[hipBLASLt] Enable MX data generation for Tensile host and support calling Tensile MX kernel#4599
Merged
Merged
Conversation
bnemanich
reviewed
Feb 18, 2026
| tensileProblem.setSwizzleTensorA(prob.swizzleA); | ||
| tensileProblem.setSwizzleTensorB(prob.swizzleB); | ||
|
|
||
| if(prob.scaleAType == RocblasltContractionProblem::ScalingFormat::Block_32_UE8M0) |
Contributor
There was a problem hiding this comment.
This should also handle Block_32_UE8M0_32_8_EXT
Contributor
Author
There was a problem hiding this comment.
Added Block_32_UE8M0_32_8_EXT, thanks!
| // NOTE: an assumption here is A & B must be both MX data types or non-MX data types. | ||
| // Mixing is not supported. | ||
| if(!problemType.useScaleAB.empty() or | ||
| (problemType.mxBlockA == 32 && problemType.mxBlockB == 32)) //kernel input data |
Contributor
There was a problem hiding this comment.
Maybe should check that mxBlockA != 0 instead.
Contributor
Author
There was a problem hiding this comment.
Changed to !=0 instead of ==32, thanks!
bnemanich
approved these changes
Feb 20, 2026
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
2 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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 enables using mxDataGenerator when Tensile is the host and supports calling FP4 kernels generated from Tensile.
Technical Details
Test Plan
Use cmake preset build with rocRoller host off (i.e., use Tensile as host) , gpu target set to gfx950 and
-DBUILD_TESTING:BOOL=OFF(turn off tensileLite test which will error out during build)Use
hipblaslt-test./clients/hipblaslt-test --gtest_filter=*matmul_tensile_fp4*Use
hipblaslt-bench./clients/hipblaslt-bench --iters 0 --cold_iters 0 --transA T --transB N --a_type f4_r --b_type f4_r --c_type f32_r --d_type f32_r -m 256 -n 256 -k 256 --alpha 2.1 --beta 0.7 --scaleA 3 --scaleB 3 --scale_type f32_r --verifyTest Result
Submission Checklist