Now that instruction optzns can update the iterator as they go, we can

have objectsize folding recursively simplify away their result when it
folds.  It is important to catch this here, because otherwise we won't
eliminate the cross-block values at isel and other times.

llvm-svn: 123524

2 files changed, 40 insertions, 10 deletions

diff --git a/llvm/lib/Transforms/Scalar/CodeGenPrepare.cpp b/llvm/lib/Transforms/Scalar/CodeGenPrepare.cpp
index ae827871b54..921cec83512 100644
--- a/llvm/lib/Transforms/Scalar/CodeGenPrepare.cpp
+++ b/llvm/lib/Transforms/Scalar/CodeGenPrepare.cpp
@@ -42,6 +42,7 @@
 #include "llvm/Support/PatternMatch.h"
 #include "llvm/Support/raw_ostream.h"
 #include "llvm/Support/IRBuilder.h"
+#include "llvm/Support/ValueHandle.h"
 using namespace llvm;
 using namespace llvm::PatternMatch;
 
@@ -634,8 +635,18 @@ bool CodeGenPrepare::OptimizeCallInst(CallInst *CI) {
     bool Min = (cast<ConstantInt>(II->getArgOperand(1))->getZExtValue() == 1);
     const Type *ReturnTy = CI->getType();
     Constant *RetVal = ConstantInt::get(ReturnTy, Min ? 0 : -1ULL);    
-    CI->replaceAllUsesWith(RetVal);
-    CI->eraseFromParent();
+    
+    // Substituting this can cause recursive simplifications, which can
+    // invalidate our iterator.  Use a WeakVH to hold onto it in case this
+    // happens.
+    WeakVH IterHandle(CurInstIterator);
+    
+    ReplaceAndSimplifyAllUses(CI, RetVal, TLI ? TLI->getTargetData() : 0, DT);
+
+    // If the iterator instruction was recursively deleted, start over at the
+    // start of the block.
+    if (IterHandle != CurInstIterator)
+      CurInstIterator = BB->begin();
     return true;
   }
 
@@ -653,6 +664,7 @@ bool CodeGenPrepare::OptimizeCallInst(CallInst *CI) {
   CodeGenPrepareFortifiedLibCalls Simplifier;
   return Simplifier.fold(CI, TD);
 }
+
 //===----------------------------------------------------------------------===//
 // Memory Optimization
 //===----------------------------------------------------------------------===//
@@ -1076,14 +1088,8 @@ bool CodeGenPrepare::OptimizeBlock(BasicBlock &BB) {
   SunkAddrs.clear();
 
   CurInstIterator = BB.begin();
-  for (BasicBlock::iterator E = BB.end(); CurInstIterator != E; ) {
-    Instruction *I = CurInstIterator++;
-
-    if (CallInst *CI = dyn_cast<CallInst>(I))
-      MadeChange |= OptimizeCallInst(CI);
-    else
-      MadeChange |= OptimizeInst(I);
-  }
+  for (BasicBlock::iterator E = BB.end(); CurInstIterator != E; )
+    MadeChange |= OptimizeInst(CurInstIterator++);
 
   return MadeChange;
 }
diff --git a/llvm/test/Transforms/CodeGenPrepare/basic.ll b/llvm/test/Transforms/CodeGenPrepare/basic.ll
new file mode 100644
index 00000000000..c3e57b4d206
--- /dev/null
+++ b/llvm/test/Transforms/CodeGenPrepare/basic.ll
@@ -0,0 +1,24 @@
+; RUN: opt -codegenprepare %s -S -o - | FileCheck %s
+
+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"
+target triple = "x86_64-apple-darwin10.0.0"
+
+; CHECK: @test1
+define i32 @test1(i8* %ptr) nounwind ssp noredzone align 2 {
+entry:
+  %0 = tail call i64 @llvm.objectsize.i64(i8* %ptr, i1 false)
+  %1 = icmp ugt i64 %0, 3
+  br i1 %1, label %T, label %trap
+
+; CHECK: br i1 true, label
+trap:                                             ; preds = %0, %entry
+  tail call void @llvm.trap() noreturn nounwind
+  unreachable
+
+T:
+  ret i32 4
+}
+
+declare i64 @llvm.objectsize.i64(i8*, i1) nounwind readonly
+
+declare void @llvm.trap() nounwind