[llvm] [RISCV] Split LMUL=8 f16 fixed vector (s/u)ittofp/fpto(s/u)i before promoting. (PR #191568)

[Luke Lau via llvm-commits]

================
@@ -8062,12 +8062,12 @@ SDValue RISCVTargetLowering::LowerOperation(SDValue Op,
           (Subtarget.hasVInstructionsF16Minimal() &&
            !Subtarget.hasVInstructionsF16())) ||
          Op.getValueType().getScalarType() == MVT::bf16)) {
-      if (isPromotedOpNeedingSplit(Op, Subtarget))
----------------
lukel97 wrote:

Should we just be checking if the container type is `nxv32f16`/`nxv32bf16` in `isPromotedOpNeedingSplit`? I'm not sure if the other callers are relying on it for fixed length vectors too

https://github.com/llvm/llvm-project/pull/191568