[llvm] [X86] Do not directly fold for VINSERTPS (PR #65718)

[Phoebe Wang via llvm-commits]

================
@@ -43,6 +43,8 @@ NOFOLD(VEXTRACTI32x8Zrrk)
 NOFOLD(VEXTRACTI64x2Z256rrk)
 NOFOLD(VEXTRACTI64x2Zrrk)
 NOFOLD(VEXTRACTI64x4Zrrk)
+NOFOLD(VINSERTPSZrr)
----------------
phoebewang wrote:

Good point, I didn't notice it before

https://github.com/llvm/llvm-project/pull/65718