[llvm] [AArch64] Improve lowering of GPR zeroing in copyPhysReg (PR #163059)

[David Tellenbach via llvm-commits]

================
@@ -39,7 +39,7 @@ body:             |
     ; CHECK-NO-ZCM-GPR32-ZCM-GPR64-NO-ZCZ-LABEL: name: f0
     ; CHECK-NO-ZCM-GPR32-ZCM-GPR64-NO-ZCZ: liveins: $x0, $lr
     ; CHECK-NO-ZCM-GPR32-ZCM-GPR64-NO-ZCZ-NEXT: {{  $}}
-    ; CHECK-NO-ZCM-GPR32-ZCM-GPR64-NO-ZCZ-NEXT: $x0 = ORRXrr $xzr, undef $xzr, implicit $wzr
+    ; CHECK-NO-ZCM-GPR32-ZCM-GPR64-NO-ZCZ-NEXT: $w0 = ORRWrr $wzr, $wzr
----------------
dtellenbach wrote:

Can we expand the testing a bit and check for different combinations of 32 and 64bit zcz? It's currently all or nothing because the test was added before splitting the feature.

https://github.com/llvm/llvm-project/pull/163059