damien-lejeune aab1d219f5 [rocm-libraries] ROCm/rocm-libraries#8350 (commit f92ded1) ...

Add tile shape for FMHA batch prefill on MI308X (on fp8,
 hdim=256) (#8350)

## Motivation

Add a tile size appropriate for FMHA batch prefill fp8/hdim256 on MI308X

## Technical Details

Appending the tile shape to the existing factory such that it can be
picked up by Aiter

## Test Plan

Ran the performance test on both MI300X and MI308X

## Test Result

MI300X performance seems unaffected by this change. MI308X does improve.

## Submission Checklist

- [X] Look over the contributing guidelines at
https://github.com/ROCm/ROCm/blob/develop/CONTRIBUTING.md#pull-requests.

Co-authored-by: Damien Lejeune <damien.lejeune@amd.com>

2026-06-15 07:00:35 +00:00

..

__init__.py

chore(copyright): update copyright header for example directory (#3273)

2025-11-24 18:02:41 -08:00

fmha_batch_prefill.py

[rocm-libraries] ROCm/rocm-libraries#8350 (commit f92ded1)

2026-06-15 07:00:35 +00:00

fmha_bwd.py

[rocm-libraries] ROCm/rocm-libraries#7586 (commit c18f2c7)

2026-05-30 00:10:26 +00:00

fmha_fwd_appendkv.py

[rocm-libraries] ROCm/rocm-libraries#6978 (commit e58096d)

2026-05-15 06:46:51 -07:00

fmha_fwd_splitkv.py

[rocm-libraries] ROCm/rocm-libraries#7500 (commit f5cd4fd)

2026-06-03 06:16:10 +00:00

fmha_fwd.py

[rocm-libraries] ROCm/rocm-libraries#7723 (commit 4ed6c51)

2026-05-28 15:58:54 +00:00

fmha_pagedkv_prefill.py

[rocm-libraries] ROCm/rocm-libraries#6978 (commit e58096d)

2026-05-15 06:46:51 -07:00