[llvm] [X86] X86FixupInstTuning - fold BLENDPS -> MOVSD (PR #144029)

[Phoebe Wang via llvm-commits]

================
@@ -235,14 +236,16 @@ bool X86FixupInstTuningPass::processInstruction(
 
   switch (Opc) {
   case X86::BLENDPDrri:
-    return ProcessBLENDToMOV(X86::MOVSDrr);
+    return ProcessBLENDToMOV(X86::MOVSDrr, 0x3, 0x1);
   case X86::VBLENDPDrri:
-    return ProcessBLENDToMOV(X86::VMOVSDrr);
+    return ProcessBLENDToMOV(X86::VMOVSDrr, 0x3, 0x1);
 
   case X86::BLENDPSrri:
-    return ProcessBLENDToMOV(X86::MOVSSrr);
+    return ProcessBLENDToMOV(X86::MOVSSrr, 0xFF, 0x1) ||
+           ProcessBLENDToMOV(X86::MOVSDrr, 0xFF, 0x3);
   case X86::VBLENDPSrri:
-    return ProcessBLENDToMOV(X86::VMOVSSrr);
+    return ProcessBLENDToMOV(X86::VMOVSSrr, 0xFF, 0x1) ||
----------------
phoebewang wrote:

I see. The 0xFF makes me thinking of it :)

https://github.com/llvm/llvm-project/pull/144029