[llvm] [X86] Fold FMSUBADD into VFCMULC for fp16 complex conjugate multiply (PR #224533)

[Rohan Shenoy via llvm-commits]

================
@@ -0,0 +1,53 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 6
+; RUN: llc < %s -mtriple=x86_64-unknown-unknown -mattr=avx512fp16,avx512vl | FileCheck %s
+
+define <8 x half> @test_128(<8 x half> %a, <8 x half> %b) {
+; CHECK-LABEL: test_128:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    vfcmulcph %xmm1, %xmm0, %xmm2
+; CHECK-NEXT:    vmovaps %xmm2, %xmm0
----------------
222rohan wrote:

@RKSimon  looks like in the tablegen it has an `@earlyclobber $dst`marked - not allowing it to use inputs

https://github.com/llvm/llvm-project/pull/224533