AMDGPU: Fix trying to skip from a block with no successors

Found while reducing bug 28550

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@275509 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/SILowerControlFlow.cpp b/lib/Target/AMDGPU/SILowerControlFlow.cpp
index adb0919231c30493b9d1de7a1c3042cc08eb4e77..d82f26ac33a6ac5c361e03ef818390e83d71ed80 100644 (file)
--- a/lib/Target/AMDGPU/SILowerControlFlow.cpp
+++ b/lib/Target/AMDGPU/SILowerControlFlow.cpp
@@ -158,6 +158,8 @@ static bool opcodeEmitsNoInsts(unsigned Opc) {
 
 bool SILowerControlFlow::shouldSkip(MachineBasicBlock *From,
                                     MachineBasicBlock *To) {
+  if (From->succ_empty())
+    return false;
 
   unsigned NumInstr = 0;
   MachineFunction *MF = From->getParent();
@@ -217,7 +219,7 @@ bool SILowerControlFlow::skipIfDead(MachineInstr &MI, MachineBasicBlock &NextBB)
     return false;
 
   MachineBasicBlock *SkipBB = insertSkipBlock(MBB, MI.getIterator());
-  SkipBB->addSuccessor(&NextBB);
+  MBB.addSuccessor(SkipBB);
 
   const DebugLoc &DL = MI.getDebugLoc();
 
@@ -493,7 +495,6 @@ MachineBasicBlock *SILowerControlFlow::insertSkipBlock(
   ++MBBI;
 
   MF->insert(MBBI, SkipBB);
-  MBB.addSuccessor(SkipBB);
 
   return SkipBB;
 }

diff --git a/test/CodeGen/AMDGPU/skip-if-dead.ll b/test/CodeGen/AMDGPU/skip-if-dead.ll
index 26733bab04c308074c1176110bbb8781ce1c07a8..a1ffba545a427392be11069e001b0a3ffc8723c8 100644 (file)
--- a/test/CodeGen/AMDGPU/skip-if-dead.ll
+++ b/test/CodeGen/AMDGPU/skip-if-dead.ll
@@ -298,6 +298,44 @@ end:
   ret void
 }
 
+; CHECK-LABEL: {{^}}no_skip_no_successors:
+; CHECK: v_cmp_nle_f32
+; CHECK: s_and_b64 vcc, exec,
+; CHECK: s_cbranch_vccz [[SKIPKILL:BB[0-9]+_[0-9]+]]
+
+; CHECK: ; BB#3: ; %bb6
+; CHECK: s_mov_b64 exec, 0
+
+; CHECK: [[SKIPKILL]]:
+; CHECK: v_cmp_nge_f32
+; CHECK: s_and_b64 vcc, exec, vcc
+; CHECK: s_cbranch_vccz [[UNREACHABLE:BB[0-9]+_[0-9]+]]
+
+; CHECK: [[UNREACHABLE]]:
+; CHECK-NEXT: .Lfunc_end{{[0-9]+}}
+define amdgpu_ps void @no_skip_no_successors(float inreg %arg, float inreg %arg1) #0 {
+bb:
+  %tmp = fcmp ult float %arg1, 0.000000e+00
+  %tmp2 = fcmp ult float %arg, 0x3FCF5C2900000000
+  br i1 %tmp, label %bb6, label %bb3
+
+bb3:                                              ; preds = %bb
+  br i1 %tmp2, label %bb5, label %bb4
+
+bb4:                                              ; preds = %bb3
+  br i1 true, label %bb5, label %bb7
+
+bb5:                                              ; preds = %bb4, %bb3
+  unreachable
+
+bb6:                                              ; preds = %bb
+  call void @llvm.AMDGPU.kill(float -1.000000e+00)
+  unreachable
+
+bb7:                                              ; preds = %bb4
+  ret void
+}
+
 declare void @llvm.AMDGPU.kill(float) #0
 
 attributes #0 = { nounwind }