hip : fix warp mask width for rocWMMA compatibility #15239
Closed
Conversation
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
github-actions
bot
added
Nvidia GPU
|
When I grep rocWMMA for "shfl" I am not finding anything. |
|
dup of #15241 please see discussion in that pr. |
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Outdated suggestions cannot be applied.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.
I ran into issues compiling w/
-DGGML_HIP_ROCWMMA_FATTN=ONwith the latest TheRock/ROCm (7.0) nightly releases. This appears to be due to a warp mask width incompatibility recently introduced between ROCm's rocWMMA library and CUDA-style sync code.ROCm's rocWMMA library recently added its own __shfl_sync and __shfl_xor_sync functions but it also requires 64-bit masks while the existing code uses hardcoded 32-bit masks (0xFFFFFFFF). This causes type conflicts and compilation failures when building with rocWMMA support enabled.
I tried fixing w/ just messing w/ etc just in hip.h and leaving the CUDA files alone but I think there really isn't a clean alternative of replacing the hard-coded masks?