[llvm] 28a3fbb - [X86][test] Pre-commit test for disabling NDD2NonNDD transform for 8/16-bit ops

[Shengchen Kan via llvm-commits]

Author: Shengchen Kan
Date: 2024-06-26T15:07:53+08:00
New Revision: 28a3fbbe8c6cd5b92fb0a2a3cc21a836a381e086

URL: https://github.com/llvm/llvm-project/commit/28a3fbbe8c6cd5b92fb0a2a3cc21a836a381e086
DIFF: https://github.com/llvm/llvm-project/commit/28a3fbbe8c6cd5b92fb0a2a3cc21a836a381e086.diff

LOG: [X86][test] Pre-commit test for disabling NDD2NonNDD transform for 8/16-bit ops

Added: 
    

Modified: 
    llvm/test/CodeGen/X86/apx/compress-evex.mir

Removed: 
    


################################################################################
diff  --git a/llvm/test/CodeGen/X86/apx/compress-evex.mir b/llvm/test/CodeGen/X86/apx/compress-evex.mir
index 5a59ab0f8a9d0..7723cac8eb326 100644
--- a/llvm/test/CodeGen/X86/apx/compress-evex.mir
+++ b/llvm/test/CodeGen/X86/apx/compress-evex.mir
@@ -127,3 +127,14 @@ body:             |
     $rbx = CFCMOV64rm_ND $rbx, $rax, 1, $noreg, 24, $noreg, 8, implicit $eflags
     RET64 $rax
 ...
+---
+name:            no_convert_8bit_16bit
+body:             |
+  bb.0.entry:
+    liveins: $di, $si, $ax
+    ; CHECK: addb    %sil, %dil                      # EVEX TO LEGACY Compression encoding: [0x40,0x00,0xf7]
+    ; CHECK: xorw    %di, %ax                        # EVEX TO LEGACY Compression encoding: [0x66,0x31,0xf8]
+    $dil = ADD8rr_ND $dil, killed $sil, implicit-def dead $eflags
+    $ax = XOR16rr_ND $ax, killed $di, implicit-def dead $eflags
+    RET64 $rax
+...