[OPENMP] Propagate alignment from original variables to the private copies.

Currently private copies of captured variables have default alignment. Patch makes private variables to have same alignment as original variables.

llvm-svn: 247260

5 files changed, 63 insertions, 63 deletions

diff --git a/clang/test/OpenMP/for_lastprivate_codegen.cpp b/clang/test/OpenMP/for_lastprivate_codegen.cpp
index cada3a06e58..9cfb73fd091 100644
--- a/clang/test/OpenMP/for_lastprivate_codegen.cpp
+++ b/clang/test/OpenMP/for_lastprivate_codegen.cpp
@@ -18,7 +18,7 @@ struct S {
   ~S() {}
 };
 
-volatile int g = 1212;
+volatile int g __attribute__((aligned(128)))= 1212;
 volatile int &g1 = g;
 float f;
 char cnt;
@@ -32,10 +32,10 @@ char cnt;
 template <typename T>
 T tmain() {
   S<T> test;
-  T t_var = T();
-  T vec[] = {1, 2};
-  S<T> s_arr[] = {1, 2};
-  S<T> &var = test;
+  T t_var __attribute__((aligned(128))) = T();
+  T vec[] __attribute__((aligned(128))) = {1, 2};
+  S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
+  S<T> &var __attribute__((aligned(128))) = test;
 #pragma omp parallel
 #pragma omp for lastprivate(t_var, vec, s_arr, var)
   for (int i = 0; i < 2; ++i) {
@@ -69,7 +69,7 @@ int main() {
     // LAMBDA: alloca i{{[0-9]+}},
     // LAMBDA: alloca i{{[0-9]+}},
     // LAMBDA: alloca i{{[0-9]+}},
-    // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
+    // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128
     // LAMBDA: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %{{.+}}
     // LAMBDA: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
     // LAMBDA: call {{.+}} @__kmpc_for_static_init_4(%{{.+}}* @{{.+}}, i32 [[GTID]], i32 34, i32* [[IS_LAST_ADDR:%.+]], i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
@@ -122,7 +122,7 @@ int main() {
     // BLOCKS: alloca i{{[0-9]+}},
     // BLOCKS: alloca i{{[0-9]+}},
     // BLOCKS: alloca i{{[0-9]+}},
-    // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
+    // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128
     // BLOCKS: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %{{.+}}
     // BLOCKS: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
     // BLOCKS: call {{.+}} @__kmpc_for_static_init_4(%{{.+}}* @{{.+}}, i32 [[GTID]], i32 34, i32* [[IS_LAST_ADDR:%.+]], i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
@@ -402,10 +402,10 @@ int main() {
 // CHECK: alloca i{{[0-9]+}},
 // CHECK: alloca i{{[0-9]+}},
 // CHECK: alloca i{{[0-9]+}},
-// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
-// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
-// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
-// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
+// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
+// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], align 128
+// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], align 128
+// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128
 // CHECK: [[VAR_PRIV_REF:%.+]] = alloca [[S_INT_TY]]*,
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
 
diff --git a/clang/test/OpenMP/parallel_copyin_codegen.cpp b/clang/test/OpenMP/parallel_copyin_codegen.cpp
index f29fb8b6419..e7c69697e29 100644
--- a/clang/test/OpenMP/parallel_copyin_codegen.cpp
+++ b/clang/test/OpenMP/parallel_copyin_codegen.cpp
@@ -17,7 +17,7 @@
 #ifndef HEADER
 #define HEADER
 
-volatile int g = 1212;
+volatile int g __attribute__((aligned(128))) = 1212;
 #pragma omp threadprivate(g)
 
 template <class T>
@@ -58,10 +58,10 @@ template <typename T>
 T tmain() {
   S<T> test;
   test = S<T>();
-  static T t_var = 333;
-  static T vec[] = {3, 3};
-  static S<T> s_arr[] = {1, 2};
-  static S<T> var(3);
+  static T t_var __attribute__((aligned(128))) = 333;
+  static T vec[] __attribute__((aligned(128))) = {3, 3};
+  static S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
+  static S<T> var __attribute__((aligned(128))) (3);
 #pragma omp threadprivate(t_var, vec, s_arr, var)
 #pragma omp parallel copyin(t_var, vec, s_arr, var)
   {
@@ -103,8 +103,8 @@ int main() {
     // LAMBDA: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[G]] to i{{[0-9]+}}), %{{.+}}
     // LAMBDA: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
     // LAMBDA: [[NOT_MASTER]]
-    // LAMBDA: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]],
-    // LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
+    // LAMBDA: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]], align 128
+    // LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
     // LAMBDA: [[DONE]]
 
     // TLS-LAMBDA-DAG: [[G_CAPTURE_SRC:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
@@ -115,7 +115,7 @@ int main() {
     // TLS-LAMBDA: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
     // TLS-LAMBDA: [[NOT_MASTER]]
     // TLS-LAMBDA: load i{{[0-9]+}}, i{{[0-9]+}}* [[G_CAPTURE_SRC]],
-    // TLS-LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]],
+    // TLS-LAMBDA: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
     // TLS-LAMBDA: [[DONE]]
 
     // LAMBDA: call {{.*}}i32 @__kmpc_cancel_barrier(
@@ -130,7 +130,7 @@ int main() {
       // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]]
 
       // TLS-LAMBDA: [[G_CAPTURE_DST:%.+]] = call i{{[0-9]+}}* [[G_CTOR]]()
-      // TLS-LAMBDA: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_CAPTURE_DST]]
+      // TLS-LAMBDA: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
     }();
   }
   }();
@@ -164,8 +164,8 @@ int main() {
     // BLOCKS: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[G]] to i{{[0-9]+}}), %{{.+}}
     // BLOCKS: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
     // BLOCKS: [[NOT_MASTER]]
-    // BLOCKS: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]],
-    // BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
+    // BLOCKS: load i{{[0-9]+}}, i{{[0-9]+}}* [[G]], align 128
+    // BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
     // BLOCKS: [[DONE]]
 
     // TLS-BLOCKS-DAG: [[G_CAPTURE_SRC:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** %
@@ -176,7 +176,7 @@ int main() {
     // TLS-BLOCKS: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
     // TLS-BLOCKS: [[NOT_MASTER]]
     // TLS-BLOCKS: load i{{[0-9]+}}, i{{[0-9]+}}* [[G_CAPTURE_SRC]],
-    // TLS-BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]],
+    // TLS-BLOCKS: store volatile i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* [[G_CAPTURE_DST]], align 128
     // TLS-BLOCKS: [[DONE]]
 
     // BLOCKS: call {{.*}}i32 @__kmpc_cancel_barrier(
@@ -381,8 +381,8 @@ int main() {
 // CHECK: icmp ne i{{[0-9]+}} ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}}), %{{.+}}
 // CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
 // CHECK: [[NOT_MASTER]]
-// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[TMAIN_T_VAR]],
-// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}},
+// CHECK: load i{{[0-9]+}}, i{{[0-9]+}}* [[TMAIN_T_VAR]], align 128
+// CHECK: store i{{[0-9]+}} %{{.+}}, i{{[0-9]+}}* %{{.+}}, align 128
 
 // TLS-CHECK: [[MASTER_REF:%.+]] = load i32*, i32** %
 // TLS-CHECK: [[MASTER_REF1:%.+]] = load [2 x i32]*, [2 x i32]** %
@@ -393,8 +393,8 @@ int main() {
 // TLS-CHECK: icmp ne i{{[0-9]+}} [[MASTER_LONG]], ptrtoint (i{{[0-9]+}}* [[TMAIN_T_VAR]] to i{{[0-9]+}})
 // TLS-CHECK: br i1 %{{.+}}, label %[[NOT_MASTER:.+]], label %[[DONE:.+]]
 // TLS-CHECK: [[NOT_MASTER]]
-// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]]
-// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[TMAIN_T_VAR]]
+// TLS-CHECK: [[MASTER_VAL:%.+]] = load i32, i32* [[MASTER_REF]],
+// TLS-CHECK: store i32 [[MASTER_VAL]], i32* [[TMAIN_T_VAR]], align 128
 
 // threadprivate_vec = vec;
 // CHECK: call {{.*}}i8* @__kmpc_threadprivate_cached({{.+}} [[TMAIN_VEC]]
diff --git a/clang/test/OpenMP/parallel_firstprivate_codegen.cpp b/clang/test/OpenMP/parallel_firstprivate_codegen.cpp
index 419af7992b7..a3f28a930c2 100644
--- a/clang/test/OpenMP/parallel_firstprivate_codegen.cpp
+++ b/clang/test/OpenMP/parallel_firstprivate_codegen.cpp
@@ -16,7 +16,7 @@ struct St {
   ~St() {}
 };
 
-volatile int g = 1212;
+volatile int g __attribute__((aligned(128))) = 1212;
 
 template <class T>
 struct S {
@@ -36,10 +36,10 @@ struct S {
 template <typename T>
 T tmain() {
   S<T> test;
-  T t_var = T();
-  T vec[] = {1, 2};
-  S<T> s_arr[] = {1, 2};
-  S<T> var(3);
+  T t_var __attribute__((aligned(128))) = T();
+  T vec[] __attribute__((aligned(128))) = {1, 2};
+  S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
+  S<T> var __attribute__((aligned(128))) (3);
 #pragma omp parallel firstprivate(t_var, vec, s_arr, var)
   {
     vec[0] = t_var;
@@ -61,7 +61,7 @@ int main() {
 #pragma omp parallel firstprivate(g)
   {
     // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}})
-    // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
+    // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128
     // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]]
     // LAMBDA: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]]
     // LAMBDA: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
@@ -93,10 +93,10 @@ int main() {
 #pragma omp parallel firstprivate(g)
   {
     // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}})
-    // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
+    // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}}, align 128
     // BLOCKS: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]]
-    // BLOCKS: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]]
-    // BLOCKS: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
+    // BLOCKS: [[G_VAL:%.+]] = load volatile i{{[0-9]+}}, i{{[0-9]+}}* [[G_REF]],
+    // BLOCKS: store i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
     // BLOCKS: call {{.*}}i32 @__kmpc_cancel_barrier(
     g = 1;
     // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
@@ -185,10 +185,10 @@ int main() {
 // CHECK: ret
 //
 // CHECK: define internal {{.*}}void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [2 x i32]* dereferenceable(8) %{{.+}}, i32* dereferenceable(4) %{{.+}}, [2 x [[S_INT_TY]]]* dereferenceable(8) %{{.+}}, [[S_INT_TY]]* dereferenceable(4) %{{.+}})
-// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
-// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
-// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
-// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
+// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
+// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], align 128
+// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], align 128
+// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]],
 
 // CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** %
diff --git a/clang/test/OpenMP/parallel_private_codegen.cpp b/clang/test/OpenMP/parallel_private_codegen.cpp
index 616d50fae9f..ec26c6066ca 100644
--- a/clang/test/OpenMP/parallel_private_codegen.cpp
+++ b/clang/test/OpenMP/parallel_private_codegen.cpp
@@ -16,17 +16,17 @@ struct S {
   ~S() {}
 };
 
-volatile int g = 1212;
+volatile int g __attribute__((aligned(128))) = 1212;
 
 // CHECK: [[S_FLOAT_TY:%.+]] = type { float }
 // CHECK: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
 template <typename T>
 T tmain() {
   S<T> test;
-  T t_var = T();
-  T vec[] = {1, 2};
-  S<T> s_arr[] = {1, 2};
-  S<T> var(3);
+  T t_var __attribute__((aligned(128))) = T();
+  T vec[] __attribute__((aligned(128))) = {1, 2};
+  S<T> s_arr[] __attribute__((aligned(128))) = {1, 2};
+  S<T> var __attribute__((aligned(128))) (3);
 #pragma omp parallel private(t_var, vec, s_arr, var)
   {
     vec[0] = t_var;
@@ -49,7 +49,7 @@ int main() {
     // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
     // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
     g = 1;
-    // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
+    // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
     // LAMBDA: store i{{[0-9]+}}* [[G_PRIVATE_ADDR]], i{{[0-9]+}}** [[G_PRIVATE_ADDR_REF]]
     // LAMBDA: call{{.*}} void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]])
@@ -78,7 +78,7 @@ int main() {
     // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}})
     // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
     g = 1;
-    // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
+    // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
     // BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
@@ -143,10 +143,10 @@ int main() {
 // CHECK: ret
 //
 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}})
-// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
-// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
-// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]],
-// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
+// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
+// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}], align 128
+// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_INT_TY]]], align 128
+// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128
 // CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
 // CHECK-NOT: [[T_VAR_PRIV]]
 // CHECK-NOT: [[VEC_PRIV]]
diff --git a/clang/test/OpenMP/parallel_reduction_codegen.cpp b/clang/test/OpenMP/parallel_reduction_codegen.cpp
index 6ecd42bb94c..2ee4239d2a5 100644
--- a/clang/test/OpenMP/parallel_reduction_codegen.cpp
+++ b/clang/test/OpenMP/parallel_reduction_codegen.cpp
@@ -8,7 +8,7 @@
 #ifndef HEADER
 #define HEADER
 
-volatile int g = 1212;
+volatile int g __attribute__((aligned(128))) = 1212;
 
 template <class T>
 struct S {
@@ -30,10 +30,10 @@ template <typename T>
 T tmain() {
   T t;
   S<T> test;
-  T t_var = T(), t_var1;
+  T t_var __attribute__((aligned(128))) = T(), t_var1 __attribute__((aligned(128)));
   T vec[] = {1, 2};
-  S<T> s_arr[] = {1, 2};
-  S<T> var(3), var1;
+  S<T> s_arr[]  = {1, 2};
+  S<T> var __attribute__((aligned(128))) (3), var1 __attribute__((aligned(128)));
 #pragma omp parallel reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1)
   {
     vec[0] = t_var;
@@ -59,9 +59,9 @@ int main() {
     // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*],
 
     // LAMBDA: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]]
-    // LAMBDA: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
+    // LAMBDA: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     g = 1;
-    // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
+    // LAMBDA: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
     // LAMBDA: store i{{[0-9]+}}* [[G_PRIVATE_ADDR]], i{{[0-9]+}}** [[G_PRIVATE_ADDR_REF]]
     // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]])
@@ -114,9 +114,9 @@ int main() {
     // BLOCKS: [[RED_LIST:%.+]] = alloca [1 x i8*],
 
     // BLOCKS: [[G_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[G_REF_ADDR:%.+]]
-    // BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
+    // BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     g = 1;
-    // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
+    // BLOCKS: store i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]], align 128
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
     // BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
     // BLOCKS-NOT: [[G]]{{[[^:word:]]}}
@@ -435,10 +435,10 @@ int main() {
 // CHECK: ret
 //
 // CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
-// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
-// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
-// CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]],
-// CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}},
+// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
+// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128
+// CHECK: [[VAR1_PRIV:%.+]] = alloca [[S_INT_TY]], align 128
+// CHECK: [[T_VAR1_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
 
 // Reduction list for runtime.
 // CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*],