[PyTorch] Add attribute version of C10_ALWAYS_INLINE#136445
[PyTorch] Add attribute version of C10_ALWAYS_INLINE#136445swolchok wants to merge 3 commits intogh/swolchok/648/basefrom
Conversation
Sometimes (such as on a lambda), you need `__attribute__((always_inline))` but not `inline`. Differential Revision: [D63266917](https://our.internmc.facebook.com/intern/diff/D63266917/) [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/136445
Note: Links to docs will display an error until the docs builds have been completed. ✅ You can merge normally! (1 Unrelated Failure)As of commit 7bf9b6f with merge base 99eb47f ( FLAKY - The following job failed but was likely due to flakiness present on trunk:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
|
This pull request was exported from Phabricator. Differential Revision: D63266917 |
albanD
left a comment
There was a problem hiding this comment.
I'm not super familiar with these. But does this mean that for a bunch of compilers, C10_ALWAYS_INLINE_ATTRIBUTE is actually not going to inline? That sounds surprising?
which compilers? it does just as much inlining as |
Sometimes (such as on a lambda), you need `__attribute__((always_inline))` but not `inline`. Differential Revision: [D63266917](https://our.internmc.facebook.com/intern/diff/D63266917/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D63266917 |
malfet
left a comment
There was a problem hiding this comment.
This sounds fine, though not entirely sure why is it important
The ones that don't have |
C10_ALWAYS_INLINE works exactly the same way.
Because it is not lambda-specific. We could alternatively attempt to replace |
Sometimes (such as on a lambda), you need `__attribute__((always_inline))` but not `inline`. Differential Revision: [D63266917](https://our.internmc.facebook.com/intern/diff/D63266917/) [ghstack-poisoned]
|
This pull request was exported from Phabricator. Differential Revision: D63266917 |
…136331) ExecuTorch's fork of BlasKernel.cpp grew bfdot support, complete with demonstration that it helps. Port it back to PyTorch. Supersedes #127488 . Includes pytorch/executorch#5444 . Differential Revision: [D63045939](https://our.internmc.facebook.com/intern/diff/D63045939/) Pull Request resolved: #136331 Approved by: https://github.com/malfet, https://github.com/albanD ghstack dependencies: #136445
Stack from ghstack (oldest at bottom):
Sometimes (such as on a lambda), you need
__attribute__((always_inline))but notinline.Differential Revision: D63266917