Repository navigation
feat(rocBLAS): enable hipBLASLt backend for F32 GEMM on gfx942 and gfx950 - #13901
Merged
Merged
Conversation
…x950 - Enable hipBLASLt for F32 (SGEMM) on all gfx950 devices - Enable hipBLASLt for F32 on both MI300A (228 CU) and MI300X (304 CU) gfx942 devices - Previously F32 was disabled on gfx950 and only enabled on 228 CU gfx942 devices - Updated CHANGELOG.md to reflect the new default backend behavior Co-Authored-By: Claude Sonnet 4 <noreply@anthropic.com>
✅ All Checks Passed — Ready for Review
📖 Need help? See the Policy FAQ for details on every check and how to fix failures. |
|
🎉 All checks passed! This PR is ready for review. |
Codecov Report✅ All modified and coverable lines are covered by tests. Additional details and impacted files@@ Coverage Diff @@
## develop #13901 +/- ##
========================================
Coverage 76.02% 76.02%
========================================
Files 2869 2869
Lines 475279 475279
Branches 71197 71197
========================================
Hits 361297 361297
Misses 89407 89407
Partials 24575 24575
*This pull request uses carry forward flags. Click here to find out more.
🚀 New features to boost your workflow:
|
TorreZuk
reviewed
Oct 19, 2026
Contributor
TheRock Submodule Bump ActivityNewest first
|
This file contains hidden or 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
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
Motivation
This PR extends hipBLASLt backend support for F32 (SGEMM) operations on both gfx942 and gfx950 architectures. Previously:
This change enables F32 GEMM to leverage the hipBLASLt backend for improved performance on:
This follows the pattern established in #13045, which defaulted various data types to hipBLASLt on specific architectures.
Technical Details
Changes to
tensile_host.cppThe
useHipBLASLt()function has been modified:gfx950 architecture (lines 1295-1301):
falsefor F32 types, forcing Tensile backendprob.handle->tryHipBLASLt()gfx942 architecture (lines 1303-1333):
only_MI300Acategory, restricting it to 228 CU devices onlyMI300A_or_MI300Xcategory alongside fp16, bf16, and fp64only_MI300AcategoryChanges to
CHANGELOG.mdUpdated the rocBLAS 5.8.0 changelog to reflect the new defaults:
f32_r" to "all input types"f32_rto the list of types enabled on both MI300A and MI300XKey Implementation Details
Impact
Architectures Affected
Backward Compatibility
ROCBLAS_USE_HIPBLASLTenvironment variable continues to force or disable the hipBLASLt backend, providing an override mechanism