[rocBLAS] Fix issues with rocBLAS 4.0.0 #448
Open
+33
−1
Conversation
This file contains 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
nilsfriess
force-pushed
the
rocblas-4.0.0-fixes
branch
from
5cbf759 to
643f610
Compare
|
@nilsfriess Thanks for the PR. I think this needs to be tested. You can contact me or infra team ( @toxicscum ) to access AMD GPUs. |
npmiller
reviewed
Jun 5, 2024
| @@ -381,10 +381,19 @@ inline void trmm(Func func, sycl::queue &queue, side left_right, uplo upper_lowe | |||
| auto a_ = sc.get_mem<rocDataType *>(a_acc); | |||
| auto b_ = sc.get_mem<rocDataType *>(b_acc); | |||
| rocblas_status err; | |||
|
|
|||
| // rocblas version 4.0.0 removed the legacy BLAS trmm implementation | |||
| #ifdef ROCBLAS_NO_LEGACY_TRMM | |||
There was a problem hiding this comment.
Suggested change
| #ifdef ROCBLAS_NO_LEGACY_TRMM | |
| #if ROCBLAS_VERSION_MAJOR >= 3 |
We can use ROCBLAS_VERSION_MAJOR, that way we don't need to add a new define, also since it's deprecated in rocBLAS 3 it gives warning so we could also use the new interface for rocBLAS 3.x.
I ran into the same issue and ended up with a similar patch before finding this PR. I didn't have any ROCBLAS_LIB_PATH issues with rocBLAS 4.x though.
|
Thank you for the PR, we have recently removed the |
2 tasks
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Outdated suggestions cannot be applied.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.
Description
rocBLAS 4.0.0 removed support for the legacy BLAS inplace trmm, so calls to trmm cause a compilation error. The old functionality is provided by just duplicating the last two arguments of trmm (see also here).
To fix this, I am defining a macro
ROCBLAS_NO_LEGACY_TRMMdepending on the rocBLAS version which is then used in an#ifdefto decide which version of the function to call.I also had to change slightly the way the path to
librocblas.sois constructed inFindrocBLAS.cmake. I'm not sure if this is also related to the rocBLAS version or if this is a general issue on some systems, so I can also put that in a separated PR if you prefer.Checklist
All Submissions
I don't have an AMD gpu to test, I'm just assessing the state of compiling oneMKL with AdaptiveCpp to see if [hipSYCL] Add hipSYCL GitHub CI #307 needs any updates and can the maybe be merged.