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Outdated suggestions cannot be applied.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.
numba-dpex does not pin the max-spirv-version that
llvm-spirv
generates and letsllvm-spriv
decide what SPIR-V code to generate based on the LLVM-IR generated by the Numba front-end.The behaviour is causing an issue with oneAPI dpcpp 2023.0 (Refer #868). The issue arises when a pre-compiled SPIR-V binary produced by compiling an OpenCL program implementing software atomic add for floating point values is linked to the SPIR-V module generated by JIT compilation inside numba-dpex.
The PR implements a workaround by setting the
spirv-max-version
flags to 1.1 when generating SPIR-V both for the pre-compiled OpenCL program and inside numba-dpex. SPIR-V version 1.1 is used as it seems to be the default used by dpcpp.