[llvm] [LLVM][NVPTX]Add BF16 vector instruction and fix lowering rules (PR #69415)

[Artem Belevich via llvm-commits]

================
@@ -436,6 +436,16 @@ NVPTXTargetLowering::NVPTXTargetLowering(const NVPTXTargetMachine &TM,
       case ISD::FADD:
       case ISD::FMUL:
       case ISD::FSUB:
+      case ISD::SELECT:
+      case ISD::SELECT_CC:
+      case ISD::SETCC:
+      case ISD::FEXP2:
+      // conversion from bf16 to bf16 only support in sm90
----------------
Artem-B wrote:

The comment is redundant -- all ops in this switch statement are available on sm_90 only, and there's a comment about that at the beginning of the switch.

https://github.com/llvm/llvm-project/pull/69415