[AArch64] Widen GPR32 zero cycle zeroing

Given a GPR32 zeroing instruction, if the target supports zero cycle zeroing for GPR64 but not for GPR32, widen the instruction to 64 bit `$xn = MOVZXi 0, 0` instead of writing to `$wn` to exploit zero cycle zeroing.

It also aligns naming in the generic zeroing test.

Author: tomershafir

llvm/lib/Target/AArch64/AArch64InstrInfo.cpp

@@ -5116,7 +5116,15 @@ void AArch64InstrInfo::copyPhysReg(MachineBasicBlock &MBB,
 
   // GPR32 zeroing
   if (AArch64::GPR32spRegClass.contains(DestReg) && SrcReg == AArch64::WZR) {
-    if (Subtarget.hasZeroCycleZeroingGPR32()) {
+    if (Subtarget.hasZeroCycleZeroingGPR64() &&
+        !Subtarget.hasZeroCycleZeroingGPR32()) {
+      MCRegister DestRegX = RI.getMatchingSuperReg(DestReg, AArch64::sub_32,
+                                                   &AArch64::GPR64spRegClass);
+      assert(DestRegX.isValid() && "Destination super-reg not valid");
+      BuildMI(MBB, I, DL, get(AArch64::MOVZXi), DestRegX)
+          .addImm(0)
+          .addImm(AArch64_AM::getShifterImm(AArch64_AM::LSL, 0));
+    } else if (Subtarget.hasZeroCycleZeroingGPR32()) {
       BuildMI(MBB, I, DL, get(AArch64::MOVZWi), DestReg)
           .addImm(0)
           .addImm(AArch64_AM::getShifterImm(AArch64_AM::LSL, 0));

llvm/test/CodeGen/AArch64/arm64-copy-phys-zero-reg.mir

@@ -35,7 +35,7 @@ body:             |
     ; CHECK-NOZCZ-GPR32-ZCZ-GPR64-LABEL: name: f0
     ; CHECK-NOZCZ-GPR32-ZCZ-GPR64: liveins: $x0, $lr
     ; CHECK-NOZCZ-GPR32-ZCZ-GPR64-NEXT: {{  $}}
-    ; CHECK-NOZCZ-GPR32-ZCZ-GPR64-NEXT: $w0 = ORRWrr $wzr, $wzr
+    ; CHECK-NOZCZ-GPR32-ZCZ-GPR64-NEXT: $x0 = MOVZXi 0, 0
     ; CHECK-NOZCZ-GPR32-ZCZ-GPR64-NEXT: BL @f2, csr_darwin_aarch64_aapcs, implicit-def dead $lr, implicit $sp, implicit $w0, implicit-def $sp, implicit-def $w0
     ;
     ; CHECK-ZCZ-GPR32-ZCZ-GPR64-LABEL: name: f0

llvm/test/CodeGen/AArch64/arm64-zero-cycle-zeroing-gpr.ll

@@ -1,41 +1,44 @@
-; RUN: llc < %s -mtriple=aarch64-linux-gnu | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR
+; RUN: llc < %s -mtriple=aarch64-linux-gnu | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR32-NOZCZ-GPR64
 ; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz-gpr32 | FileCheck %s -check-prefixes=ALL,ZCZ-GPR32
-; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz-gpr64 | FileCheck %s -check-prefixes=ALL,ZCZ-GPR64
-; RUN: llc < %s -mtriple=arm64-apple-macosx -mcpu=generic | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR
+; RUN: llc < %s -mtriple=aarch64-linux-gnu -mattr=+zcz-gpr64 | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR32-ZCZ-GPR64
+; RUN: llc < %s -mtriple=arm64-apple-macosx -mcpu=generic | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR32-NOZCZ-GPR64
 ; RUN: llc < %s -mtriple=arm64-apple-ios -mcpu=cyclone | FileCheck %s -check-prefixes=ALL,ZCZ-GPR32,ZCZ-GPR64
 ; RUN: llc < %s -mtriple=arm64-apple-macosx -mcpu=apple-m1 | FileCheck %s -check-prefixes=ALL,ZCZ-GPR32,ZCZ-GPR64
-; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=exynos-m3 | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR
+; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=exynos-m3 | FileCheck %s -check-prefixes=ALL,NOZCZ-GPR32-NOZCZ-GPR64
 ; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=kryo | FileCheck %s -check-prefixes=ALL,ZCZ-GPR32,ZCZ-GPR64
 ; RUN: llc < %s -mtriple=aarch64-linux-gnu -mcpu=falkor | FileCheck %s -check-prefixes=ALL,ZCZ-GPR32,ZCZ-GPR64
 
 define i8 @ti8() {
 entry:
 ; ALL-LABEL: ti8:
-; NOZCZ-GPR: mov w0, wzr
+; NOZCZ-GPR32-NOZCZ-GPR64: mov w0, wzr
 ; ZCZ-GPR32: mov w0, #0
+; NOZCZ-GPR32-ZCZ-GPR64: mov x0, #0
   ret i8 0
 }
 
 define i16 @ti16() {
 entry:
 ; ALL-LABEL: ti16:
-; NOZCZ-GPR: mov w0, wzr
+; NOZCZ-GPR32-NOZCZ-GPR64: mov w0, wzr
 ; ZCZ-GPR32: mov w0, #0
+; NOZCZ-GPR32-ZCZ-GPR64: mov x0, #0
   ret i16 0
 }
 
 define i32 @ti32() {
 entry:
 ; ALL-LABEL: ti32:
-; NOZCZ-GPR: mov w0, wzr
+; NOZCZ-GPR32-NOZCZ-GPR64: mov w0, wzr
 ; ZCZ-GPR32: mov w0, #0
+; NOZCZ-GPR32-ZCZ-GPR64: mov x0, #0
   ret i32 0
 }
 
 define i64 @ti64() {
 entry:
 ; ALL-LABEL: ti64:
-; NOZCZ-GPR: mov x0, xzr
+; NOZCZ-GPR32-NOZCZ-GPR64 mov x0, xzr
 ; ZCZ-GPR64: mov x0, #0
   ret i64 0
 }