From e96fec2e436003714a6bf3739d594d77701da1e5 Mon Sep 17 00:00:00 2001
From: Arnold Schwaighofer <aschwaighofer@apple.com>
Date: Wed, 15 Jan 2014 03:35:46 +0000
Subject: [PATCH] LoopVectorize: Only strip casts from integer types when
 replacing symbolic strides

Fixes PR18480.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@199291 91177308-0d34-0410-b5e6-96231b3b80d8
---
 lib/Transforms/Vectorize/LoopVectorize.cpp    |  9 +++--
 .../LoopVectorize/version-mem-access.ll       | 37 +++++++++++++++++++
 2 files changed, 42 insertions(+), 4 deletions(-)

diff --git a/lib/Transforms/Vectorize/LoopVectorize.cpp b/lib/Transforms/Vectorize/LoopVectorize.cpp
index bdfa0f06476..c05288bd070 100644
--- a/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -1072,9 +1072,10 @@ struct LoopVectorize : public LoopPass {
 // LoopVectorizationCostModel.
 //===----------------------------------------------------------------------===//
 
-static Value *stripCast(Value *V) {
+static Value *stripIntegerCast(Value *V) {
   if (CastInst *CI = dyn_cast<CastInst>(V))
-    return CI->getOperand(0);
+    if (CI->getOperand(0)->getType()->isIntegerTy())
+      return CI->getOperand(0);
   return V;
 }
 
@@ -1095,7 +1096,7 @@ static const SCEV *replaceSymbolicStrideSCEV(ScalarEvolution *SE,
     Value *StrideVal = SI->second;
 
     // Strip casts.
-    StrideVal = stripCast(StrideVal);
+    StrideVal = stripIntegerCast(StrideVal);
 
     // Replace symbolic stride by one.
     Value *One = ConstantInt::get(StrideVal->getType(), 1);
@@ -1551,7 +1552,7 @@ InnerLoopVectorizer::addStrideCheck(Instruction *Loc) {
   for (SmallPtrSet<Value *, 8>::iterator SI = Legal->strides_begin(),
                                          SE = Legal->strides_end();
        SI != SE; ++SI) {
-    Value *Ptr = stripCast(*SI);
+    Value *Ptr = stripIntegerCast(*SI);
     Value *C = ChkBuilder.CreateICmpNE(Ptr, ConstantInt::get(Ptr->getType(), 1),
                                        "stride.chk");
     // Store the first instruction we create.
diff --git a/test/Transforms/LoopVectorize/version-mem-access.ll b/test/Transforms/LoopVectorize/version-mem-access.ll
index e7127281111..51d20e227dd 100644
--- a/test/Transforms/LoopVectorize/version-mem-access.ll
+++ b/test/Transforms/LoopVectorize/version-mem-access.ll
@@ -48,3 +48,40 @@ for.end.loopexit:
 for.end:
   ret void
 }
+
+; We used to crash on this function because we removed the fptosi cast when
+; replacing the symbolic stride '%conv'.
+; PR18480
+
+; CHECK-LABEL: fn1
+; CHECK: load <2 x double>
+
+define void @fn1(double* noalias %x, double* noalias %c, double %a) {
+entry:
+  %conv = fptosi double %a to i32
+  %cmp8 = icmp sgt i32 %conv, 0
+  br i1 %cmp8, label %for.body.preheader, label %for.end
+
+for.body.preheader:
+  br label %for.body
+
+for.body:
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
+  %0 = trunc i64 %indvars.iv to i32
+  %mul = mul nsw i32 %0, %conv
+  %idxprom = sext i32 %mul to i64
+  %arrayidx = getelementptr inbounds double* %x, i64 %idxprom
+  %1 = load double* %arrayidx, align 8
+  %arrayidx3 = getelementptr inbounds double* %c, i64 %indvars.iv
+  store double %1, double* %arrayidx3, align 8
+  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, %conv
+  br i1 %exitcond, label %for.end.loopexit, label %for.body
+
+for.end.loopexit:
+  br label %for.end
+
+for.end:
+  ret void
+}
-- 
2.34.1