fix miss isa bool check

[xuhancn]

New cpp builder missed ISA bool(dry-compile) check.

@jgong5 Found this missing and then I submit this PR to fix it.

cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @voznesenskym @penguinwu @EikanWang @Guobing-Chen @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @peterbell10 @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/128274

• 📄 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.

❗ 1 Active SEVs

There are 1 currently active SEVs. If your PR is affected, please view them below:

• Rebase your PRs: Unstable CUDA signal in CI caused by cudnn 9 update

✅ You can merge normally! (2 Unrelated Failures)

As of commit 521be4c with merge base 917387f ():

BROKEN TRUNK - The following job failed but were present on the merge base:

👉 Rebase onto the `viable/strict` branch to avoid these failures

• inductor / cuda12.1-py3.10-gcc9-sm86 / test (dynamic_inductor_timm, 1, 2, linux.g5.4xlarge.nvidia.gpu) (gh) (trunk failure)
  gluon_inception_v3

UNSTABLE - The following job failed but was likely due to flakiness present on trunk and has been marked as unstable:

• inductor / linux-jammy-cpu-py3.8-gcc11-inductor / test (inductor_torchbench_cpu_smoketest_perf, 1, 1, linux.24xl.spr-metal, unstable) (gh) (#126993)
  Process completed with exit code 1.

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

[ezyang]

Test? I don't understand what the change is doing, why are there falsey isa in the list of isa?

[xuhancn]

Test? I don't understand what the change is doing, why are there falsey isa in the list of isa?

Hi @ezyang
I merge the new cpp_builder #125849 and miss the isa bool check. @jgong5 find the missing code, and then I submit this PR.

falsey fail are cuda related not caused by cpu isa.
I just testing whether re-run failed job can fix env related UT.

This PR reviewed by @jgong5 , don't worry.

[xuhancn]

@pytorchbot rebase

[xuhancn]

rebase to fix unrelated fail case.

[pytorchmergebot]

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

[pytorchmergebot]

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

[xuhancn]

@ezyang previous fail was bringed from trunk, after rebase they are gone.

[ezyang]

You asked me for a review, and now you're getting a review from me :P

I don't have anything fundamentally against this patch. But I would like to understand it more. That is what the code review process is for. I'll ask again:

1. Under what circumstances can isa be falsey? How come it can be falsey but also be in the list of supported isas?
2. This seems like a plausible bug to me. But how does one trigger this situation? Can we test it in CI? An explanation why it's not easy to test is also acceptable.

[xuhancn]

You asked me for a review, and now you're getting a review from me :P

I don't have anything fundamentally against this patch. But I would like to understand it more. That is what the code review process is for. I'll ask again:

1. Under what circumstances can isa be falsey? How come it can be falsey but also be in the list of supported isas?
2. This seems like a plausible bug to me. But how does one trigger this situation? Can we test it in CI? An explanation why it's not easy to test is also acceptable.

Hi @ezyang
Previous PR lost and isa, it seems will lead to lost call class VecISA's def __bool__(self) -> bool:. This function will check compiler's capability for isa.

1. If the compiler too old to support the checking isa, def __bool__(self) -> bool: will return False and skip the checking isa. Please check the compiler capability here: https://github.com/intel/intel-extension-for-pytorch/blob/main/docs/tutorials/features/isa_dynamic_dispatch.md#cpu-isa-build-compiler-requirement . If we use gcc 7.x for AVX512, this version gcc can't support AVX512, but we miss the checking and fail to catch the compiler error.
2. Current CI at least using gcc 9, it can cover all testing isa. So we can't find my missing code: and isa.

@jgong5 tell me it is a serious bug, I need fix it ASAP before branch cut.

[xuhancn]

@pytorchbot merge

[pytorchmergebot]

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