Make the variant of __builtin_shufflevector that takes the shuffle indexes as a vector actually usable. Patch by David Neto. PR12465.

llvm-svn: 154128

2 files changed, 38 insertions, 3 deletions

diff --git a/clang/lib/CodeGen/CGExprScalar.cpp b/clang/lib/CodeGen/CGExprScalar.cpp
index e9de43dd66b..5723541c62d 100644
--- a/clang/lib/CodeGen/CGExprScalar.cpp
+++ b/clang/lib/CodeGen/CGExprScalar.cpp
@@ -754,8 +754,8 @@ Value *ScalarExprEmitter::VisitShuffleVectorExpr(ShuffleVectorExpr *E) {
                                                         MTy->getNumElements());
     Value* NewV = llvm::UndefValue::get(RTy);
     for (unsigned i = 0, e = MTy->getNumElements(); i != e; ++i) {
-      Value *Indx = Builder.getInt32(i);
-      Indx = Builder.CreateExtractElement(Mask, Indx, "shuf_idx");
+      Value *IIndx = Builder.getInt32(i);
+      Value *Indx = Builder.CreateExtractElement(Mask, IIndx, "shuf_idx");
       Indx = Builder.CreateZExt(Indx, CGF.Int32Ty, "idx_zext");
       
       // Handle vec3 special since the index will be off by one for the RHS.
@@ -767,7 +767,7 @@ Value *ScalarExprEmitter::VisitShuffleVectorExpr(ShuffleVectorExpr *E) {
         Indx = Builder.CreateSelect(cmpIndx, newIndx, Indx, "sel_shuf_idx");
       }
       Value *VExt = Builder.CreateExtractElement(LHS, Indx, "shuf_elt");
-      NewV = Builder.CreateInsertElement(NewV, VExt, Indx, "shuf_ins");
+      NewV = Builder.CreateInsertElement(NewV, VExt, IIndx, "shuf_ins");
     }
     return NewV;
   }
diff --git a/clang/test/CodeGen/builtinshufflevector2.c b/clang/test/CodeGen/builtinshufflevector2.c
new file mode 100644
index 00000000000..faf7a3ec1a1
--- /dev/null
+++ b/clang/test/CodeGen/builtinshufflevector2.c
@@ -0,0 +1,35 @@
+// RUN: %clang -emit-llvm -S -o - %s | FileCheck %s
+
+typedef float float4 __attribute__((ext_vector_type(4)));
+typedef unsigned int uint4 __attribute__((ext_vector_type(4)));
+
+// CHECK: define void @clang_shufflevector_v_v(
+void clang_shufflevector_v_v( float4* A, float4 x, uint4 mask ) {
+// CHECK: [[MASK:%.*]] = and <4 x i32> {{%.*}}, <i32 3, i32 3, i32 3, i32 3>
+// CHECK: [[I:%.*]] = extractelement <4 x i32> [[MASK]], i32 0
+// CHECK: [[E:%.*]] = extractelement <4 x float> [[X:%.*]], i32 [[I]]
+//
+// Here is where ToT Clang code generation makes a mistake.  
+// It uses [[I]] as the insertion index instead of 0.
+// Similarly on the remaining insertelement.
+// CHECK: [[V:%[a-zA-Z0-9._]+]] = insertelement <4 x float> undef, float [[E]], i32 0
+
+// CHECK: [[I:%.*]] = extractelement <4 x i32> [[MASK]], i32 1
+// CHECK: [[E:%.*]] = extractelement <4 x float> [[X]], i32 [[I]]
+// CHECK: [[V:%.*]] = insertelement <4 x float> [[V]], float [[E]], i32 1
+// CHECK: [[I:%.*]] = extractelement <4 x i32> [[MASK]], i32 2
+// CHECK: [[E:%.*]] = extractelement <4 x float> [[X]], i32 [[I]]
+// CHECK: [[V:%.*]] = insertelement <4 x float> [[V]], float [[E]], i32 2
+// CHECK: [[I:%.*]] = extractelement <4 x i32> [[MASK]], i32 3
+// CHECK: [[E:%.*]] = extractelement <4 x float> [[X]], i32 [[I]]
+// CHECK: [[V:%.*]] = insertelement <4 x float> [[V]], float [[E]], i32 3
+// CHECK: store <4 x float> [[V]], <4 x float>* {{%.*}},
+  *A = __builtin_shufflevector( x, mask );
+}
+
+// CHECK: define void @clang_shufflevector_v_v_c(
+void clang_shufflevector_v_v_c( float4* A, float4 x, float4 y, uint4 mask ) {
+// CHECK: [[V:%.*]] = shufflevector <4 x float> {{%.*}}, <4 x float> {{%.*}}, <4 x i32> <i32 0, i32 4, i32 1, i32 5>
+// CHECK: store <4 x float> [[V]], <4 x float>* {{%.*}}
+  *A = __builtin_shufflevector( x, y, 0, 4, 1, 5 );
+}