[GPU] Add col_major optional attribute to MMAAttr #19860
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
|
Nice! You can use the documentation from here if you like: https://github.com/iree-org/iree/blob/main/compiler/src/iree/compiler/Codegen/LLVMGPU/LLVMGPUConfigureTensorLayouts.cpp#L582 We already do this for attention. Also, this should really always be the canonical form of MFMA instructions. It has A, B, C all in row major format. The only thing this can messup is the broadcast semantics of MFMA instructions because they are not symmetric for A and B, however we should never really expose the broadcast semantics by themselves anyway, so this is okay, but an additional comment for it would be good |
qedawkins
force-pushed
the
swap_mfma_operands
branch
2 times, most recently
from
c68cec0 to
3799a2e
Compare
For MMA intrinsics with symmetric register layouts we have the freedom to choose between row major and column major output layouts by simply swapping the input operands. This adds a flag to `MMAAttr` that allows specifying whether the result should be column major. A follow up PR will turn the column major variant on by default for mfma as this enables vectorization of writes in non-transpose cases.
qedawkins
force-pushed
the
swap_mfma_operands
branch
from
3799a2e to
4eb2c15
Compare
qedawkins
changed the title
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Outdated suggestions cannot be applied.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.
For MMA intrinsics with symmetric register layouts we have the freedom
to choose between row major and column major output layouts by simply
swapping the input operands. This adds a flag to
MMAAttrthat allowsspecifying whether the result should be column major.
A follow up PR will turn the column major variant on by default for
mfma as this enables vectorization of writes in non-transpose cases.