Fix a bug in the code that builds SDNodes from vector GEPs.

When the GEP index is a vector of pointers, the code that calculated the size
of the element started from the vector type, and not the contained pointer type.
As a result, instead of looking at the data element pointed by the vector, this
code used the size of the vector. This works for 32bit members (on 32bit
systems), but not for other types. Added code to peel the vector type and
added a test.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@151626 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
index ca64c903f39a68379173955a57f9f30f2741777f..5eba2c3ea7c13a8972bbe07b1f953727e77061ad 100644 (file)
--- a/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
+++ b/lib/CodeGen/SelectionDAG/SelectionDAGBuilder.cpp
@@ -3078,6 +3078,10 @@ void SelectionDAGBuilder::visitExtractValue(const ExtractValueInst &I) {
 void SelectionDAGBuilder::visitGetElementPtr(const User &I) {
   SDValue N = getValue(I.getOperand(0));
   Type *Ty = I.getOperand(0)->getType();
+  // If this is a vector of pointers, use the size of the
+  // vector element and not the size of the pointer.
+  if (VectorType *VTy = dyn_cast<VectorType>(Ty))
+    Ty = VTy->getElementType();
 
   for (GetElementPtrInst::const_op_iterator OI = I.op_begin()+1, E = I.op_end();
        OI != E; ++OI) {

diff --git a/test/CodeGen/X86/vector-gep.ll b/test/CodeGen/X86/vector-gep.ll
index d032eda88bec882390917f1c138e0a99ef7e61e5..3476e36c646fd96ea837a2aca5f4718e7b22de49 100644 (file)
--- a/test/CodeGen/X86/vector-gep.ll
+++ b/test/CodeGen/X86/vector-gep.ll
@@ -1,27 +1,27 @@
 ; RUN: llc < %s -march=x86 -mcpu=corei7-avx | FileCheck %s
 ; RUN: opt -instsimplify %s -disable-output
 
-;CHECK: AGEP0
+;CHECK: AGEP0:
 define <4 x i32*> @AGEP0(i32* %ptr) nounwind {
 entry:
   %vecinit.i = insertelement <4 x i32*> undef, i32* %ptr, i32 0
   %vecinit2.i = insertelement <4 x i32*> %vecinit.i, i32* %ptr, i32 1
   %vecinit4.i = insertelement <4 x i32*> %vecinit2.i, i32* %ptr, i32 2
   %vecinit6.i = insertelement <4 x i32*> %vecinit4.i, i32* %ptr, i32 3
-;CHECK: pslld
+;CHECK: pslld $2
 ;CHECK: padd
   %A2 = getelementptr <4 x i32*> %vecinit6.i, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
-;CHECK: pslld
+;CHECK: pslld $2
 ;CHECK: padd
   %A3 = getelementptr <4 x i32*> %A2, <4 x i32> <i32 10, i32 14, i32 19, i32 233>
   ret <4 x i32*> %A3
 ;CHECK: ret
 }
 
-;CHECK: AGEP1
+;CHECK: AGEP1:
 define i32 @AGEP1(<4 x i32*> %param) nounwind {
 entry:
-;CHECK: pslld
+;CHECK: pslld $2
 ;CHECK: padd
   %A2 = getelementptr <4 x i32*> %param, <4 x i32> <i32 1, i32 2, i32 3, i32 4>
   %k = extractelement <4 x i32*> %A2, i32 3
@@ -30,10 +30,10 @@ entry:
 ;CHECK: ret
 }
 
-;CHECK: AGEP2
+;CHECK: AGEP2:
 define i32 @AGEP2(<4 x i32*> %param, <4 x i32> %off) nounwind {
 entry:
-;CHECK: pslld
+;CHECK: pslld $2
 ;CHECK: padd
   %A2 = getelementptr <4 x i32*> %param, <4 x i32> %off
   %k = extractelement <4 x i32*> %A2, i32 3
@@ -42,10 +42,10 @@ entry:
 ;CHECK: ret
 }
 
-;CHECK: AGEP3
+;CHECK: AGEP3:
 define <4 x i32*> @AGEP3(<4 x i32*> %param, <4 x i32> %off) nounwind {
 entry:
-;CHECK: pslld
+;CHECK: pslld $2
 ;CHECK: padd
   %A2 = getelementptr <4 x i32*> %param, <4 x i32> %off
   %v = alloca i32
@@ -54,24 +54,35 @@ entry:
 ;CHECK: ret
 }
 
-;CHECK: AGEP4
-define <4 x i8*> @AGEP4(<4 x i8*> %param, <4 x i32> %off) nounwind {
+;CHECK: AGEP4:
+define <4 x i16*> @AGEP4(<4 x i16*> %param, <4 x i32> %off) nounwind {
 entry:
-;CHECK: pslld
+; Multiply offset by two (add it to itself).
 ;CHECK: padd
-  %A = getelementptr <4 x i8*> %param, <4 x i32> %off
-  ret <4 x i8*> %A
+; add the base to the offset
+;CHECK: padd
+  %A = getelementptr <4 x i16*> %param, <4 x i32> %off
+  ret <4 x i16*> %A
 ;CHECK: ret
 }
 
-;CHECK: AGEP5
+;CHECK: AGEP5:
 define <4 x i8*> @AGEP5(<4 x i8*> %param, <4 x i8> %off) nounwind {
 entry:
-;CHECK: pslld
-;CHECK: padd
+;CHECK: paddd
   %A = getelementptr <4 x i8*> %param, <4 x i8> %off
   ret <4 x i8*> %A
 ;CHECK: ret
 }
 
 
+; The size of each element is 1 byte. No need to multiply by element size.
+;CHECK: AGEP6:
+define <4 x i8*> @AGEP6(<4 x i8*> %param, <4 x i32> %off) nounwind {
+entry:
+;CHECK-NOT: pslld
+  %A = getelementptr <4 x i8*> %param, <4 x i32> %off
+  ret <4 x i8*> %A
+;CHECK: ret
+}
+