vendor/clang/clang-trunk-r308421

17 files changed, 777 insertions, 98 deletions

diff --git a/test/OpenMP/distribute_parallel_for_if_codegen.cpp b/test/OpenMP/distribute_parallel_for_if_codegen.cpp
index e6cd7215f585..9ee909e0139c 100644
--- a/test/OpenMP/distribute_parallel_for_if_codegen.cpp
+++ b/test/OpenMP/distribute_parallel_for_if_codegen.cpp
@@ -1,6 +1,6 @@
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s
 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
-// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix=CHECK %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -fopenmp-targets=powerpc64le-ibm-linux-gnu -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
diff --git a/test/OpenMP/for_reduction_codegen.cpp b/test/OpenMP/for_reduction_codegen.cpp
index 274546b06547..d5afae990bf4 100644
--- a/test/OpenMP/for_reduction_codegen.cpp
+++ b/test/OpenMP/for_reduction_codegen.cpp
@@ -497,15 +497,12 @@ int main() {
 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
 
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
+// CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}**
+// CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]],
+// CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]],
 
-// CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
-// CHECK: [[LB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
-// CHECK: [[LB1_0:%.+]] = getelementptr inbounds i32, i32* [[LB1]], i64 0
-// CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
-// CHECK: [[UB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
-// CHECK: [[UB1_UP:%.+]] = getelementptr inbounds i32, i32* [[UB1]], i64 %
-// CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP]] to i64
-// CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0]] to i64
+// CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64
+// CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64
 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]]
 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1
@@ -704,7 +701,7 @@ int main() {
 
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
 
-// CHECK: [[ARR_SIZE:%.+]] = mul nuw i64 %{{.+}}, 4
+// CHECK: [[ARR_SIZE:%.+]] = udiv exact i64
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
 
@@ -718,7 +715,6 @@ int main() {
 // CHECK: br i1 [[DONE]],
 
 // Check initialization of private copy.
-// CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0
 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]]
@@ -727,6 +723,7 @@ int main() {
 // CHECK: call void @_ZN1SIfEC1Ev([[S_FLOAT_TY]]* %
 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
 // CHECK: br i1 [[DONE]],
+// CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]*
 
 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
@@ -895,23 +892,15 @@ int main() {
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]],
 
-// CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 0
-// CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
-// CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 1
-// CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 4
-// CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 6
-// CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
-// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
+// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
+// CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -932,8 +921,6 @@ int main() {
 // CHECK: [[VVAR2_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]],
 
 // CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0
-// CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 4
-// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
@@ -941,6 +928,7 @@ int main() {
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -960,19 +948,15 @@ int main() {
 
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 2
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
-// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
+// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
+// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -997,11 +981,11 @@ int main() {
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0
 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 2
 
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** %
+// CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 
 // CHECK: ret void
 
diff --git a/test/OpenMP/for_reduction_codegen_UDR.cpp b/test/OpenMP/for_reduction_codegen_UDR.cpp
index 699e06dda85a..95b04f44d468 100644
--- a/test/OpenMP/for_reduction_codegen_UDR.cpp
+++ b/test/OpenMP/for_reduction_codegen_UDR.cpp
@@ -306,15 +306,12 @@ int main() {
 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],
 
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
+// CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}**
+// CHECK: store i{{[0-9]+}}* %{{.+}}, i{{[0-9]+}}** [[ARR_ADDR:%.+]],
+// CHECK: [[ARR:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[ARR_ADDR]],
 
-// CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
-// CHECK: [[LB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
-// CHECK: [[LB1_0:%.+]] = getelementptr inbounds i32, i32* [[LB1]], i64 0
-// CHECK: [[IDX1:%.+]] = mul nsw i64 1, %{{.+}}
-// CHECK: [[UB1:%.+]] = getelementptr inbounds i32, i32* %{{.+}}, i64 [[IDX1]]
-// CHECK: [[UB1_UP:%.+]] = getelementptr inbounds i32, i32* [[UB1]], i64 %
-// CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP]] to i64
-// CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0]] to i64
+// CHECK: [[UB_CAST:%.+]] = ptrtoint i32* [[UB1_UP:%.+]] to i64
+// CHECK: [[LB_CAST:%.+]] = ptrtoint i32* [[LB1_0:%.+]] to i64
 // CHECK: [[DIFF:%.+]] = sub i64 [[UB_CAST]], [[LB_CAST]]
 // CHECK: [[SIZE_1:%.+]] = sdiv exact i64 [[DIFF]], ptrtoint (i32* getelementptr (i32, i32* null, i32 1) to i64)
 // CHECK: [[ARR_SIZE:%.+]] = add nuw i64 [[SIZE_1]], 1
@@ -508,7 +505,7 @@ int main() {
 
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
 
-// CHECK: [[ARR_SIZE:%.+]] = mul nuw i64 %{{.+}}, 4
+// CHECK: [[ARR_SIZE:%.+]] = udiv exact i64
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[ARR_PRIV:%.+]] = alloca i32, i64 [[ARR_SIZE]],
 
@@ -522,8 +519,8 @@ int main() {
 // CHECK: br i1 [[DONE]],
 
 // Check initialization of private copy.
-// CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
 // CHECK: [[BEGIN:%.+]] = getelementptr inbounds [10 x [4 x [[S_FLOAT_TY]]]], [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]], i32 0, i32 0, i32 0
+// CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
 // CHECK: [[END:%.+]] = getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* [[BEGIN]], i64 40
 // CHECK: [[ISEMPTY:%.+]] = icmp eq [[S_FLOAT_TY]]* [[BEGIN]], [[END]]
 // CHECK: br i1 [[ISEMPTY]],
@@ -531,6 +528,7 @@ int main() {
 // CHECK: call void @_Z4initR6BaseS1RKS_(%
 // CHECK: [[DONE:%.+]] = icmp eq [[S_FLOAT_TY]]* %{{.+}}, [[END]]
 // CHECK: br i1 [[DONE]],
+// CHECK: [[LHS_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* %{{.+}} to [[S_FLOAT_TY]]*
 // CHECK: [[ARRS_PRIV_BEGIN:%.+]] = bitcast [10 x [4 x [[S_FLOAT_TY]]]]* [[ARRS_PRIV]] to [[S_FLOAT_TY]]*
 
 // CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]]
@@ -695,23 +693,15 @@ int main() {
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
 // CHECK: [[VAR2_ORIG:%.+]] = load [[S_FLOAT_TY]]***, [[S_FLOAT_TY]]**** [[VAR2_ORIG_ADDR]],
 
-// CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 0
-// CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
-// CHECK: [[LOW:%.+]] = getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 1
-// CHECK: load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %{{.+}}, i64 4
-// CHECK: load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** %
-// CHECK: getelementptr inbounds [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 6
-// CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
-// CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
-// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
+// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[LD:%.+]] = load [[S_FLOAT_TY]]**, [[S_FLOAT_TY]]*** [[VAR2_ORIG]],
+// CHECK: [[ORIG_START:%.+]] = load [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]** [[LD]],
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -732,8 +722,6 @@ int main() {
 // CHECK: [[VVAR2_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VVAR2_ORIG_ADDR]],
 
 // CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 0
-// CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]], i64 0, i64 4
-// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
 // CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
@@ -741,6 +729,7 @@ int main() {
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VVAR2_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VVAR2_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -760,19 +749,15 @@ int main() {
 
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: [[LOW:%.+]] = getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 1
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], i64 0, i64 2
-// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[LAST:%.+]] = ptrtoint [[S_FLOAT_TY]]* %{{.+}} to i64
-// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
+// CHECK: [[FIRST:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW:%.+]] to i64
 // CHECK: [[BYTE_DIF:%.+]] = sub i64 [[LAST]], [[FIRST]]
 // CHECK: [[DIF:%.+]] = sdiv exact i64 [[BYTE_DIF]], ptrtoint ([[S_FLOAT_TY]]* getelementptr ([[S_FLOAT_TY]], [[S_FLOAT_TY]]* null, i32 1) to i64)
 // CHECK: [[SIZE:%.+]] = add nuw i64 [[DIF]], 1
 // CHECK: call i8* @llvm.stacksave()
 // CHECK: [[VAR3_PRIV:%.+]] = alloca [[S_FLOAT_TY]], i64 [[SIZE]],
+// CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
+// CHECK: [[ORIG_START:%.+]] = bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: [[START:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[ORIG_START]] to i64
 // CHECK: [[LOW_BOUND:%.+]] = ptrtoint [[S_FLOAT_TY]]* [[LOW]] to i64
 // CHECK: [[OFFSET_BYTES:%.+]] = sub i64 [[START]], [[LOW_BOUND]]
@@ -797,8 +782,8 @@ int main() {
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]], [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR:%.+]],
 // CHECK: [[VAR3_ORIG:%.+]] = load [2 x [[S_FLOAT_TY]]]*, [2 x [[S_FLOAT_TY]]]** [[VAR3_ORIG_ADDR]],
-// CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: getelementptr inbounds [2 x [[S_FLOAT_TY]]], [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], i32 0, i32 0
+// CHECK: bitcast [2 x [[S_FLOAT_TY]]]* [[VAR3_ORIG]] to [[S_FLOAT_TY]]*
 // CHECK: getelementptr [[S_FLOAT_TY]], [[S_FLOAT_TY]]* %{{.+}}, i64 2
 
 // CHECK: store [2 x [[S_FLOAT_TY]]]* [[VAR3_PRIV]], [2 x [[S_FLOAT_TY]]]** %
diff --git a/test/OpenMP/parallel_if_codegen.cpp b/test/OpenMP/parallel_if_codegen.cpp
index af154c0e36dd..87dcf1dde943 100644
--- a/test/OpenMP/parallel_if_codegen.cpp
+++ b/test/OpenMP/parallel_if_codegen.cpp
@@ -1,6 +1,6 @@
 // RUN: %clang_cc1 -verify -fopenmp -fopenmp-version=45 -x c++ -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck %s
 // RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -std=c++11 -triple %itanium_abi_triple -emit-pch -o %t %s
-// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck --check-prefix=CHECK %s
+// RUN: %clang_cc1 -fopenmp -fopenmp-version=45 -x c++ -triple %itanium_abi_triple -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s
 // expected-no-diagnostics
 #ifndef HEADER
 #define HEADER
diff --git a/test/OpenMP/taskgroup_ast_print.cpp b/test/OpenMP/taskgroup_ast_print.cpp
index 683141e686ec..4292eedfb56e 100644
--- a/test/OpenMP/taskgroup_ast_print.cpp
+++ b/test/OpenMP/taskgroup_ast_print.cpp
@@ -8,6 +8,62 @@
 
 void foo() {}
 
+struct S1 {
+  S1(): a(0) {}
+  S1(int v) : a(v) {}
+  int a;
+  typedef int type;
+  S1& operator +(const S1&);
+  S1& operator *(const S1&);
+  S1& operator &&(const S1&);
+  S1& operator ^(const S1&);
+};
+
+template <typename T>
+class S7 : public T {
+protected:
+  T a;
+  T b[100];
+  S7() : a(0) {}
+
+public:
+  S7(typename T::type v) : a(v) {
+#pragma omp taskgroup task_reduction(+ : a) task_reduction(*: b[:])
+    for (int k = 0; k < a.a; ++k)
+      ++this->a.a;
+  }
+  S7 &operator=(S7 &s) {
+#pragma omp taskgroup task_reduction(&& : this->a) task_reduction(^: b[s.a.a])
+    for (int k = 0; k < s.a.a; ++k)
+      ++s.a.a;
+    return *this;
+  }
+};
+
+// CHECK: #pragma omp taskgroup task_reduction(+: this->a) task_reduction(*: this->b[:])
+// CHECK: #pragma omp taskgroup task_reduction(&&: this->a) task_reduction(^: this->b[s.a.a])
+// CHECK: #pragma omp taskgroup task_reduction(+: this->a) task_reduction(*: this->b[:])
+
+class S8 : public S7<S1> {
+  S8() {}
+
+public:
+  S8(int v) : S7<S1>(v){
+#pragma omp taskgroup task_reduction(^ : S7 < S1 > ::a) task_reduction(+ : S7 < S1 > ::b[ : S7 < S1 > ::a.a])
+    for (int k = 0; k < a.a; ++k)
+      ++this->a.a;
+  }
+  S8 &operator=(S8 &s) {
+#pragma omp taskgroup task_reduction(* : this->a) task_reduction(&&:this->b[a.a:])
+    for (int k = 0; k < s.a.a; ++k)
+      ++s.a.a;
+    return *this;
+  }
+};
+
+// CHECK: #pragma omp taskgroup task_reduction(^: this->S7<S1>::a) task_reduction(+: this->S7<S1>::b[:this->S7<S1>::a.a])
+// CHECK: #pragma omp taskgroup task_reduction(*: this->a) task_reduction(&&: this->b[this->a.a:])
+
 int main (int argc, char **argv) {
   int b = argc, c, d, e, f, g;
   static int a;
@@ -18,9 +74,9 @@ int main (int argc, char **argv) {
 // CHECK-NEXT: a = 2;
 // CHECK-NEXT: ++a;
   ++a;
-#pragma omp taskgroup
+#pragma omp taskgroup task_reduction(min: a)
   foo();
-// CHECK-NEXT: #pragma omp taskgroup
+// CHECK-NEXT: #pragma omp taskgroup task_reduction(min: a)
 // CHECK-NEXT: foo();
 // CHECK-NEXT: return 0;
   return 0;
diff --git a/test/OpenMP/taskgroup_messages.cpp b/test/OpenMP/taskgroup_messages.cpp
index e893da4be3fa..23820734a551 100644
--- a/test/OpenMP/taskgroup_messages.cpp
+++ b/test/OpenMP/taskgroup_messages.cpp
@@ -63,5 +63,7 @@ int foo() {
     foo();
   }
 
+#pragma omp taskgroup init // expected-warning {{extra tokens at the end of '#pragma omp taskgroup' are ignored}}
+  ;
   return 0;
 }
diff --git a/test/OpenMP/taskgroup_task_reduction_messages.cpp b/test/OpenMP/taskgroup_task_reduction_messages.cpp
new file mode 100644
index 000000000000..9c8177b0bd43
--- /dev/null
+++ b/test/OpenMP/taskgroup_task_reduction_messages.cpp
@@ -0,0 +1,258 @@
+// RUN: %clang_cc1 -verify -fopenmp -ferror-limit 150 -o - %s
+// RUN: %clang_cc1 -verify -fopenmp -std=c++98 -ferror-limit 150 -o - %s
+// RUN: %clang_cc1 -verify -fopenmp -std=c++11 -ferror-limit 150 -o - %s
+
+void foo() {
+}
+
+bool foobool(int argc) {
+  return argc;
+}
+
+void foobar(int &ref) {
+#pragma omp taskgroup task_reduction(+:ref)
+  foo();
+}
+
+struct S1; // expected-note {{declared here}} expected-note 4 {{forward declaration of 'S1'}}
+extern S1 a;
+class S2 {
+  mutable int a;
+  S2 &operator+(const S2 &arg) { return (*this); } // expected-note 3 {{implicitly declared private here}}
+
+public:
+  S2() : a(0) {}
+  S2(S2 &s2) : a(s2.a) {}
+  static float S2s; // expected-note 2 {{static data member is predetermined as shared}}
+  static const float S2sc;
+};
+const float S2::S2sc = 0; // expected-note 2 {{'S2sc' defined here}}
+S2 b;                     // expected-note 3 {{'b' defined here}}
+const S2 ba[5];           // expected-note 2 {{'ba' defined here}}
+class S3 {
+  int a;
+
+public:
+  int b;
+  S3() : a(0) {}
+  S3(const S3 &s3) : a(s3.a) {}
+  S3 operator+(const S3 &arg1) { return arg1; }
+};
+int operator+(const S3 &arg1, const S3 &arg2) { return 5; }
+S3 c;               // expected-note 3 {{'c' defined here}}
+const S3 ca[5];     // expected-note 2 {{'ca' defined here}}
+extern const int f; // expected-note 4 {{'f' declared here}}
+class S4 {
+  int a;
+  S4(); // expected-note {{implicitly declared private here}}
+  S4(const S4 &s4);
+  S4 &operator+(const S4 &arg) { return (*this); }
+
+public:
+  S4(int v) : a(v) {}
+};
+S4 &operator&=(S4 &arg1, S4 &arg2) { return arg1; }
+class S5 {
+  int a;
+  S5() : a(0) {} // expected-note {{implicitly declared private here}}
+  S5(const S5 &s5) : a(s5.a) {}
+  S5 &operator+(const S5 &arg);
+
+public:
+  S5(int v) : a(v) {}
+};
+class S6 { // expected-note 3 {{candidate function (the implicit copy assignment operator) not viable: no known conversion from 'int' to 'const S6' for 1st argument}}
+#if __cplusplus >= 201103L // C++11 or later
+// expected-note@-2 3 {{candidate function (the implicit move assignment operator) not viable}}
+#endif
+  int a;
+
+public:
+  S6() : a(6) {}
+  operator int() { return 6; }
+} o;
+
+S3 h, k;
+#pragma omp threadprivate(h) // expected-note 2 {{defined as threadprivate or thread local}}
+
+template <class T>       // expected-note {{declared here}}
+T tmain(T argc) {
+  const T d = T();       // expected-note 4 {{'d' defined here}}
+  const T da[5] = {T()}; // expected-note 2 {{'da' defined here}}
+  T qa[5] = {T()};
+  T i;
+  T &j = i;                    // expected-note 2 {{'j' defined here}}
+  S3 &p = k;                   // expected-note 2 {{'p' defined here}}
+  const T &r = da[(int)i];     // expected-note 2 {{'r' defined here}}
+  T &q = qa[(int)i];
+  T fl;
+#pragma omp taskgroup task_reduction // expected-error {{expected '(' after 'task_reduction'}}
+  foo();
+#pragma omp taskgroup task_reduction + // expected-error {{expected '(' after 'task_reduction'}} expected-warning {{extra tokens at the end of '#pragma omp taskgroup' are ignored}}
+  foo();
+#pragma omp taskgroup task_reduction( // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+#pragma omp taskgroup task_reduction(- // expected-warning {{missing ':' after reduction identifier - ignoring}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+#pragma omp taskgroup task_reduction() // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(*) // expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(\) // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(& : argc // expected-error {{expected ')'}} expected-note {{to match this '('}} expected-error {{invalid operands to binary expression ('float' and 'float')}}
+  foo();
+#pragma omp taskgroup task_reduction(| : argc, // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}} expected-error {{invalid operands to binary expression ('float' and 'float')}}
+  foo();
+#pragma omp taskgroup task_reduction(|| : argc ? i : argc) // expected-error 2 {{expected variable name, array element or array section}}
+  foo();
+#pragma omp taskgroup task_reduction(foo : argc) //expected-error {{incorrect reduction identifier, expected one of '+', '-', '*', '&', '|', '^', '&&', '||', 'min' or 'max' or declare reduction for type 'float'}} expected-error {{incorrect reduction identifier, expected one of '+', '-', '*', '&', '|', '^', '&&', '||', 'min' or 'max' or declare reduction for type 'int'}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : argc)
+  foo();
+#pragma omp taskgroup task_reduction(^ : T) // expected-error {{'T' does not refer to a value}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : a, b, c, d, f) // expected-error {{a reduction list item with incomplete type 'S1'}} expected-error 3 {{const-qualified list item cannot be reduction}} expected-error 2 {{'operator+' is a private member of 'S2'}}
+  foo();
+#pragma omp taskgroup task_reduction(min : a, b, c, d, f) // expected-error {{a reduction list item with incomplete type 'S1'}} expected-error 4 {{arguments of OpenMP clause 'task_reduction' for 'min' or 'max' must be of arithmetic type}} expected-error 3 {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(max : h.b) // expected-error {{expected variable name, array element or array section}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : ba) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(* : ca) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(- : da) // expected-error {{const-qualified list item cannot be reduction}} expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : S2::S2sc) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : h, k) // expected-error {{threadprivate or thread local variable cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : o) // expected-error 2 {{no viable overloaded '='}}
+  foo();
+#pragma omp parallel private(k)
+#pragma omp taskgroup task_reduction(+ : p), task_reduction(+ : p) // expected-error 2 {{argument of OpenMP clause 'task_reduction' must reference the same object in all threads}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : p), task_reduction(+ : p) // expected-error 2 {{variable can appear only once in OpenMP 'task_reduction' clause}} expected-note 2 {{previously referenced here}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : r) // expected-error 2 {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp parallel shared(i)
+#pragma omp parallel reduction(min : i)
+#pragma omp taskgroup task_reduction(max : j) // expected-error 2 {{argument of OpenMP clause 'task_reduction' must reference the same object in all threads}}
+  foo();
+#pragma omp parallel
+#pragma omp for private(fl)
+  for (int i = 0; i < 10; ++i)
+#pragma omp taskgroup task_reduction(+ : fl)
+    foo();
+#pragma omp parallel
+#pragma omp for reduction(- : fl)
+  for (int i = 0; i < 10; ++i)
+#pragma omp taskgroup task_reduction(+ : fl)
+    foo();
+
+  return T();
+}
+
+namespace A {
+double x;
+#pragma omp threadprivate(x) // expected-note {{defined as threadprivate or thread local}}
+}
+namespace B {
+using A::x;
+}
+
+int main(int argc, char **argv) {
+  const int d = 5;       // expected-note 2 {{'d' defined here}}
+  const int da[5] = {0}; // expected-note {{'da' defined here}}
+  int qa[5] = {0};
+  S4 e(4);
+  S5 g(5);
+  int i;
+  int &j = i;                  // expected-note {{'j' defined here}}
+  S3 &p = k;                   // expected-note 2 {{'p' defined here}}
+  const int &r = da[i];        // expected-note {{'r' defined here}}
+  int &q = qa[i];
+  float fl;
+#pragma omp taskgroup task_reduction // expected-error {{expected '(' after 'task_reduction'}}
+  foo();
+#pragma omp taskgroup task_reduction + // expected-error {{expected '(' after 'task_reduction'}} expected-warning {{extra tokens at the end of '#pragma omp taskgroup' are ignored}}
+  foo();
+#pragma omp taskgroup task_reduction( // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+#pragma omp taskgroup task_reduction(- // expected-warning {{missing ':' after reduction identifier - ignoring}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+#pragma omp taskgroup task_reduction() // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(*) // expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(\) // expected-error {{expected unqualified-id}} expected-warning {{missing ':' after reduction identifier - ignoring}}
+  foo();
+#pragma omp taskgroup task_reduction(foo : argc // expected-error {{expected ')'}} expected-note {{to match this '('}} expected-error {{incorrect reduction identifier, expected one of '+', '-', '*', '&', '|', '^', '&&', '||', 'min' or 'max'}}
+  foo();
+#pragma omp taskgroup task_reduction(| : argc, // expected-error {{expected expression}} expected-error {{expected ')'}} expected-note {{to match this '('}}
+  foo();
+#pragma omp taskgroup task_reduction(|| : argc > 0 ? argv[1] : argv[2]) // expected-error {{expected variable name, array element or array section}}
+  foo();
+#pragma omp taskgroup task_reduction(~ : argc) // expected-error {{expected unqualified-id}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : argc)
+  foo();
+#pragma omp taskgroup task_reduction(^ : S1) // expected-error {{'S1' does not refer to a value}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : a, b, c, d, f) // expected-error {{a reduction list item with incomplete type 'S1'}} expected-error 2 {{const-qualified list item cannot be reduction}} expected-error {{'operator+' is a private member of 'S2'}}
+  foo();
+#pragma omp taskgroup task_reduction(min : a, b, c, d, f) // expected-error {{a reduction list item with incomplete type 'S1'}} expected-error 2 {{arguments of OpenMP clause 'task_reduction' for 'min' or 'max' must be of arithmetic type}} expected-error 2 {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(max : h.b) // expected-error {{expected variable name, array element or array section}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : ba) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(* : ca) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(- : da) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(^ : fl) // expected-error {{invalid operands to binary expression ('float' and 'float')}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : S2::S2s) // expected-error {{shared variable cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(&& : S2::S2sc) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(& : e, g) // expected-error {{calling a private constructor of class 'S4'}} expected-error {{nvalid operands to binary expression ('S4' and 'S4')}} expected-error {{calling a private constructor of class 'S5'}} expected-error {{invalid operands to binary expression ('S5' and 'S5')}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : h, k, B::x) // expected-error 2 {{threadprivate or thread local variable cannot be reduction}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : o) // expected-error {{no viable overloaded '='}}
+  foo();
+#pragma omp parallel private(k)
+#pragma omp taskgroup task_reduction(+ : p), task_reduction(+ : p) // expected-error 2 {{argument of OpenMP clause 'task_reduction' must reference the same object in all threads}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : p), task_reduction(+ : p) // expected-error {{variable can appear only once in OpenMP 'task_reduction' clause}} expected-note {{previously referenced here}}
+  foo();
+#pragma omp taskgroup task_reduction(+ : r) // expected-error {{const-qualified list item cannot be reduction}}
+  foo();
+#pragma omp parallel shared(i)
+#pragma omp parallel reduction(min : i)
+#pragma omp taskgroup task_reduction(max : j) // expected-error {{argument of OpenMP clause 'task_reduction' must reference the same object in all threads}}
+  foo();
+#pragma omp parallel
+#pragma omp for private(fl)
+  for (int i = 0; i < 10; ++i)
+#pragma omp taskgroup task_reduction(+ : fl)
+    foo();
+#pragma omp parallel
+#pragma omp for reduction(- : fl)
+  for (int i = 0; i < 10; ++i)
+#pragma omp taskgroup task_reduction(+ : fl)
+    foo();
+  static int m;
+#pragma omp taskgroup task_reduction(+ : m) // OK
+  m++;
+
+  return tmain(argc) + tmain(fl); // expected-note {{in instantiation of function template specialization 'tmain<int>' requested here}} expected-note {{in instantiation of function template specialization 'tmain<float>' requested here}}
+}
diff --git a/test/OpenMP/taskloop_codegen.cpp b/test/OpenMP/taskloop_codegen.cpp
index b1b45455802a..8de52023f33d 100644
--- a/test/OpenMP/taskloop_codegen.cpp
+++ b/test/OpenMP/taskloop_codegen.cpp
@@ -9,7 +9,7 @@
 int main(int argc, char **argv) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
@@ -24,7 +24,7 @@ int main(int argc, char **argv) {
 #pragma omp taskloop priority(argc)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
@@ -40,7 +40,7 @@ int main(int argc, char **argv) {
   for (int i = 0; i < 10; ++i)
     ;
 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[IF:%.+]] = icmp ne i32 %{{.+}}, 0
@@ -144,7 +144,7 @@ struct S {
   int a;
   S(int c) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
diff --git a/test/OpenMP/taskloop_firstprivate_codegen.cpp b/test/OpenMP/taskloop_firstprivate_codegen.cpp
index f4358db99b8c..6e8453050987 100644
--- a/test/OpenMP/taskloop_firstprivate_codegen.cpp
+++ b/test/OpenMP/taskloop_firstprivate_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type {{.*}}{ [2 x i32]*, i32, {{.*}}[2 x [[S_DOUBLE_TY]]]*, [[S_DOUBLE_TY]]*, i{{[0-9]+}}
@@ -57,7 +57,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[PRIVATES]], i{{.+}} 0, i{{.+}} 0
 // LAMBDA: [[G_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 0
@@ -101,7 +101,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[PRIVATES]], i{{.+}} 0, i{{.+}} 0
   // BLOCKS: [[G_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 0
@@ -190,7 +190,7 @@ int main() {
 //         [[KMP_TASK_T]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // Fill kmp_task_t->shareds by copying from original capture argument.
diff --git a/test/OpenMP/taskloop_lastprivate_codegen.cpp b/test/OpenMP/taskloop_lastprivate_codegen.cpp
index e689feba20ac..837977eb57c0 100644
--- a/test/OpenMP/taskloop_lastprivate_codegen.cpp
+++ b/test/OpenMP/taskloop_lastprivate_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { [2 x i32]*, i32*, [2 x [[S_DOUBLE_TY]]]*, [[S_DOUBLE_TY]]*, i{{[0-9]+}}* }
@@ -57,7 +57,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 
 // LAMBDA: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* bitcast (void ([[KMP_TASK_MAIN_TY:%[^*]+]]*, [[KMP_TASK_MAIN_TY]]*, i32)* [[MAIN_DUP:@.+]] to i8*))
@@ -98,7 +98,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* bitcast (void ([[KMP_TASK_MAIN_TY:%[^*]+]]*, [[KMP_TASK_MAIN_TY]]*, i32)* [[MAIN_DUP:@.+]] to i8*))
   // BLOCKS: ret
@@ -182,7 +182,7 @@ int main() {
 //         [[KMP_TASK_T]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // Fill kmp_task_t->shareds by copying from original capture argument.
diff --git a/test/OpenMP/taskloop_private_codegen.cpp b/test/OpenMP/taskloop_private_codegen.cpp
index dad9a17767fc..6ca7c402a12b 100644
--- a/test/OpenMP/taskloop_private_codegen.cpp
+++ b/test/OpenMP/taskloop_private_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { i8 }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
@@ -55,7 +55,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 // LAMBDA: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
 // LAMBDA: ret
@@ -91,7 +91,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
   // BLOCKS: ret
@@ -157,7 +157,7 @@ int main() {
 //         [[KMP_TASK_T_TY]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // CHECK: [[TASK:%.+]] = getelementptr inbounds [[KMP_TASK_MAIN_TY]], [[KMP_TASK_MAIN_TY]]* [[RES_KMP_TASK]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
diff --git a/test/OpenMP/taskloop_reduction_codegen.cpp b/test/OpenMP/taskloop_reduction_codegen.cpp
new file mode 100644
index 000000000000..af071ae8bd4e
--- /dev/null
+++ b/test/OpenMP/taskloop_reduction_codegen.cpp
@@ -0,0 +1,197 @@
+// RUN: %clang_cc1 -fopenmp -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu | FileCheck %s
+// expected-no-diagnostics
+
+struct S {
+  float a;
+  S() : a(0.0f) {}
+  ~S() {}
+};
+
+#pragma omp declare reduction(+:S:omp_out.a += omp_in.a) initializer(omp_priv = omp_orig)
+
+float g;
+
+int a;
+#pragma omp threadprivate(a)
+int main (int argc, char *argv[])
+{
+int   i, n;
+float a[100], b[100], sum, e[argc + 100];
+S c[100];
+float &d = g;
+
+/* Some initializations */
+n = 100;
+for (i=0; i < n; i++)
+  a[i] = b[i] = i * 1.0;
+sum = 0.0;
+
+#pragma omp taskloop reduction(+:sum, c[:n], d, e)
+  for (i=0; i < n; i++) {
+    sum = sum + (a[i] * b[i]);
+    c[i].a = i*i;
+    d += i*i;
+    e[i] = i;
+  }
+
+}
+
+// CHECK-LABEL: @main(
+// CHECK:    [[RETVAL:%.*]] = alloca i32,
+// CHECK:    [[ARGC_ADDR:%.*]] = alloca i32,
+// CHECK:    [[ARGV_ADDR:%.*]] = alloca i8**,
+// CHECK:    [[I:%.*]] = alloca i32,
+// CHECK:    [[N:%.*]] = alloca i32,
+// CHECK:    [[A:%.*]] = alloca [100 x float],
+// CHECK:    [[B:%.*]] = alloca [100 x float],
+// CHECK:    [[SUM:%.*]] = alloca float,
+// CHECK:    [[SAVED_STACK:%.*]] = alloca i8*,
+// CHECK:    [[C:%.*]] = alloca [100 x %struct.S],
+// CHECK:    [[D:%.*]] = alloca float*,
+// CHECK:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]],
+// CHECK:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%ident_t*
+// CHECK:    [[DOTRD_INPUT_:%.*]] = alloca [4 x %struct.kmp_task_red_input_t],
+// CHECK:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32,
+// CHECK:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32,
+// CHECK:    store i32 0, i32* [[RETVAL]],
+// CHECK:    store i32 [[ARGC:%.*]], i32* [[ARGC_ADDR]],
+// CHECK:    store i8** [[ARGV:%.*]], i8*** [[ARGV_ADDR]],
+// CHECK:    [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]],
+// CHECK:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 100
+// CHECK:    [[TMP2:%.*]] = zext i32 [[ADD]] to i64
+// CHECK:    [[VLA:%.+]] = alloca float, i64 %
+
+// CHECK:    call void @__kmpc_taskgroup(%ident_t*
+// CHECK-DAG:    [[TMP21:%.*]] = bitcast float* [[SUM]] to i8*
+// CHECK-DAG:    store i8* [[TMP21]], i8** [[TMP20:%[^,]+]],
+// CHECK-DAG:    [[TMP20]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T:%.+]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
+// CHECK-DAG:    store i64 4, i64* [[TMP22]],
+// CHECK-DAG:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT1:@.+]] to i8*), i8** [[TMP23]],
+// CHECK-DAG:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP24]],
+// CHECK-DAG:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB1:@.+]] to i8*), i8** [[TMP25]],
+// CHECK-DAG:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
+// CHECK-DAG:    [[TMP27:%.*]] = bitcast i32* [[TMP26]] to i8*
+// CHECK-DAG:    call void @llvm.memset.p0i8.i64(i8* [[TMP27]], i8 0, i64 4, i32 8, i1 false)
+// CHECK-DAG:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 0
+// CHECK-DAG:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, %
+// CHECK-DAG:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 [[LB_ADD_LEN]]
+// CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
+// CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
+// CHECK-DAG:    [[TMP28]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
+// CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
+// CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
+// CHECK-DAG:    [[TMP38]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
+// CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT2:@.+]] to i8*), i8** [[TMP39]],
+// CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_FINI2:@.+]] to i8*), i8** [[TMP40]],
+// CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB2:@.+]] to i8*), i8** [[TMP41]],
+// CHECK-DAG:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
+// CHECK-DAG:    store i32 1, i32* [[TMP42]],
+// CHECK-DAG:    [[TMP44:%.*]] = load float*, float** [[D]],
+// CHECK-DAG:    [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8*
+// CHECK-DAG:    store i8* [[TMP45]], i8** [[TMP43:%[^,]+]],
+// CHECK-DAG:    [[TMP43]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
+// CHECK-DAG:    store i64 4, i64* [[TMP46]],
+// CHECK-DAG:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT3:@.+]] to i8*), i8** [[TMP47]],
+// CHECK-DAG:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP48]],
+// CHECK-DAG:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB3:@.+]] to i8*), i8** [[TMP49]],
+// CHECK-DAG:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
+// CHECK-DAG:    [[TMP51:%.*]] = bitcast i32* [[TMP50]] to i8*
+// CHECK-DAG:    call void @llvm.memset.p0i8.i64(i8* [[TMP51]], i8 0, i64 4, i32 8, i1 false)
+// CHECK-DAG:    [[TMP53:%.*]] = bitcast float* [[VLA]] to i8*
+// CHECK-DAG:    store i8* [[TMP53]], i8** [[TMP52:%[^,]+]],
+// CHECK-DAG:    [[TMP52]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP54:%.*]] = mul nuw i64 [[TMP2]], 4
+// CHECK-DAG:    [[TMP55:%.*]] = udiv exact i64 [[TMP54]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    store i64 [[TMP54]], i64* [[TMP56:%[^,]+]],
+// CHECK-DAG:    [[TMP56]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 1
+// CHECK-DAG:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT4:@.+]] to i8*), i8** [[TMP57]],
+// CHECK-DAG:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP58]],
+// CHECK-DAG:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB4:@.+]] to i8*), i8** [[TMP59]],
+// CHECK-DAG:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 5
+// CHECK-DAG:    store i32 1, i32* [[TMP60]],
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_4]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_7]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_8]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK:    [[TMP61:%.*]] = bitcast [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]] to i8*
+// CHECK:    [[TMP62:%.*]] = call i8* @__kmpc_task_reduction_init(i32 [[TMP0]], i32 4, i8* [[TMP61]])
+// CHECK:    [[TMP63:%.*]] = load i32, i32* [[N]],
+// CHECK:    store i32 [[TMP63]], i32* [[DOTCAPTURE_EXPR_]],
+// CHECK:    [[TMP64:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]],
+// CHECK:    [[SUB:%.*]] = sub nsw i32 [[TMP64]], 0
+// CHECK:    [[SUB10:%.*]] = sub nsw i32 [[SUB]], 1
+// CHECK:    [[ADD11:%.*]] = add nsw i32 [[SUB10]], 1
+// CHECK:    [[DIV:%.*]] = sdiv i32 [[ADD11]], 1
+// CHECK:    [[SUB12:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK:    store i32 [[SUB12]], i32* [[DOTCAPTURE_EXPR_9]],
+// CHECK:    [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%ident_t* %{{.+}}, i32 [[TMP0]], i32 1, i64 888, i64 72, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @{{.+}} to i32 (i32, i8*)*))
+// CHECK:    call void @__kmpc_taskloop(%ident_t* %{{.+}}, i32 [[TMP0]], i8* [[TMP65]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
+// CHECK:    call void @__kmpc_end_taskgroup(%ident_t*
+
+// CHECK:    ret i32
+
+// CHECK: define internal void [[RED_INIT1]](i8*)
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB1]](i8*, i8*)
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT2]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_FINI2]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call void @
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB2]](i8*, i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT3]](i8*)
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB3]](i8*, i8*)
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT4]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB4]](i8*, i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
diff --git a/test/OpenMP/taskloop_simd_codegen.cpp b/test/OpenMP/taskloop_simd_codegen.cpp
index 48c6f479e9fa..f787689688e3 100644
--- a/test/OpenMP/taskloop_simd_codegen.cpp
+++ b/test/OpenMP/taskloop_simd_codegen.cpp
@@ -9,7 +9,7 @@
 int main(int argc, char **argv) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 33, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK1:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
@@ -24,7 +24,7 @@ int main(int argc, char **argv) {
 #pragma omp taskloop simd priority(argc)
   for (int i = 0; i < 10; ++i)
     ;
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK2:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
@@ -40,7 +40,7 @@ int main(int argc, char **argv) {
   for (int i = 0; i < 10; ++i)
     ;
 // CHECK: call void @__kmpc_taskgroup(%ident_t* [[DEFLOC]], i32 [[GTID]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 24, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK3:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[IF:%.+]] = icmp ne i32 %{{.+}}, 0
@@ -146,7 +146,7 @@ struct S {
   int a;
   S(int c) {
 // CHECK: [[GTID:%.+]] = call i32 @__kmpc_global_thread_num(%ident_t* [[DEFLOC:@.+]])
-// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 72, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
+// CHECK: [[TASKV:%.+]] = call i8* @__kmpc_omp_task_alloc(%ident_t* [[DEFLOC]], i32 [[GTID]], i32 1, i64 80, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, [[TDP_TY:%.+]]*)* [[TASK4:@.+]] to i32 (i32, i8*)*))
 // CHECK: [[TASK:%.+]] = bitcast i8* [[TASKV]] to [[TDP_TY]]*
 // CHECK: [[TASK_DATA:%.+]] = getelementptr inbounds [[TDP_TY]], [[TDP_TY]]* [[TASK]], i32 0, i32 0
 // CHECK: [[DOWN:%.+]] = getelementptr inbounds [[TD_TY:%.+]], [[TD_TY]]* [[TASK_DATA]], i32 0, i32 5
diff --git a/test/OpenMP/taskloop_simd_firstprivate_codegen.cpp b/test/OpenMP/taskloop_simd_firstprivate_codegen.cpp
index 230ef980e64d..9e268d8a1739 100644
--- a/test/OpenMP/taskloop_simd_firstprivate_codegen.cpp
+++ b/test/OpenMP/taskloop_simd_firstprivate_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type {{.*}}{ [2 x i32]*, i32, {{.*}}[2 x [[S_DOUBLE_TY]]]*, [[S_DOUBLE_TY]]*, i{{[0-9]+}}
@@ -57,7 +57,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[PRIVATES]], i{{.+}} 0, i{{.+}} 0
 // LAMBDA: [[G_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 0
@@ -101,7 +101,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[PRIVATES]], i{{.+}} 0, i{{.+}} 0
   // BLOCKS: [[G_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 0
@@ -190,7 +190,7 @@ int main() {
 //         [[KMP_TASK_T]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // Fill kmp_task_t->shareds by copying from original capture argument.
diff --git a/test/OpenMP/taskloop_simd_lastprivate_codegen.cpp b/test/OpenMP/taskloop_simd_lastprivate_codegen.cpp
index 41f04f9bdb3c..8d5ebb216808 100644
--- a/test/OpenMP/taskloop_simd_lastprivate_codegen.cpp
+++ b/test/OpenMP/taskloop_simd_lastprivate_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { [2 x i32]*, i32*, [2 x [[S_DOUBLE_TY]]]*, [[S_DOUBLE_TY]]*, i{{[0-9]+}}* }
@@ -57,7 +57,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 
 // LAMBDA: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* bitcast (void ([[KMP_TASK_MAIN_TY:%[^*]+]]*, [[KMP_TASK_MAIN_TY]]*, i32)* [[MAIN_DUP:@.+]] to i8*))
@@ -98,7 +98,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 16, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* bitcast (void ([[KMP_TASK_MAIN_TY:%[^*]+]]*, [[KMP_TASK_MAIN_TY]]*, i32)* [[MAIN_DUP:@.+]] to i8*))
   // BLOCKS: ret
@@ -182,7 +182,7 @@ int main() {
 //         [[KMP_TASK_T]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 40, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // Fill kmp_task_t->shareds by copying from original capture argument.
diff --git a/test/OpenMP/taskloop_simd_private_codegen.cpp b/test/OpenMP/taskloop_simd_private_codegen.cpp
index 493137ade65f..3833b5ea9fb3 100644
--- a/test/OpenMP/taskloop_simd_private_codegen.cpp
+++ b/test/OpenMP/taskloop_simd_private_codegen.cpp
@@ -23,7 +23,7 @@ struct S {
 
 volatile double g;
 
-// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32 }
+// CHECK-DAG: [[KMP_TASK_T_TY:%.+]] = type { i8*, i32 (i32, i8*)*, i32, %union{{.+}}, %union{{.+}}, i64, i64, i64, i32, i8* }
 // CHECK-DAG: [[S_DOUBLE_TY:%.+]] = type { double }
 // CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { i8 }
 // CHECK-DAG: [[PRIVATES_MAIN_TY:%.+]] = type {{.?}}{ [2 x [[S_DOUBLE_TY]]], [[S_DOUBLE_TY]], i32, [2 x i32]
@@ -55,7 +55,7 @@ int main() {
   // LAMBDA: call{{( x86_thiscallcc)?}} void [[OUTER_LAMBDA:@.+]](
   [&]() {
   // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
-  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // LAMBDA: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // LAMBDA: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
 // LAMBDA: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
 // LAMBDA: ret
@@ -91,7 +91,7 @@ int main() {
   // BLOCKS: call void {{%.+}}(i8
   ^{
   // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
-  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 88, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+  // BLOCKS: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc(%{{[^ ]+}} @{{[^,]+}}, i32 %{{[^,]+}}, i32 1, i64 96, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, %{{[^*]+}}*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
   // BLOCKS: [[PRIVATES:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* %{{.+}}, i{{.+}} 0, i{{.+}} 1
   // BLOCKS: call void @__kmpc_taskloop(%{{.+}}* @{{.+}}, i32 %{{.+}}, i8* [[RES]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
   // BLOCKS: ret
@@ -157,7 +157,7 @@ int main() {
 //         [[KMP_TASK_T_TY]] task_data;
 //         [[KMP_TASK_MAIN_TY]] privates;
 //       };
-// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 112, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
+// CHECK: [[RES:%.+]] = call i8* @__kmpc_omp_task_alloc([[LOC]], i32 [[GTID]], i32 9, i64 120, i64 1, i32 (i32, i8*)* bitcast (i32 (i32, [[KMP_TASK_MAIN_TY]]*)* [[TASK_ENTRY:@[^ ]+]] to i32 (i32, i8*)*))
 // CHECK: [[RES_KMP_TASK:%.+]] = bitcast i8* [[RES]] to [[KMP_TASK_MAIN_TY]]*
 
 // CHECK: [[TASK:%.+]] = getelementptr inbounds [[KMP_TASK_MAIN_TY]], [[KMP_TASK_MAIN_TY]]* [[RES_KMP_TASK]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
diff --git a/test/OpenMP/taskloop_simd_reduction_codegen.cpp b/test/OpenMP/taskloop_simd_reduction_codegen.cpp
new file mode 100644
index 000000000000..b96c8d50a3ff
--- /dev/null
+++ b/test/OpenMP/taskloop_simd_reduction_codegen.cpp
@@ -0,0 +1,197 @@
+// RUN: %clang_cc1 -fopenmp -x c++ %s -verify -debug-info-kind=limited -emit-llvm -o - -triple powerpc64le-unknown-linux-gnu | FileCheck %s
+// expected-no-diagnostics
+
+struct S {
+  float a;
+  S() : a(0.0f) {}
+  ~S() {}
+};
+
+#pragma omp declare reduction(+:S:omp_out.a += omp_in.a) initializer(omp_priv = omp_orig)
+
+float g;
+
+int a;
+#pragma omp threadprivate(a)
+int main (int argc, char *argv[])
+{
+int   i, n;
+float a[100], b[100], sum, e[argc + 100];
+S c[100];
+float &d = g;
+
+/* Some initializations */
+n = 100;
+for (i=0; i < n; i++)
+  a[i] = b[i] = i * 1.0;
+sum = 0.0;
+
+#pragma omp taskloop simd reduction(+:sum, c[:n], d, e)
+  for (i=0; i < n; i++) {
+    sum = sum + (a[i] * b[i]);
+    c[i].a = i*i;
+    d += i*i;
+    e[i] = i;
+  }
+
+}
+
+// CHECK-LABEL: @main(
+// CHECK:    [[RETVAL:%.*]] = alloca i32,
+// CHECK:    [[ARGC_ADDR:%.*]] = alloca i32,
+// CHECK:    [[ARGV_ADDR:%.*]] = alloca i8**,
+// CHECK:    [[I:%.*]] = alloca i32,
+// CHECK:    [[N:%.*]] = alloca i32,
+// CHECK:    [[A:%.*]] = alloca [100 x float],
+// CHECK:    [[B:%.*]] = alloca [100 x float],
+// CHECK:    [[SUM:%.*]] = alloca float,
+// CHECK:    [[SAVED_STACK:%.*]] = alloca i8*,
+// CHECK:    [[C:%.*]] = alloca [100 x %struct.S],
+// CHECK:    [[D:%.*]] = alloca float*,
+// CHECK:    [[AGG_CAPTURED:%.*]] = alloca [[STRUCT_ANON:%.*]],
+// CHECK:    [[TMP0:%.*]] = call i32 @__kmpc_global_thread_num(%ident_t*
+// CHECK:    [[DOTRD_INPUT_:%.*]] = alloca [4 x %struct.kmp_task_red_input_t],
+// CHECK:    [[DOTCAPTURE_EXPR_:%.*]] = alloca i32,
+// CHECK:    [[DOTCAPTURE_EXPR_9:%.*]] = alloca i32,
+// CHECK:    store i32 0, i32* [[RETVAL]],
+// CHECK:    store i32 [[ARGC:%.*]], i32* [[ARGC_ADDR]],
+// CHECK:    store i8** [[ARGV:%.*]], i8*** [[ARGV_ADDR]],
+// CHECK:    [[TMP1:%.*]] = load i32, i32* [[ARGC_ADDR]],
+// CHECK:    [[ADD:%.*]] = add nsw i32 [[TMP1]], 100
+// CHECK:    [[TMP2:%.*]] = zext i32 [[ADD]] to i64
+// CHECK:    [[VLA:%.+]] = alloca float, i64 %
+
+// CHECK:    call void @__kmpc_taskgroup(%ident_t*
+// CHECK-DAG:    [[TMP21:%.*]] = bitcast float* [[SUM]] to i8*
+// CHECK-DAG:    store i8* [[TMP21]], i8** [[TMP20:%[^,]+]],
+// CHECK-DAG:    [[TMP20]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T:%.+]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP22:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 1
+// CHECK-DAG:    store i64 4, i64* [[TMP22]],
+// CHECK-DAG:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT1:@.+]] to i8*), i8** [[TMP23]],
+// CHECK-DAG:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP24]],
+// CHECK-DAG:    [[TMP25:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB1:@.+]] to i8*), i8** [[TMP25]],
+// CHECK-DAG:    [[TMP26:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_]], i32 0, i32 5
+// CHECK-DAG:    [[TMP27:%.*]] = bitcast i32* [[TMP26]] to i8*
+// CHECK-DAG:    call void @llvm.memset.p0i8.i64(i8* [[TMP27]], i8 0, i64 4, i32 8, i1 false)
+// CHECK-DAG:    [[ARRAYIDX5:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 0
+// CHECK-DAG:    [[LB_ADD_LEN:%.*]] = add nsw i64 -1, %
+// CHECK-DAG:    [[ARRAYIDX6:%.*]] = getelementptr inbounds [100 x %struct.S], [100 x %struct.S]* [[C]], i64 0, i64 [[LB_ADD_LEN]]
+// CHECK-DAG:    [[TMP31:%.*]] = bitcast %struct.S* [[ARRAYIDX5]] to i8*
+// CHECK-DAG:    store i8* [[TMP31]], i8** [[TMP28:%[^,]+]],
+// CHECK-DAG:    [[TMP28]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP32:%.*]] = ptrtoint %struct.S* [[ARRAYIDX6]] to i64
+// CHECK-DAG:    [[TMP33:%.*]] = ptrtoint %struct.S* [[ARRAYIDX5]] to i64
+// CHECK-DAG:    [[TMP34:%.*]] = sub i64 [[TMP32]], [[TMP33]]
+// CHECK-DAG:    [[TMP35:%.*]] = sdiv exact i64 [[TMP34]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    [[TMP36:%.*]] = add nuw i64 [[TMP35]], 1
+// CHECK-DAG:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    store i64 [[TMP37]], i64* [[TMP38:%[^,]+]],
+// CHECK-DAG:    [[TMP38]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 1
+// CHECK-DAG:    [[TMP39:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT2:@.+]] to i8*), i8** [[TMP39]],
+// CHECK-DAG:    [[TMP40:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 3
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_FINI2:@.+]] to i8*), i8** [[TMP40]],
+// CHECK-DAG:    [[TMP41:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB2:@.+]] to i8*), i8** [[TMP41]],
+// CHECK-DAG:    [[TMP42:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_4]], i32 0, i32 5
+// CHECK-DAG:    store i32 1, i32* [[TMP42]],
+// CHECK-DAG:    [[TMP44:%.*]] = load float*, float** [[D]],
+// CHECK-DAG:    [[TMP45:%.*]] = bitcast float* [[TMP44]] to i8*
+// CHECK-DAG:    store i8* [[TMP45]], i8** [[TMP43:%[^,]+]],
+// CHECK-DAG:    [[TMP43]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP46:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 1
+// CHECK-DAG:    store i64 4, i64* [[TMP46]],
+// CHECK-DAG:    [[TMP47:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT3:@.+]] to i8*), i8** [[TMP47]],
+// CHECK-DAG:    [[TMP48:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP48]],
+// CHECK-DAG:    [[TMP49:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB3:@.+]] to i8*), i8** [[TMP49]],
+// CHECK-DAG:    [[TMP50:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_7]], i32 0, i32 5
+// CHECK-DAG:    [[TMP51:%.*]] = bitcast i32* [[TMP50]] to i8*
+// CHECK-DAG:    call void @llvm.memset.p0i8.i64(i8* [[TMP51]], i8 0, i64 4, i32 8, i1 false)
+// CHECK-DAG:    [[TMP53:%.*]] = bitcast float* [[VLA]] to i8*
+// CHECK-DAG:    store i8* [[TMP53]], i8** [[TMP52:%[^,]+]],
+// CHECK-DAG:    [[TMP52]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8:%.+]], i32 0, i32 0
+// CHECK-DAG:    [[TMP54:%.*]] = mul nuw i64 [[TMP2]], 4
+// CHECK-DAG:    [[TMP55:%.*]] = udiv exact i64 [[TMP54]], ptrtoint (float* getelementptr (float, float* null, i32 1) to i64)
+// CHECK-DAG:    store i64 [[TMP54]], i64* [[TMP56:%[^,]+]],
+// CHECK-DAG:    [[TMP56]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 1
+// CHECK-DAG:    [[TMP57:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 2
+// CHECK-DAG:    store i8* bitcast (void (i8*)* [[RED_INIT4:@.+]] to i8*), i8** [[TMP57]],
+// CHECK-DAG:    [[TMP58:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 3
+// CHECK-DAG:    store i8* null, i8** [[TMP58]],
+// CHECK-DAG:    [[TMP59:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 4
+// CHECK-DAG:    store i8* bitcast (void (i8*, i8*)* [[RED_COMB4:@.+]] to i8*), i8** [[TMP59]],
+// CHECK-DAG:    [[TMP60:%.*]] = getelementptr inbounds [[STRUCT_KMP_TASK_RED_INPUT_T]], %struct.kmp_task_red_input_t* [[DOTRD_INPUT_GEP_8]], i32 0, i32 5
+// CHECK-DAG:    store i32 1, i32* [[TMP60]],
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_4]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_7]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK-DAG:    [[DOTRD_INPUT_GEP_8]] = getelementptr inbounds [4 x %struct.kmp_task_red_input_t], [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]], i64 0, i64
+// CHECK:    [[TMP61:%.*]] = bitcast [4 x %struct.kmp_task_red_input_t]* [[DOTRD_INPUT_]] to i8*
+// CHECK:    [[TMP62:%.*]] = call i8* @__kmpc_task_reduction_init(i32 [[TMP0]], i32 4, i8* [[TMP61]])
+// CHECK:    [[TMP63:%.*]] = load i32, i32* [[N]],
+// CHECK:    store i32 [[TMP63]], i32* [[DOTCAPTURE_EXPR_]],
+// CHECK:    [[TMP64:%.*]] = load i32, i32* [[DOTCAPTURE_EXPR_]],
+// CHECK:    [[SUB:%.*]] = sub nsw i32 [[TMP64]], 0
+// CHECK:    [[SUB10:%.*]] = sub nsw i32 [[SUB]], 1
+// CHECK:    [[ADD11:%.*]] = add nsw i32 [[SUB10]], 1
+// CHECK:    [[DIV:%.*]] = sdiv i32 [[ADD11]], 1
+// CHECK:    [[SUB12:%.*]] = sub nsw i32 [[DIV]], 1
+// CHECK:    store i32 [[SUB12]], i32* [[DOTCAPTURE_EXPR_9]],
+// CHECK:    [[TMP65:%.*]] = call i8* @__kmpc_omp_task_alloc(%ident_t* %{{.+}}, i32 [[TMP0]], i32 1, i64 888, i64 72, i32 (i32, i8*)* bitcast (i32 (i32, %struct.kmp_task_t_with_privates*)* @{{.+}} to i32 (i32, i8*)*))
+// CHECK:    call void @__kmpc_taskloop(%ident_t* %{{.+}}, i32 [[TMP0]], i8* [[TMP65]], i32 1, i64* %{{.+}}, i64* %{{.+}}, i64 %{{.+}}, i32 0, i32 0, i64 0, i8* null)
+// CHECK:    call void @__kmpc_end_taskgroup(%ident_t*
+
+// CHECK:    ret i32
+
+// CHECK: define internal void [[RED_INIT1]](i8*)
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB1]](i8*, i8*)
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT2]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_FINI2]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: call void @
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB2]](i8*, i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT3]](i8*)
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB3]](i8*, i8*)
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_INIT4]](i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: store float 0.000000e+00, float* %
+// CHECK: ret void
+
+// CHECK: define internal void [[RED_COMB4]](i8*, i8*)
+// CHECK: call i8* @__kmpc_threadprivate_cached(
+// CHECK: fadd float %
+// CHECK: store float %{{.+}}, float* %
+// CHECK: ret void
+