Re-apply r124518 with fix. Watch out for invalidated iterator.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@124526 91177308-0d34-0410-b5e6-96231b3b80d8

13 files changed, 226 insertions, 73 deletions

diff --git a/include/llvm/Transforms/Utils/BasicBlockUtils.h b/include/llvm/Transforms/Utils/BasicBlockUtils.h
index 6de8b85070..5335860287 100644
--- a/include/llvm/Transforms/Utils/BasicBlockUtils.h
+++ b/include/llvm/Transforms/Utils/BasicBlockUtils.h
@@ -22,9 +22,10 @@
 
 namespace llvm {
 
+class AliasAnalysis;
 class Instruction;
 class Pass;
-class AliasAnalysis;
+class ReturnInst;
 
 /// DeleteDeadBlock - Delete the specified block, which must have no
 /// predecessors.
@@ -171,7 +172,15 @@ BasicBlock *SplitBlock(BasicBlock *Old, Instruction *SplitPt, Pass *P);
 BasicBlock *SplitBlockPredecessors(BasicBlock *BB, BasicBlock *const *Preds,
                                    unsigned NumPreds, const char *Suffix,
                                    Pass *P = 0);
-  
+
+/// FoldReturnIntoUncondBranch - This method duplicates the specified return
+/// instruction into a predecessor which ends in an unconditional branch. If
+/// the return instruction returns a value defined by a PHI, propagate the
+/// right value into the return. It returns the new return instruction in the
+/// predecessor.
+ReturnInst *FoldReturnIntoUncondBranch(ReturnInst *RI, BasicBlock *BB,
+                                       BasicBlock *Pred);
+
 } // End llvm namespace
 
 #endif
diff --git a/lib/CodeGen/TailDuplication.cpp b/lib/CodeGen/TailDuplication.cpp
index ce4b1be854..15aed3436c 100644
--- a/lib/CodeGen/TailDuplication.cpp
+++ b/lib/CodeGen/TailDuplication.cpp
@@ -465,9 +465,12 @@ TailDuplicatePass::TailDuplicate(MachineBasicBlock *TailBB, MachineFunction &MF,
     MaxDuplicateCount = TailDuplicateSize;
 
   if (PreRegAlloc) {
-      // Pre-regalloc tail duplication hurts compile time and doesn't help
-      // much except for indirect branches.
-    if (TailBB->empty() || !TailBB->back().getDesc().isIndirectBranch())
+    if (TailBB->empty())
+      return false;
+    const TargetInstrDesc &TID = TailBB->back().getDesc();
+    // Pre-regalloc tail duplication hurts compile time and doesn't help
+    // much except for indirect branches and returns.
+    if (!TID.isIndirectBranch() && !TID.isReturn())
       return false;
     // If the target has hardware branch prediction that can handle indirect
     // branches, duplicating them can often make them predictable when there
@@ -502,7 +505,7 @@ TailDuplicatePass::TailDuplicate(MachineBasicBlock *TailBB, MachineFunction &MF,
   }
   // Heuristically, don't tail-duplicate calls if it would expand code size,
   // as it's less likely to be worth the extra cost.
-  if (InstrCount > 1 && HasCall)
+  if (InstrCount > 1 && (PreRegAlloc && HasCall))
     return false;
 
   DEBUG(dbgs() << "\n*** Tail-duplicating BB#" << TailBB->getNumber() << '\n');
diff --git a/lib/Transforms/Scalar/TailRecursionElimination.cpp b/lib/Transforms/Scalar/TailRecursionElimination.cpp
index 23514fd51b..61f230804f 100644
--- a/lib/Transforms/Scalar/TailRecursionElimination.cpp
+++ b/lib/Transforms/Scalar/TailRecursionElimination.cpp
@@ -52,11 +52,13 @@
 
 #define DEBUG_TYPE "tailcallelim"
 #include "llvm/Transforms/Scalar.h"
+#include "llvm/Transforms/Utils/BasicBlockUtils.h"
 #include "llvm/Transforms/Utils/Local.h"
 #include "llvm/Constants.h"
 #include "llvm/DerivedTypes.h"
 #include "llvm/Function.h"
 #include "llvm/Instructions.h"
+#include "llvm/IntrinsicInst.h"
 #include "llvm/Pass.h"
 #include "llvm/Analysis/CaptureTracking.h"
 #include "llvm/Analysis/InlineCost.h"
@@ -64,7 +66,9 @@
 #include "llvm/Analysis/Loads.h"
 #include "llvm/Support/CallSite.h"
 #include "llvm/Support/CFG.h"
+#include "llvm/Support/Debug.h"
 #include "llvm/ADT/Statistic.h"
+#include "llvm/ADT/STLExtras.h"
 using namespace llvm;
 
 STATISTIC(NumEliminated, "Number of tail calls removed");
@@ -80,6 +84,18 @@ namespace {
     virtual bool runOnFunction(Function &F);
 
   private:
+    CallInst *FindTRECandidate(Instruction *I,
+                               bool CannotTailCallElimCallsMarkedTail);
+    bool EliminateRecursiveTailCall(CallInst *CI, ReturnInst *Ret,
+                                    BasicBlock *&OldEntry,
+                                    bool &TailCallsAreMarkedTail,
+                                    SmallVector<PHINode*, 8> &ArgumentPHIs,
+                                    bool CannotTailCallElimCallsMarkedTail);
+    bool FoldReturnAndProcessPred(BasicBlock *BB,
+                                  ReturnInst *Ret, BasicBlock *&OldEntry,
+                                  bool &TailCallsAreMarkedTail,
+                                  SmallVector<PHINode*, 8> &ArgumentPHIs,
+                                  bool CannotTailCallElimCallsMarkedTail);
     bool ProcessReturningBlock(ReturnInst *RI, BasicBlock *&OldEntry,
                                bool &TailCallsAreMarkedTail,
                                SmallVector<PHINode*, 8> &ArgumentPHIs,
@@ -136,7 +152,6 @@ bool TailCallElim::runOnFunction(Function &F) {
   bool TailCallsAreMarkedTail = false;
   SmallVector<PHINode*, 8> ArgumentPHIs;
   bool MadeChange = false;
-
   bool FunctionContainsEscapingAllocas = false;
 
   // CannotTCETailMarkedCall - If true, we cannot perform TCE on tail calls
@@ -163,10 +178,17 @@ bool TailCallElim::runOnFunction(Function &F) {
     return false;
 
   // Second pass, change any tail calls to loops.
-  for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
-    if (ReturnInst *Ret = dyn_cast<ReturnInst>(BB->getTerminator()))
-      MadeChange |= ProcessReturningBlock(Ret, OldEntry, TailCallsAreMarkedTail,
+  for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB) {
+    if (ReturnInst *Ret = dyn_cast<ReturnInst>(BB->getTerminator())) {
+      bool Change = ProcessReturningBlock(Ret, OldEntry, TailCallsAreMarkedTail,
                                           ArgumentPHIs,CannotTCETailMarkedCall);
+      if (!Change && BB->getFirstNonPHIOrDbg() == Ret)
+        Change = FoldReturnAndProcessPred(BB, Ret, OldEntry,
+                                          TailCallsAreMarkedTail, ArgumentPHIs,
+                                          CannotTCETailMarkedCall);
+      MadeChange |= Change;
+    }
+  }
 
   // If we eliminated any tail recursions, it's possible that we inserted some
   // silly PHI nodes which just merge an initial value (the incoming operand)
@@ -325,41 +347,47 @@ Value *TailCallElim::CanTransformAccumulatorRecursion(Instruction *I,
   return getCommonReturnValue(cast<ReturnInst>(I->use_back()), CI);
 }
 
-bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
-                                         bool &TailCallsAreMarkedTail,
-                                         SmallVector<PHINode*, 8> &ArgumentPHIs,
-                                       bool CannotTailCallElimCallsMarkedTail) {
-  BasicBlock *BB = Ret->getParent();
+static Instruction *FirstNonDbg(BasicBlock::iterator I) {
+  while (isa<DbgInfoIntrinsic>(I))
+    ++I;
+  return &*I;
+}
+
+CallInst*
+TailCallElim::FindTRECandidate(Instruction *TI,
+                               bool CannotTailCallElimCallsMarkedTail) {
+  BasicBlock *BB = TI->getParent();
   Function *F = BB->getParent();
 
-  if (&BB->front() == Ret) // Make sure there is something before the ret...
-    return false;
+  if (&BB->front() == TI) // Make sure there is something before the terminator.
+    return 0;
   
   // Scan backwards from the return, checking to see if there is a tail call in
   // this block.  If so, set CI to it.
-  CallInst *CI;
-  BasicBlock::iterator BBI = Ret;
-  while (1) {
+  CallInst *CI = 0;
+  BasicBlock::iterator BBI = TI;
+  while (true) {
     CI = dyn_cast<CallInst>(BBI);
     if (CI && CI->getCalledFunction() == F)
       break;
 
     if (BBI == BB->begin())
-      return false;          // Didn't find a potential tail call.
+      return 0;          // Didn't find a potential tail call.
     --BBI;
   }
 
   // If this call is marked as a tail call, and if there are dynamic allocas in
   // the function, we cannot perform this optimization.
   if (CI->isTailCall() && CannotTailCallElimCallsMarkedTail)
-    return false;
+    return 0;
 
   // As a special case, detect code like this:
   //   double fabs(double f) { return __builtin_fabs(f); } // a 'fabs' call
   // and disable this xform in this case, because the code generator will
   // lower the call to fabs into inline code.
   if (BB == &F->getEntryBlock() && 
-      &BB->front() == CI && &*++BB->begin() == Ret &&
+      FirstNonDbg(BB->front()) == CI &&
+      FirstNonDbg(llvm::next(BB->begin())) == TI &&
       callIsSmall(F)) {
     // A single-block function with just a call and a return. Check that
     // the arguments match.
@@ -370,9 +398,17 @@ bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
     for (; I != E && FI != FE; ++I, ++FI)
       if (*I != &*FI) break;
     if (I == E && FI == FE)
-      return false;
+      return 0;
   }
 
+  return CI;
+}
+
+bool TailCallElim::EliminateRecursiveTailCall(CallInst *CI, ReturnInst *Ret,
+                                       BasicBlock *&OldEntry,
+                                       bool &TailCallsAreMarkedTail,
+                                       SmallVector<PHINode*, 8> &ArgumentPHIs,
+                                       bool CannotTailCallElimCallsMarkedTail) {
   // If we are introducing accumulator recursion to eliminate operations after
   // the call instruction that are both associative and commutative, the initial
   // value for the accumulator is placed in this variable.  If this value is set
@@ -390,7 +426,8 @@ bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
   // tail call if all of the instructions between the call and the return are
   // movable to above the call itself, leaving the call next to the return.
   // Check that this is the case now.
-  for (BBI = CI, ++BBI; &*BBI != Ret; ++BBI) {
+  BasicBlock::iterator BBI = CI;
+  for (++BBI; &*BBI != Ret; ++BBI) {
     if (CanMoveAboveCall(BBI, CI)) continue;
     
     // If we can't move the instruction above the call, it might be because it
@@ -427,6 +464,9 @@ bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
       return false;
   }
 
+  BasicBlock *BB = Ret->getParent();
+  Function *F = BB->getParent();
+
   // OK! We can transform this tail call.  If this is the first one found,
   // create the new entry block, allowing us to branch back to the old entry.
   if (OldEntry == 0) {
@@ -536,3 +576,52 @@ bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
   ++NumEliminated;
   return true;
 }
+
+bool TailCallElim::FoldReturnAndProcessPred(BasicBlock *BB,
+                                       ReturnInst *Ret, BasicBlock *&OldEntry,
+                                       bool &TailCallsAreMarkedTail,
+                                       SmallVector<PHINode*, 8> &ArgumentPHIs,
+                                       bool CannotTailCallElimCallsMarkedTail) {
+  bool Change = false;
+
+  // If the return block contains nothing but the return and PHI's,
+  // there might be an opportunity to duplicate the return in its
+  // predecessors and perform TRC there. Look for predecessors that end
+  // in unconditional branch and recursive call(s).
+  SmallVector<BranchInst*, 8> UncondBranchPreds;
+  for (pred_iterator PI = pred_begin(BB), E = pred_end(BB); PI != E; ++PI) {
+    BasicBlock *Pred = *PI;
+    TerminatorInst *PTI = Pred->getTerminator();
+    if (BranchInst *BI = dyn_cast<BranchInst>(PTI))
+      if (BI->isUnconditional())
+        UncondBranchPreds.push_back(BI);
+  }
+
+  while (!UncondBranchPreds.empty()) {
+    BranchInst *BI = UncondBranchPreds.pop_back_val();
+    BasicBlock *Pred = BI->getParent();
+    if (CallInst *CI = FindTRECandidate(BI, CannotTailCallElimCallsMarkedTail)){
+      DEBUG(dbgs() << "FOLDING: " << *BB
+            << "INTO UNCOND BRANCH PRED: " << *Pred);
+      EliminateRecursiveTailCall(CI, FoldReturnIntoUncondBranch(Ret, BB, Pred),
+                                 OldEntry, TailCallsAreMarkedTail, ArgumentPHIs,
+                                 CannotTailCallElimCallsMarkedTail);
+      Change = true;
+    }
+  }
+
+  return Change;
+}
+
+bool TailCallElim::ProcessReturningBlock(ReturnInst *Ret, BasicBlock *&OldEntry,
+                                         bool &TailCallsAreMarkedTail,
+                                         SmallVector<PHINode*, 8> &ArgumentPHIs,
+                                       bool CannotTailCallElimCallsMarkedTail) {
+  CallInst *CI = FindTRECandidate(Ret, CannotTailCallElimCallsMarkedTail);
+  if (!CI)
+    return false;
+
+  return EliminateRecursiveTailCall(CI, Ret, OldEntry, TailCallsAreMarkedTail,
+                                    ArgumentPHIs,
+                                    CannotTailCallElimCallsMarkedTail);
+}
diff --git a/lib/Transforms/Utils/BasicBlockUtils.cpp b/lib/Transforms/Utils/BasicBlockUtils.cpp
index 5d3af3759a..acaea195e7 100644
--- a/lib/Transforms/Utils/BasicBlockUtils.cpp
+++ b/lib/Transforms/Utils/BasicBlockUtils.cpp
@@ -509,7 +509,32 @@ void llvm::FindFunctionBackedges(const Function &F,
       // Go up one level.
       InStack.erase(VisitStack.pop_back_val().first);
     }
-  } while (!VisitStack.empty());
-  
-  
+  } while (!VisitStack.empty()); 
+}
+
+/// FoldReturnIntoUncondBranch - This method duplicates the specified return
+/// instruction into a predecessor which ends in an unconditional branch. If
+/// the return instruction returns a value defined by a PHI, propagate the
+/// right value into the return. It returns the new return instruction in the
+/// predecessor.
+ReturnInst *llvm::FoldReturnIntoUncondBranch(ReturnInst *RI, BasicBlock *BB,
+                                             BasicBlock *Pred) {
+  Instruction *UncondBranch = Pred->getTerminator();
+  // Clone the return and add it to the end of the predecessor.
+  Instruction *NewRet = RI->clone();
+  Pred->getInstList().push_back(NewRet);
+      
+  // If the return instruction returns a value, and if the value was a
+  // PHI node in "BB", propagate the right value into the return.
+  for (User::op_iterator i = NewRet->op_begin(), e = NewRet->op_end();
+       i != e; ++i)
+    if (PHINode *PN = dyn_cast<PHINode>(*i))
+      if (PN->getParent() == BB)
+        *i = PN->getIncomingValueForBlock(Pred);
+      
+  // Update any PHI nodes in the returning block to realize that we no
+  // longer branch to them.
+  BB->removePredecessor(Pred);
+  UncondBranch->eraseFromParent();
+  return cast<ReturnInst>(NewRet);
 }
diff --git a/lib/Transforms/Utils/SimplifyCFG.cpp b/lib/Transforms/Utils/SimplifyCFG.cpp
index f6d7d76dbf..b9432c2e6e 100644
--- a/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -28,6 +28,7 @@
 #include "llvm/ADT/Statistic.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/CFG.h"
+#include "llvm/Support/CommandLine.h"
 #include "llvm/Support/ConstantRange.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/raw_ostream.h"
@@ -36,6 +37,10 @@
 #include <map>
 using namespace llvm;
 
+static cl::opt<bool>
+DupRet("simplifycfg-dup-ret", cl::Hidden, cl::init(false),
+       cl::desc("Duplicate return instructions into unconditional branches"));
+
 STATISTIC(NumSpeculations, "Number of speculative executed instructions");
 
 namespace {
@@ -2027,28 +2032,12 @@ bool SimplifyCFGOpt::SimplifyReturn(ReturnInst *RI) {
   }
   
   // If we found some, do the transformation!
-  if (!UncondBranchPreds.empty()) {
+  if (!UncondBranchPreds.empty() && DupRet) {
     while (!UncondBranchPreds.empty()) {
       BasicBlock *Pred = UncondBranchPreds.pop_back_val();
       DEBUG(dbgs() << "FOLDING: " << *BB
             << "INTO UNCOND BRANCH PRED: " << *Pred);
-      Instruction *UncondBranch = Pred->getTerminator();
-      // Clone the return and add it to the end of the predecessor.
-      Instruction *NewRet = RI->clone();
-      Pred->getInstList().push_back(NewRet);
-      
-      // If the return instruction returns a value, and if the value was a
-      // PHI node in "BB", propagate the right value into the return.
-      for (User::op_iterator i = NewRet->op_begin(), e = NewRet->op_end();
-           i != e; ++i)
-        if (PHINode *PN = dyn_cast<PHINode>(*i))
-          if (PN->getParent() == BB)
-            *i = PN->getIncomingValueForBlock(Pred);
-      
-      // Update any PHI nodes in the returning block to realize that we no
-      // longer branch to them.
-      BB->removePredecessor(Pred);
-      UncondBranch->eraseFromParent();
+      (void)FoldReturnIntoUncondBranch(RI, BB, Pred);
     }
     
     // If we eliminated all predecessors of the block, delete the block now.
diff --git a/test/CodeGen/X86/loop-blocks.ll b/test/CodeGen/X86/loop-blocks.ll
index 354d082069..faba630071 100644
--- a/test/CodeGen/X86/loop-blocks.ll
+++ b/test/CodeGen/X86/loop-blocks.ll
@@ -70,6 +70,7 @@ exit:
 
 ; Same as slightly_more_involved, but block_a is now a CFG diamond with
 ; fallthrough edges which should be preserved.
+; "callq block_a_merge_func" is tail duped.
 
 ; CHECK: yet_more_involved:
 ;      CHECK:   jmp .LBB2_1
@@ -78,12 +79,12 @@ exit:
 ; CHECK-NEXT:   callq bar99
 ; CHECK-NEXT:   callq get
 ; CHECK-NEXT:   cmpl $2999, %eax
-; CHECK-NEXT:   jg .LBB2_6
-; CHECK-NEXT:   callq block_a_true_func
-; CHECK-NEXT:   jmp .LBB2_7
-; CHECK-NEXT: .LBB2_6:
+; CHECK-NEXT:   jle .LBB2_5
 ; CHECK-NEXT:   callq block_a_false_func
-; CHECK-NEXT: .LBB2_7:
+; CHECK-NEXT:   callq block_a_merge_func
+; CHECK-NEXT:   jmp .LBB2_1
+; CHECK-NEXT: .LBB2_5:
+; CHECK-NEXT:   callq block_a_true_func
 ; CHECK-NEXT:   callq block_a_merge_func
 ; CHECK-NEXT: .LBB2_1:
 ; CHECK-NEXT:   callq body
diff --git a/test/Transforms/JumpThreading/and-and-cond.ll b/test/Transforms/JumpThreading/and-and-cond.ll
index e6db9ee5a3..765d940cc7 100644
--- a/test/Transforms/JumpThreading/and-and-cond.ll
+++ b/test/Transforms/JumpThreading/and-and-cond.ll
@@ -1,14 +1,14 @@
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | grep {ret i32 %v1}
-; There should be no uncond branches left.
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | not grep {br label}
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | FileCheck %s
 
 declare i32 @f1()
 declare i32 @f2()
 declare void @f3()
 
 define i32 @test(i1 %cond, i1 %cond2, i1 %cond3) {
+; CHECK: test
 	br i1 %cond, label %T1, label %F1
 
+; CHECK-NOT: T1:
 T1:
 	%v1 = call i32 @f1()
 	br label %Merge
@@ -18,6 +18,10 @@ F1:
 	br label %Merge
 
 Merge:
+; CHECK: Merge:
+; CHECK: %v1 = call i32 @f1()
+; CHECK-NEXT: %D = and i1 %cond2, %cond3
+; CHECK-NEXT: br i1 %D
 	%A = phi i1 [true, %T1], [false, %F1]
 	%B = phi i32 [%v1, %T1], [%v2, %F1]
 	%C = and i1 %A, %cond2
diff --git a/test/Transforms/JumpThreading/and-cond.ll b/test/Transforms/JumpThreading/and-cond.ll
index 58dbec72a7..0159bb3bb7 100644
--- a/test/Transforms/JumpThreading/and-cond.ll
+++ b/test/Transforms/JumpThreading/and-cond.ll
@@ -1,14 +1,14 @@
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | grep {ret i32 %v1}
-; There should be no uncond branches left.
-; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | not grep {br label}
+; RUN: opt < %s -jump-threading -mem2reg -instcombine -simplifycfg  -S | FileCheck %s
 
 declare i32 @f1()
 declare i32 @f2()
 declare void @f3()
 
 define i32 @test(i1 %cond, i1 %cond2) {
+; CHECK: test
 	br i1 %cond, label %T1, label %F1
 
+; CHECK-NOT: T1
 T1:
 	%v1 = call i32 @f1()
 	br label %Merge
@@ -18,6 +18,9 @@ F1:
 	br label %Merge
 
 Merge:
+; CHECK: Merge:
+; CHECK: %v1 = call i32 @f1()
+; CHECK-NEXT: br i1 %cond2
 	%A = phi i1 [true, %T1], [false, %F1]
 	%B = phi i32 [%v1, %T1], [%v2, %F1]
 	%C = and i1 %A, %cond2
diff --git a/test/Transforms/JumpThreading/thread-loads.ll b/test/Transforms/JumpThreading/thread-loads.ll
index 96ba701046..cce23ea319 100644
--- a/test/Transforms/JumpThreading/thread-loads.ll
+++ b/test/Transforms/JumpThreading/thread-loads.ll
@@ -1,4 +1,4 @@
-; RUN: opt < %s -jump-threading -simplifycfg -S | grep {ret i32 1}
+; RUN: opt < %s -jump-threading -S | FileCheck %s
 ; rdar://6402033
 
 ; Test that we can thread through the block with the partially redundant load (%2).
@@ -6,12 +6,16 @@ target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f3
 target triple = "i386-apple-darwin7"
 
 define i32 @foo(i32* %P) nounwind {
+; CHECK: foo
 entry:
 	%0 = tail call i32 (...)* @f1() nounwind		; <i32> [#uses=1]
 	%1 = icmp eq i32 %0, 0		; <i1> [#uses=1]
 	br i1 %1, label %bb1, label %bb
 
 bb:		; preds = %entry
+; CHECK: bb1.thread:
+; CHECK: store
+; CHECK: br label %bb3
 	store i32 42, i32* %P, align 4
 	br label %bb1
 
@@ -26,6 +30,9 @@ bb2:		; preds = %bb1
 	ret i32 %res.0
 
 bb3:		; preds = %bb1
+; CHECK: bb3:
+; CHECK: %res.01 = phi i32 [ 1, %bb1.thread ], [ 0, %bb1 ]
+; CHECK: ret i32 %res.01
 	ret i32 %res.0
 }
 
diff --git a/test/Transforms/SimplifyCFG/MagicPointer.ll b/test/Transforms/SimplifyCFG/MagicPointer.ll
index 54e5b14880..93b9a276ea 100644
--- a/test/Transforms/SimplifyCFG/MagicPointer.ll
+++ b/test/Transforms/SimplifyCFG/MagicPointer.ll
@@ -8,7 +8,6 @@
 ; CHECK: i64 2, label
 ; CHECK: i64 3, label
 ; CHECK: i64 4, label
-; CHECK-NOT: br
 ; CHECK: }
 
 target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64-s0:64:64-f80:128:128-n8:16:32:64"
diff --git a/test/Transforms/SimplifyCFG/basictest.ll b/test/Transforms/SimplifyCFG/basictest.ll
index b485f6ab05..052e10667d 100644
--- a/test/Transforms/SimplifyCFG/basictest.ll
+++ b/test/Transforms/SimplifyCFG/basictest.ll
@@ -25,16 +25,6 @@ define void @test3(i1 %T) {
 }
 
 
-define void @test4() {
-  br label %return
-return:
-  ret void
-; CHECK: @test4
-; CHECK-NEXT: ret void
-}
-@test4g = global i8* blockaddress(@test4, %return)
-
-
 ; PR5795
 define void @test5(i32 %A) {
   switch i32 %A, label %return [
diff --git a/test/Transforms/SimplifyCFG/switch_create.ll b/test/Transforms/SimplifyCFG/switch_create.ll
index da7f65a6ca..4e199bc859 100644
--- a/test/Transforms/SimplifyCFG/switch_create.ll
+++ b/test/Transforms/SimplifyCFG/switch_create.ll
@@ -147,7 +147,7 @@ UnifiedReturnBlock:             ; preds = %shortcirc_done.4, %shortcirc_next.4
 ; CHECK:       i32 16, label %UnifiedReturnBlock
 ; CHECK:       i32 17, label %UnifiedReturnBlock
 ; CHECK:       i32 18, label %UnifiedReturnBlock
-; CHECK:       i32 19, label %switch.edge
+; CHECK:       i32 19, label %UnifiedReturnBlock
 ; CHECK:     ]
 }
 
@@ -441,3 +441,29 @@ if.end:
 ; CHECK-NOT: switch
 ; CHECK: ret void
 }
+
+; PR8675
+; rdar://5134905
+define zeroext i1 @test16(i32 %x) nounwind {
+entry:
+; CHECK: @test16
+; CHECK: switch i32 %x, label %lor.rhs [
+; CHECK:   i32 1, label %lor.end
+; CHECK:   i32 2, label %lor.end
+; CHECK:   i32 3, label %lor.end
+; CHECK: ]
+  %cmp.i = icmp eq i32 %x, 1
+  br i1 %cmp.i, label %lor.end, label %lor.lhs.false
+
+lor.lhs.false:
+  %cmp.i2 = icmp eq i32 %x, 2
+  br i1 %cmp.i2, label %lor.end, label %lor.rhs
+
+lor.rhs:
+  %cmp.i1 = icmp eq i32 %x, 3
+  br label %lor.end
+
+lor.end:
+  %0 = phi i1 [ true, %lor.lhs.false ], [ true, %entry ], [ %cmp.i1, %lor.rhs ]
+  ret i1 %0
+}
diff --git a/test/Transforms/SimplifyCFG/switch_formation.dbg.ll b/test/Transforms/SimplifyCFG/switch_formation.dbg.ll
index f1c820ec43..357ffb60e1 100644
--- a/test/Transforms/SimplifyCFG/switch_formation.dbg.ll
+++ b/test/Transforms/SimplifyCFG/switch_formation.dbg.ll
@@ -1,5 +1,4 @@
-; RUN: opt < %s -simplifycfg -S | not grep br
-
+; RUN: opt < %s -simplifycfg -S | FileCheck %s
 
         %llvm.dbg.anchor.type = type { i32, i32 }
         %llvm.dbg.compile_unit.type = type { i32, { }*, i32, i8*, i8*, i8*, i1, i1, i8* }
@@ -13,7 +12,16 @@
 
 declare void @llvm.dbg.stoppoint(i32, i32, { }*) nounwind
 
-define i1 @_ZN4llvm11SetCondInst7classofEPKNS_11InstructionE({ i32, i32 }* %I) {
+define i1 @t({ i32, i32 }* %I) {
+; CHECK: t
+; CHECK:  switch i32 %tmp.2.i, label %shortcirc_next.4 [
+; CHECK:    i32 14, label %UnifiedReturnBlock
+; CHECK:    i32 15, label %UnifiedReturnBlock
+; CHECK:    i32 16, label %UnifiedReturnBlock
+; CHECK:    i32 17, label %UnifiedReturnBlock
+; CHECK:    i32 18, label %UnifiedReturnBlock
+; CHECK:    i32 19, label %UnifiedReturnBlock
+; CHECK:  ]
 entry:
         %tmp.1.i = getelementptr { i32, i32 }* %I, i64 0, i32 1         ; <i32*> [#uses=1]
         %tmp.2.i = load i32* %tmp.1.i           ; <i32> [#uses=6]