[llvm] 5a8d2dd - [AMDGPU] Handle subregisters properly in generic operand legalizer (#108496)

[via llvm-commits]

Author: Aditi Medhane
Date: 2024-09-18T13:14:49+05:30
New Revision: 5a8d2dd1f97017253be5d2262b25a9cf9d002546

URL: https://github.com/llvm/llvm-project/commit/5a8d2dd1f97017253be5d2262b25a9cf9d002546
DIFF: https://github.com/llvm/llvm-project/commit/5a8d2dd1f97017253be5d2262b25a9cf9d002546.diff

LOG: [AMDGPU] Handle subregisters properly in generic operand legalizer (#108496)

Fix for the issue found during COPY introduction during legalization of
PHI operands for sgpr to vgpr copy when subreg is involved.

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
    llvm/test/CodeGen/AMDGPU/phi-vgpr-input-moveimm.mir

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index e4a679f6a3ef8f..30aa36be99c95f 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -6231,10 +6231,9 @@ void SIInstrInfo::legalizeGenericOperand(MachineBasicBlock &InsertMBB,
     return;
 
   Register DstReg = MRI.createVirtualRegister(DstRC);
-  auto Copy = BuildMI(InsertMBB, I, DL, get(AMDGPU::COPY), DstReg).add(Op);
-
+  auto Copy =
+      BuildMI(InsertMBB, I, DL, get(AMDGPU::COPY), DstReg).addReg(OpReg);
   Op.setReg(DstReg);
-  Op.setSubReg(0);
 
   MachineInstr *Def = MRI.getVRegDef(OpReg);
   if (!Def)

diff  --git a/llvm/test/CodeGen/AMDGPU/phi-vgpr-input-moveimm.mir b/llvm/test/CodeGen/AMDGPU/phi-vgpr-input-moveimm.mir
index dab4c9d401407b..d21dbd290accea 100644
--- a/llvm/test/CodeGen/AMDGPU/phi-vgpr-input-moveimm.mir
+++ b/llvm/test/CodeGen/AMDGPU/phi-vgpr-input-moveimm.mir
@@ -73,13 +73,13 @@ body:             |
   ; GCN-NEXT:   successors: %bb.2(0x80000000)
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT:   [[S_ADD_U:%[0-9]+]]:sreg_64 = S_ADD_U64_PSEUDO [[COPY]], [[COPY1]], implicit-def $scc
-  ; GCN-NEXT:   [[COPY2:%[0-9]+]]:vreg_64 = COPY [[S_ADD_U]].sub0, implicit $exec
+  ; GCN-NEXT:   [[COPY2:%[0-9]+]]:vreg_64 = COPY [[S_ADD_U]], implicit $exec
   ; GCN-NEXT:   S_BRANCH %bb.2
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.2:
   ; GCN-NEXT:   successors: %bb.3(0x80000000)
   ; GCN-NEXT: {{  $}}
-  ; GCN-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI [[V_MOV_B64_e32_]].sub0, %bb.3, [[COPY2]], %bb.1
+  ; GCN-NEXT:   [[PHI:%[0-9]+]]:vgpr_32 = PHI [[V_MOV_B64_e32_]].sub0, %bb.3, [[COPY2]].sub0, %bb.1
   ; GCN-NEXT:   S_BRANCH %bb.3
   ; GCN-NEXT: {{  $}}
   ; GCN-NEXT: bb.3: