[NVGPU] Propagate layouts through DotWait ops. #3324
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
jlebar
force-pushed
the
dev-jlebar/propagate-dot-wait
branch
from
d38c20b to
b2cae8a
Compare
ThomasRaoux
approved these changes
Mar 8, 2024
#3301 converts some tt.dot ops into ttng.dot_async+wait. We need to propagate layouts through these the same way we propagate them through regular tt.dot ops. Otherwise...well, eventually we create unsupported IR, trying to convert from blocked to MMAv3 dot-operand layout, which we canot do. This fixes an OAI-internal test. GPC: propagate-dot-wait
jlebar
force-pushed
the
dev-jlebar/propagate-dot-wait
branch
from
b2cae8a to
5fd0d0c
Compare
|
Thank you, Thomas. |
htyu
pushed a commit
to htyu/triton
that referenced
this pull request
Mar 20, 2024
[NVGPU] Propagate layouts through DotWait ops. triton-lang#3301 converts some tt.dot ops into ttng.dot_async+wait. We need to propagate layouts through these the same way we propagate them through regular tt.dot ops. Otherwise...well, eventually we create unsupported IR, trying to convert from blocked to MMAv3 dot-operand layout, which we cannot do. This fixes an OAI-internal test.
karupayun
pushed a commit
to openxla/triton
that referenced
this pull request
Apr 3, 2024
[NVGPU] Propagate layouts through DotWait ops. triton-lang#3301 converts some tt.dot ops into ttng.dot_async+wait. We need to propagate layouts through these the same way we propagate them through regular tt.dot ops. Otherwise...well, eventually we create unsupported IR, trying to convert from blocked to MMAv3 dot-operand layout, which we cannot do. This fixes an OAI-internal test.
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.
[NVGPU] Propagate layouts through DotWait ops.
#3301 converts some tt.dot ops into
ttng.dot_async+wait. We need to propagate layouts through these the same way
we propagate them through regular tt.dot ops. Otherwise...well, eventually we
create unsupported IR, trying to convert from blocked to MMAv3 dot-operand
layout, which we canot do.
This fixes an OAI-internal test.
PR chain