[llvm] [X86] Support APX CMOV/CFCMOV instructions (PR #82592)

[via llvm-commits]

================
@@ -298,9 +298,9 @@ define i8 @addflag8rr(i8 noundef %a, i8 noundef %b) {
 ; CHECK-LABEL: addflag8rr:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    addb %sil, %dil, %al # encoding: [0x62,0xf4,0x7c,0x18,0x00,0xf7]
-; CHECK-NEXT:    movzbl %al, %ecx # encoding: [0x0f,0xb6,0xc8]
-; CHECK-NEXT:    movl $255, %eax # encoding: [0xb8,0xff,0x00,0x00,0x00]
-; CHECK-NEXT:    cmovael %ecx, %eax # encoding: [0x0f,0x43,0xc1]
+; CHECK-NEXT:    movzbl %al, %eax # encoding: [0x0f,0xb6,0xc0]
+; CHECK-NEXT:    movl $255, %ecx # encoding: [0xb9,0xff,0x00,0x00,0x00]
+; CHECK-NEXT:    cmovbl %ecx, %eax # EVEX TO LEGACY Compression encoding: [0x0f,0x42,0xc1]
----------------
XinWang10 wrote:

Might because RA? Value in regs also changed, so the result is correct.

https://github.com/llvm/llvm-project/pull/82592