================ ---------------- MadhurKumar004 wrote: The changes are just specific to f32 and makes no changes in half/bfloat/double. Would you suggest fixing codegen for other types too or should i just push the f32 ones? https://github.com/llvm/llvm-project/pull/207178