Use the AliasAnalysis isIdentifiedObj because it also understands mallocs and c++ news.

PR14158.

llvm-svn: 166491

2 files changed, 48 insertions, 19 deletions

diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index f944d9b4fcd..bead39225b8 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -259,10 +259,6 @@ private:
   /// Returns true if BB is vectorizable
   bool canVectorizeMemory(BasicBlock &BB);
 
-  // Check if a pointer value is known to be disjoint.
-  // Example: Alloca, Global, NoAlias.
-  bool isIdentifiedSafeObject(Value* Val);
-
   /// Returns True, if 'Phi' is the kind of reduction variable for type
   /// 'Kind'. If this is a reduction variable, it adds it to ReductionList.
   bool AddReductionVar(PHINode *Phi, ReductionKind Kind);
@@ -1178,7 +1174,7 @@ bool LoopVectorizationLegality::canVectorizeMemory(BasicBlock &BB) {
     GetUnderlyingObjects(*I, TempObjects, DL);
     for (ValueVector::iterator it=TempObjects.begin(), e=TempObjects.end();
          it != e; ++it) {
-      if (!isIdentifiedSafeObject(*it)) {
+      if (!isIdentifiedObject(*it)) {
         DEBUG(dbgs() << "LV: Found an unidentified write ptr:"<< **it <<"\n");
         return false;
       }
@@ -1196,7 +1192,7 @@ bool LoopVectorizationLegality::canVectorizeMemory(BasicBlock &BB) {
     GetUnderlyingObjects(*I, TempObjects, DL);
     for (ValueVector::iterator it=TempObjects.begin(), e=TempObjects.end();
          it != e; ++it) {
-      if (!isIdentifiedSafeObject(*it)) {
+      if (!isIdentifiedObject(*it)) {
         DEBUG(dbgs() << "LV: Found an unidentified read ptr:"<< **it <<"\n");
         return false;
       }
@@ -1213,19 +1209,6 @@ bool LoopVectorizationLegality::canVectorizeMemory(BasicBlock &BB) {
   return true;
 }
 
-/// Checks if the value is a Global variable or if it is an Arguments
-/// marked with the NoAlias attribute.
-bool LoopVectorizationLegality::isIdentifiedSafeObject(Value* Val) {
-  assert(Val && "Invalid value");
-  if (isa<GlobalValue>(Val))
-    return true;
-  if (isa<AllocaInst>(Val))
-    return true;
-  if (Argument *A = dyn_cast<Argument>(Val))
-    return A->hasNoAliasAttr();
-  return false;
-}
-
 bool LoopVectorizationLegality::AddReductionVar(PHINode *Phi,
                                                 ReductionKind Kind) {
   if (Phi->getNumIncomingValues() != 2)
diff --git a/llvm/test/Transforms/LoopVectorize/cpp-new-array.ll b/llvm/test/Transforms/LoopVectorize/cpp-new-array.ll
new file mode 100644
index 00000000000..12a91b07914
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/cpp-new-array.ll
@@ -0,0 +1,46 @@
+; RUN: opt < %s  -loop-vectorize -dce -instcombine -licm -S | 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-S128"
+target triple = "x86_64-apple-macosx10.8.0"
+
+;CHECK: @cpp_new_arrays
+;CHECK: insertelement <4 x i32>
+;CHECK: load <4 x float>
+;CHECK: fadd <4 x float>
+;CHECK: ret i32
+define i32 @cpp_new_arrays() uwtable ssp {
+entry:
+  %call = call noalias i8* @_Znwm(i64 4)
+  %0 = bitcast i8* %call to float*
+  store float 1.000000e+03, float* %0, align 4
+  %call1 = call noalias i8* @_Znwm(i64 4)
+  %1 = bitcast i8* %call1 to float*
+  store float 1.000000e+03, float* %1, align 4
+  %call3 = call noalias i8* @_Znwm(i64 4)
+  %2 = bitcast i8* %call3 to float*
+  store float 1.000000e+03, float* %2, align 4
+  br label %for.body
+
+for.body:                                         ; preds = %entry, %for.body
+  %i.01 = phi i32 [ 0, %entry ], [ %inc, %for.body ]
+  %idxprom = sext i32 %i.01 to i64
+  %arrayidx = getelementptr inbounds float* %0, i64 %idxprom
+  %3 = load float* %arrayidx, align 4
+  %idxprom5 = sext i32 %i.01 to i64
+  %arrayidx6 = getelementptr inbounds float* %1, i64 %idxprom5
+  %4 = load float* %arrayidx6, align 4
+  %add = fadd float %3, %4
+  %idxprom7 = sext i32 %i.01 to i64
+  %arrayidx8 = getelementptr inbounds float* %2, i64 %idxprom7
+  store float %add, float* %arrayidx8, align 4
+  %inc = add nsw i32 %i.01, 1
+  %cmp = icmp slt i32 %inc, 1000
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:                                          ; preds = %for.body
+  %5 = load float* %2, align 4
+  %conv10 = fptosi float %5 to i32
+  ret i32 %conv10
+}
+
+declare noalias i8* @_Znwm(i64)