[llvm] [DAG] SimplifyDemandedVectorElts - add handling for INT<->FP conversions (PR #117884)

[Simon Pilgrim via llvm-commits]

================
@@ -141,56 +141,61 @@ declare <8 x half> @llvm.ldexp.v8f16.v8i16(<8 x half>, <8 x i16>)
 define <8 x half> @fmul_pow2_8xhalf(<8 x i16> %i) {
 ; CHECK-SSE-LABEL: fmul_pow2_8xhalf:
 ; CHECK-SSE:       # %bb.0:
-; CHECK-SSE-NEXT:    subq $88, %rsp
-; CHECK-SSE-NEXT:    .cfi_def_cfa_offset 96
+; CHECK-SSE-NEXT:    subq $104, %rsp
----------------
RKSimon wrote:

Minor regression due to complexity the scalarisation of the f16 conversions - I'm not convinced this is a showstopper tbh.

https://github.com/llvm/llvm-project/pull/117884