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

[via llvm-commits]

peterbell10 wrote:

The spec says
> `add{.rnd}.bf16` and `add{.rnd}.bf16x2` requires `sm_90` or higher.

I don't see any suggestion that that only applies for specific PTX versions.

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