Fused kernel support for AMD (using JIT) #1156
Closed
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Outdated suggestions cannot be applied.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.
Overview
This PR aims to bring JIT support into building fused kernels. The main reason for the swtich is that hip extension acts weird (making ninja generates different rules for the same target) with the traditional
CUDAExtension
setup, thus in this PR we usecpp_extension.load()
, this enables no code modification when building across AMD and NVIDIA platformsHeads-up
This PR also bring in the
cpp_extention_load_helper
from here, note that we JIT load fused kernels on demand and depends onneox_args
, this may introduce additional complexity in testing, for fused kernels testing, I recommend keeping the originalsetup.py
file as a backup.Aim to close #954