[llvm-branch-commits] [llvm-branch] r185869 - Fix for a regression caused by the LoopVectorizer when

[Tom Stellard]

Author: tstellar
Date: Mon Jul  8 16:20:34 2013
New Revision: 185869

URL: http://llvm.org/viewvc/llvm-project?rev=185869&view=rev
Log:
Fix for a regression caused by the LoopVectorizer when
vectorizing loops with memory accesses to non-zero address spaces. It
simply dropped the AS info. Fixes PR16306.


Merged from r184103
Author: Pekka Jaaskelainen <pekka.jaaskelainen at tut.fi>
Date:   Mon Jun 17 18:49:06 2013 +0000

Added:
    llvm/branches/release_33/test/Transforms/LoopVectorize/multiple-address-spaces.ll
Modified:
    llvm/branches/release_33/lib/Transforms/Vectorize/LoopVectorize.cpp

Modified: llvm/branches/release_33/lib/Transforms/Vectorize/LoopVectorize.cpp
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/release_33/lib/Transforms/Vectorize/LoopVectorize.cpp?rev=185869&r1=185868&r2=185869&view=diff
==============================================================================
--- llvm/branches/release_33/lib/Transforms/Vectorize/LoopVectorize.cpp (original)
+++ llvm/branches/release_33/lib/Transforms/Vectorize/LoopVectorize.cpp Mon Jul  8 16:20:34 2013
@@ -964,7 +964,7 @@ void InnerLoopVectorizer::vectorizeMemor
   Type *DataTy = VectorType::get(ScalarDataTy, VF);
   Value *Ptr = LI ? LI->getPointerOperand() : SI->getPointerOperand();
   unsigned Alignment = LI ? LI->getAlignment() : SI->getAlignment();
-
+  unsigned AddressSpace = Ptr->getType()->getPointerAddressSpace();
   unsigned ScalarAllocatedSize = DL->getTypeAllocSize(ScalarDataTy);
   unsigned VectorElementSize = DL->getTypeStoreSize(DataTy)/VF;
 
@@ -1039,7 +1039,7 @@ void InnerLoopVectorizer::vectorizeMemor
         PartPtr = Builder.CreateGEP(PartPtr, Builder.getInt32(1 - VF));
       }
 
-      Value *VecPtr = Builder.CreateBitCast(PartPtr, DataTy->getPointerTo());
+      Value *VecPtr = Builder.CreateBitCast(PartPtr, DataTy->getPointerTo(AddressSpace));
       Builder.CreateStore(StoredVal[Part], VecPtr)->setAlignment(Alignment);
     }
   }
@@ -1055,7 +1055,7 @@ void InnerLoopVectorizer::vectorizeMemor
       PartPtr = Builder.CreateGEP(PartPtr, Builder.getInt32(1 - VF));
     }
 
-    Value *VecPtr = Builder.CreateBitCast(PartPtr, DataTy->getPointerTo());
+    Value *VecPtr = Builder.CreateBitCast(PartPtr, DataTy->getPointerTo(AddressSpace));
     Value *LI = Builder.CreateLoad(VecPtr, "wide.load");
     cast<LoadInst>(LI)->setAlignment(Alignment);
     Entry[Part] = Reverse ? reverseVector(LI) :  LI;

Added: llvm/branches/release_33/test/Transforms/LoopVectorize/multiple-address-spaces.ll
URL: http://llvm.org/viewvc/llvm-project/llvm/branches/release_33/test/Transforms/LoopVectorize/multiple-address-spaces.ll?rev=185869&view=auto
==============================================================================
--- llvm/branches/release_33/test/Transforms/LoopVectorize/multiple-address-spaces.ll (added)
+++ llvm/branches/release_33/test/Transforms/LoopVectorize/multiple-address-spaces.ll Mon Jul  8 16:20:34 2013
@@ -0,0 +1,47 @@
+; RUN: opt < %s  -loop-vectorize -force-vector-unroll=1 -force-vector-width=4 -dce -instcombine -S | FileCheck %s
+
+; From a simple program with two address spaces:
+; char Y[4*10000] __attribute__((address_space(1)));
+; char X[4*10000];
+; int main() {
+;    for (int i = 0; i < 4*10000; ++i)
+;        X[i] = Y[i] + 1;
+;    return 0;
+;}
+
+
+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-unknown-linux-gnu"
+
+ at Y = common addrspace(1) global [40000 x i8] zeroinitializer, align 16
+ at X = common global [40000 x i8] zeroinitializer, align 16
+
+;CHECK: @main
+;CHECK: bitcast i8 addrspace(1)* %{{.*}} to <4 x i8> addrspace(1)*
+;CHECK: bitcast i8* %{{.*}} to <4 x i8>*
+
+; Function Attrs: nounwind uwtable
+define i32 @main() #0 {
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %for.body, %entry
+  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds [40000 x i8] addrspace(1)* @Y, i64 0, i64 %indvars.iv
+  %0 = load i8 addrspace(1)* %arrayidx, align 1, !tbaa !0
+  %add = add i8 %0, 1
+  %arrayidx3 = getelementptr inbounds [40000 x i8]* @X, i64 0, i64 %indvars.iv
+  store i8 %add, i8* %arrayidx3, align 1, !tbaa !0
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, 40000
+  br i1 %exitcond, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body
+  ret i32 0
+}
+
+attributes #0 = { nounwind uwtable "less-precise-fpmad"="false" "no-frame-pointer-elim"="false" "no-frame-pointer-elim-non-leaf"="false" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "unsafe-fp-math"="false" "use-soft-float"="false" }
+
+!0 = metadata !{metadata !"omnipotent char", metadata !1}
+!1 = metadata !{metadata !"Simple C/C++ TBAA"}