3 files changed, 31 insertions, 3 deletions
diff --git a/llvm/lib/Target/AMDGPU/SIInsertSkips.cpp b/llvm/lib/Target/AMDGPU/SIInsertSkips.cpp
index f23fa02bf8a..ba21a5ce129 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertSkips.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertSkips.cpp
@@ -476,7 +476,7 @@ bool SIInsertSkips::runOnMachineFunction(MachineFunction &MF) {
         kill(MI);
 
         if (ExecBranchStack.empty()) {
-          if (skipIfDead(MI, *NextBB)) {
+          if (NextBB != BE && skipIfDead(MI, *NextBB)) {
             HaveSkipBlock = true;
             NextBB = std::next(BI);
             BE = MF.end();
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index eb04338d4d3..7aef9fde480 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -1632,7 +1632,34 @@ bool SIInstrInfo::analyzeBranch(MachineBasicBlock &MBB, MachineBasicBlock *&TBB,
                                 SmallVectorImpl<MachineOperand> &Cond,
                                 bool AllowModify) const {
   MachineBasicBlock::iterator I = MBB.getFirstTerminator();
-  if (I == MBB.end())
+  auto E = MBB.end();
+  if (I == E)
+    return false;
+
+  // Skip over the instructions that are artificially terminators for special
+  // exec management.
+  while (I != E && !I->isBranch() && !I->isReturn() &&
+         I->getOpcode() != AMDGPU::SI_MASK_BRANCH) {
+    switch (I->getOpcode()) {
+    case AMDGPU::SI_MASK_BRANCH:
+    case AMDGPU::S_MOV_B64_term:
+    case AMDGPU::S_XOR_B64_term:
+    case AMDGPU::S_ANDN2_B64_term:
+      break;
+    case AMDGPU::SI_IF:
+    case AMDGPU::SI_ELSE:
+    case AMDGPU::SI_KILL_I1_TERMINATOR:
+    case AMDGPU::SI_KILL_F32_COND_IMM_TERMINATOR:
+      // FIXME: It's messy that these need to be considered here at all.
+      return true;
+    default:
+      llvm_unreachable("unexpected non-branch terminator inst");
+    }
+
+    ++I;
+  }
+
+  if (I == E)
     return false;
 
   if (I->getOpcode() != AMDGPU::SI_MASK_BRANCH)
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index 11136e0a1df..9b0edf4c4e0 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -247,7 +247,7 @@ def SI_LOOP : CFPseudoInstSI <
   (outs), (ins SReg_64:$saved, brtarget:$target),
   [(AMDGPUloop i64:$saved, bb:$target)], 1, 1> {
   let Size = 8;
-  let isBranch = 0;
+  let isBranch = 1;
   let hasSideEffects = 1;
 }
 
@@ -307,6 +307,7 @@ def SI_ILLEGAL_COPY : SPseudoInstSI <
 def SI_BR_UNDEF : SPseudoInstSI <(outs), (ins sopp_brtarget:$simm16)> {
   let isTerminator = 1;
   let usesCustomInserter = 1;
+  let isBranch = 1;
 }
 
 def SI_PS_LIVE : PseudoInstSI <