Bin Bao

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D54258088).

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D54258088).

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D54258088).

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D54258088).

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D54258088).

Verified reproducible with pytorch 2.1, but not on the nightly or my local build (hash `f2d476843ee`).

We do have an environment variable for that, `AOT_INDUCTOR_DEBUG_COMPILE=1`. https://github.com/pytorch/pytorch/blob/65efece3a4acf23fd3c38f5217c545cd989f9cda/torch/_inductor/config.py#L605

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D52328687).

@desertfire has imported this pull request. If you are a Meta employee, you can view this diff [on Phabricator](https://www.internalfb.com/diff/D52328687).

Fixing this at the decomposition time seems to cause quite a few pattern matching optimization to fail, see https://github.com/pytorch/pytorch/pull/125330. Probably not worth to do that for fixing this small issue.