Skip to content

GEMM: support 2d grouped scales #4478

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.

Sign up for GitHub

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

Jump to bottom
Merged
kealan-barbieri merged 7 commits into from
Jan 8, 2026
Merged

GEMM: support 2d grouped scales #4478

kealan-barbieri merged 7 commits into from
Jan 8, 2026

Conversation

@kealan-barbieri
Copy link
Contributor

@kealan-barbieri kealan-barbieri commented Dec 22, 2025 •
edited
Loading

Description

Enable scales with non-trivial M/N group dim in existing scales use cases. Fix for improper bf16 special path usage.

Fixes # MFDNN-14021

Checklist

General

  • Do all unit and benchdnn tests (make test and make test_benchdnn_*) pass locally for each commit?
  • Have you formatted the code using clang-format?

@kealan-barbieri kealan-barbieri requested review from a team as code owners December 22, 2025 23:55
@github-actions github-actions bot added platform:gpu-intel Codeowner: @oneapi-src/onednn-gpu-intel component:tests Codeowner: @oneapi-src/onednn-arch component:common labels Dec 22, 2025
dzarukin
dzarukin reviewed Jan 5, 2026
View reviewed changes
@kealan-barbieri
Copy link
Contributor Author

kealan-barbieri commented Jan 6, 2026

make test
set test_scope=NIGHTLY
disable benchdnn_all
enable benchdnn_matmul
enable arch_gpu_xe-hpc
enable arch_gpu_xe-hpg-atsm
enable arch_gpu_xe-hpg-dg2
enable arch_gpu_xe-lp
enable arch_gpu_xe-lpg
enable arch_gpu_xe-lpg+
enable arch_gpu_xe2-hpg-bmg
enable arch_gpu_xe2-lpg
enable arch_gpu_xe3-lpg

@kealan-barbieri
Copy link
Contributor Author

kealan-barbieri commented Jan 6, 2026

make test
set test_scope=NIGHTLY
disable benchdnn_all
enable benchdnn_matmul
enable arch_gpu_xe-hpc
enable arch_gpu_xe-hpg-atsm
enable arch_gpu_xe-hpg-dg2
enable arch_gpu_xe-lp
enable arch_gpu_xe-lpg
enable arch_gpu_xe-lpg+
enable arch_gpu_xe2-hpg-bmg
enable arch_gpu_xe2-lpg
enable arch_gpu_xe3-lpg

Simonsays095
Simonsays095 reviewed Jan 7, 2026
View reviewed changes
@kealan-barbieri kealan-barbieri force-pushed the kealanba/double_block branch 5 times, most recently from 960a50e to a165b12 Compare January 8, 2026 01:15
Simonsays095
Simonsays095 approved these changes Jan 8, 2026
View reviewed changes
@kealan-barbieri
Copy link
Contributor Author

kealan-barbieri commented Jan 8, 2026

make test
set test_scope=NIGHTLY
disable benchdnn_all
enable benchdnn_matmul
enable arch_gpu_xe-hpc
enable arch_gpu_xe-hpg-atsm
enable arch_gpu_xe-hpg-dg2
enable arch_gpu_xe-lp
enable arch_gpu_xe-lpg
enable arch_gpu_xe-lpg+
enable arch_gpu_xe2-hpg-bmg
enable arch_gpu_xe2-lpg
enable arch_gpu_xe3-lpg

dzarukin
dzarukin approved these changes Jan 8, 2026
View reviewed changes
tests/benchdnn/inputs/matmul/shapes_2d_attr Outdated
2x2048x256:2x256x128_n"NCF:1*1"
2x2048x128:1x128x128_n"NCF:2*1"
2x896x1024:2x1024x4096_n"RNN-T:Encoder_cell1_Hidden*11"
2x896x320:1x320x1280_n"RNN-T:Prediction_Input*12"
Copy link
Contributor

@dzarukin dzarukin Jan 8, 2026

Choose a reason for hiding this comment

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

320 doesn't divide by 128 to the best of my knowledge. Does benchdnn skip as invalid or as unimplemented?

Copy link
Contributor Author

@kealan-barbieri kealan-barbieri Jan 8, 2026

Choose a reason for hiding this comment

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

it skips as invalid, will drop the case just to reduce noise.

@kealan-barbieri kealan-barbieri merged commit a10ac3a into main Jan 8, 2026
25 of 30 checks passed
@kealan-barbieri kealan-barbieri deleted the kealanba/double_block branch January 8, 2026 22:12
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@Simonsays095 Simonsays095 Simonsays095 approved these changes

@dzarukin dzarukin dzarukin approved these changes

Assignees

No one assigned

Labels

component:common component:tests Codeowner: @oneapi-src/onednn-arch platform:gpu-intel Codeowner: @oneapi-src/onednn-gpu-intel

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

3 participants

@kealan-barbieri @dzarukin @Simonsays095