[llvm] [NVPTX] Select bfloat16 add/mul/sub as fma on SM80 (PR #121065)

[via llvm-commits]

peterbell10 wrote:

> PTX < 7.8 (otherwise FMUL is available, and we don't need the patch).

This is where your logic doesn't make sense. Why would PTX >= 7.8 mean that sm_80 doesn't benefit from this patch.

https://github.com/llvm/llvm-project/pull/121065