleef functions and could be used as a subset that is statically compiled.

Test plan

Removing the header makes the static link to pass without a need to have a sleef lib.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/executorch/13079

• 📄 Preview Python docs built from this PR

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

❌ 3 New Failures, 1 Pending

As of commit 74c887e with merge base 27fa18e ():

NEW FAILURES - The following jobs have failed:

• Build documentation / build (buck2) / Build doc (gh)
  At least one of the pre-conditions you specified did not hold
• pull / unittest-arm-backend-with-no-fvp (test_pytest_models) / linux-job (gh)
  RuntimeError: Command docker exec -t 019c4556388f0e9ec59f0f42c470c2563ca2f8832ea01d8d2e67b931a6d57c39 /exec failed with exit code 1
• pull / unittest-arm-backend-with-no-fvp (test_pytest_ops) / linux-job (gh)
  RuntimeError: Command docker exec -t a905c6a57c15f155174aa7413cfb825629bf0d1306631f7c287f1a94417d71da /exec failed with exit code 1

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

[github-actions]

This PR needs a release notes: label

If your change should be included in the release notes (i.e. would users of this library care about this change?), please use a label starting with release notes:. This helps us keep track and include your important work in the next release notes.

To add a label, you can comment to pytorchbot, for example
@pytorchbot label "release notes: none"

For more information, see
https://github.com/pytorch/pytorch/wiki/PyTorch-AutoLabel-Bot#why-categorize-for-release-notes-and-how-does-it-work.

[digantdesai]

The 2 particular optimized kernels are not using any sleef functions and could be used as a subset that is statically compiled.

Good. yeah we use at::native::vectorized_gelu now for gelu for instance.
If the CI is clean perhaps we can merge. Do you mind fixing the Buck target file as well?

[facebook-github-bot]

@digantdesai has imported this pull request. If you are a Meta employee, you can view this in D79648326.

[digantdesai]

Running CI now

[ykhrustalev]

The 2 particular optimized kernels are not using any sleef functions and could be used as a subset that is statically compiled.

Good. yeah we use at::native::vectorized_gelu now for gelu for instance. If the CI is clean perhaps we can merge. Do you mind fixing the Buck target file as well?

@digantdesai the linking comes from the "header" file, not from the buck or cmake. Can you please elaborate the Buck change you mean?

[digantdesai]

NVM we may not need any update to the BUCK file.

[kimishpatel]

wait but gelu.h from aten native is all header only impl which pulls in other headers like Vectorized.h. So inclusion of sleef should likely happen there, which does include sleef under these conditions, https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/cpu/vec/vec128/vec128_float_neon.h#L10-L12. So if you are avoiding use of sleef tahts fine (as long as it doesnt break other builds), but it will cost perf on windows

[ykhrustalev]

wait but gelu.h from aten native is all header only impl which pulls in other headers like Vectorized.h. So inclusion of sleef should likely happen there, which does include sleef under these conditions, https://github.com/pytorch/pytorch/blob/main/aten/src/ATen/cpu/vec/vec128/vec128_float_neon.h#L10-L12. So if you are avoiding use of sleef tahts fine (as long as it doesnt break other builds), but it will cost perf on windows

That's fine, we are using only a subset of kernels for that case, and the 2 kernels in question are not using Sleef in their implementation. Other sleef kernels are skipped via compilation flags.

[digantdesai]

Once the CI is green (after rebase) I can merge this. @kimishpatel - OK with this?

[kimishpatel]

yeah ok with me

[facebook-github-bot]

@digantdesai has imported this pull request. If you are a Meta employee, you can view this in D79648326.