PTX: Default to 32-bit indexing of pointers. #444
Merged
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Outdated suggestions cannot be applied.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.
This change sets the default index size for address calculations in PTX code to 32bit, in order to avoid extensions of 32-bit indices to 64-bit as part of instcombine when generating new GEP instructions (see JuliaLLVM/LLVM.jl#342 (comment)).
I'm not entirely sure this is correct, but it looks safe. It doesn't limit address calculations to 4GB, as you can still use 64-bit indices, but just defaults to 32-bit as these are generally more GPU friendly.
FWIW, the NVPTX back-end docs leave the index size undefined, and have it default to 64 bits.
x-ref JuliaGPU/CUDA.jl#1895
cc @vchuravy