[llvm] [DAGCombiner][AMDGPU] Track signedness in ByteProviders (PR #65995)

David Stuttard via llvm-commits llvm-commits at lists.llvm.org
Tue Sep 12 07:39:22 PDT 2023

dstutt wrote:

> I should say this code looks a bit strange, since we are mixing types (signed / unsigned) in the mul operation. In general, I would think the unsigned type "wins" as this is the spirit of the c++ standard, but, since that was causing failures, I have just disabled the combine in such cases.

It's not C++ - it's a Vulkan operation https://registry.khronos.org/SPIR-V/specs/unified1/SPIRV.html#OpSUDot that's used in the failing test.


More information about the llvm-commits mailing list