[AMDGPU] Fixed addReg() in SIOptimizeExecMaskingPreRA.cpp

The second argument is flags, not subreg.

Differential Revision: https://reviews.llvm.org/D61031

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@359017 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp b/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
index c8a146741ec57b2f35e84de277a14f7b64abe488..6340615244c5a422f65b07164c911c54586775a9 100644 (file)
--- a/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
+++ b/lib/Target/AMDGPU/SIOptimizeExecMaskingPreRA.cpp
@@ -246,7 +246,7 @@ static unsigned optimizeVcndVcmpPair(MachineBasicBlock &MBB,
   MachineInstr *Andn2 = BuildMI(MBB, *And, And->getDebugLoc(),
                                 TII->get(Andn2Opc), And->getOperand(0).getReg())
                             .addReg(ExecReg)
-                            .addReg(CCReg, CC->getSubReg());
+                            .addReg(CCReg, 0, CC->getSubReg());
   And->eraseFromParent();
   LIS->InsertMachineInstrInMaps(*Andn2);
 

diff --git a/test/CodeGen/AMDGPU/optimize-negated-cond-exec-masking.mir b/test/CodeGen/AMDGPU/optimize-negated-cond-exec-masking.mir
index a9b8d94b59eee67f7423717d1dd12a8773593db5..4986f5153b6eaffe11f6c4b9221dc2bd183e13ac 100644 (file)
--- a/test/CodeGen/AMDGPU/optimize-negated-cond-exec-masking.mir
+++ b/test/CodeGen/AMDGPU/optimize-negated-cond-exec-masking.mir
@@ -463,3 +463,25 @@ body:             |
   bb.4:
     S_ENDPGM 0
 ...
+
+# GCN: name: negated_cond_subreg
+# GCN:      %0.sub0_sub1:sreg_128 = IMPLICIT_DEF
+# GCN-NEXT: $vcc = S_ANDN2_B64 $exec, %0.sub0_sub1, implicit-def $scc
+# GCN-NEXT: S_CBRANCH_VCCNZ %bb.2, implicit $vcc
+---
+name:            negated_cond_subreg
+body:             |
+  bb.0:
+    %0.sub0_sub1:sreg_128 = IMPLICIT_DEF
+    %1:vgpr_32 = V_CNDMASK_B32_e64 0, 0, 0, 1, %0.sub0_sub1, implicit $exec
+    %2.sub0_sub1:sreg_128 = V_CMP_NE_U32_e64 %1, 1, implicit $exec
+    $vcc = S_AND_B64 $exec, killed %2.sub0_sub1:sreg_128, implicit-def dead $scc
+    S_CBRANCH_VCCNZ %bb.2, implicit killed $vcc
+    S_BRANCH %bb.1
+
+  bb.1:
+    S_BRANCH %bb.0
+
+  bb.2:
+    S_ENDPGM 0
+...