[AMD] Remove cuda2gcn and opencl #3565
Merged
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
zhanglx13
commented
Apr 4, 2024
| @@ -52,7 +52,7 @@ def __post_init__(self): | |||
| # Ignore user-defined warp size for gfx9 | |||
| warp_size = 32 if 'gfx10' in self.arch or 'gfx11' in self.arch else 64 | |||
| object.__setattr__(self, 'warp_size', warp_size) | |||
| libs = ["cuda2gcn", "opencl", "ocml", "ockl"] | |||
| libs = ["ocml", "ockl"] | |||
There was a problem hiding this comment.
@antiagainst Did you leave opencl for some reason when you cleaned up the bc files?
There was a problem hiding this comment.
No particular reason. I was just dropping the ones obviously not needed before due to my change. Didn't inspect whether this one is still needed. Great to know that it can be dropped too!
antiagainst
approved these changes
Apr 4, 2024
micmelesse
added a commit
to micmelesse/triton
that referenced
this pull request
Apr 10, 2024
This reverts commit 70fe526.
zhanglx13
added a commit
that referenced
this pull request
Apr 12, 2024
This reverts commit 70fe526.
zahimoud
pushed a commit
that referenced
this pull request
Apr 12, 2024
Reverts #3565 Pytorch-inductor's triton codegen utilises math functions not implemented in tl.math e.g. isinf, signbit (https://github.com/pytorch/pytorch/blob/41613a0803f7cde7956f039bc80f94253b0843f9/torch/_inductor/codegen/triton.py#L837) The removal of cuda2gcn causes functionality breakages in inductor until an alternative lowering is implemented. This is important for the ongoing effort to utilise upstream triton in pytorch instead of the ROCm fork.
jataylo
added a commit
to jataylo/triton
that referenced
this pull request
Apr 15, 2024
This reverts commit 70fe526.
zhanglx13
pushed a commit
that referenced
this pull request
Apr 15, 2024
This reverts commit 70fe526. Initially just cuda2gcn was reverted here 2556127 but some libdevice lowerings in inductor rely on opencl.bc Example code: https://gist.github.com/jataylo/a5f804d48d8f831395a208836fd4740a
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Since the math functions are no longer lowered to __nv_xxx first before lowered to __ocml/ockl_xxx, we don't need cuda2gcn translation any more on AMD backend.