Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Revert "Revert "[LLVMGPU][ROCm] Add MFMA_F32_16x16x4_F32 instruction"… #17921

Merged
merged 2 commits into from
Jul 24, 2024

Conversation

pashu123
Copy link
Contributor

… (#17894)"

This reverts commit 02c2000.

@pashu123 pashu123 force-pushed the wmma_rebased branch 2 times, most recently from 836a821 to 2ee6d68 Compare July 23, 2024 16:10
@pashu123 pashu123 marked this pull request as ready for review July 23, 2024 17:16
@pashu123 pashu123 requested a review from raikonenfnu July 23, 2024 17:16
// TODO: Generalize to other dimensions.
// Try to search for pad value and check only filter dimension is blocked.
SmallVector<std::array<int64_t, 3>> mnkPaddingCandidates;
for (const GPUMatmulShapeType &intrinsic : intrinsics) {

if (inpElemType != intrinsic.aType && kernelElemType != intrinsic.bType) {
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I think this should be || no? because if either of this is not true then we want to skip.

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, thanks for correcting.

if lhs_rhs_type.value.upper() not in schedule.intrinsic:
# Search for the lhs_rhs type in the first part of intrinsic
# e.g., MFMA_F32_16x16x4_F32 -> MFMA_F32
if lhs_rhs_type.value.upper() not in schedule.intrinsic[:8]:
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would this work with MFMA_F8E4M3FNUZ intrinsic?

Copy link
Contributor Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The current logic won't. I've updated the script to find the right pattern. PTAL.

Copy link
Collaborator

@raikonenfnu raikonenfnu left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM!

@pashu123 pashu123 merged commit ae00c4f into iree-org:main Jul 24, 2024
43 checks passed
LLITCHEV pushed a commit to LLITCHEV/iree that referenced this pull request Jul 30, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

2 participants