[OPENMP 4.5] Codegen for data members in 'reduction' clause.
OpenMP 4.5 allows to privatize non-static data members of current class
in non-static member functions. Patch supports codegen for non-static
data members in 'reduction' clauses.
llvm-svn: 262460
diff --git a/clang/test/OpenMP/parallel_reduction_codegen.cpp b/clang/test/OpenMP/parallel_reduction_codegen.cpp
index 05224d0..703750c 100644
--- a/clang/test/OpenMP/parallel_reduction_codegen.cpp
+++ b/clang/test/OpenMP/parallel_reduction_codegen.cpp
@@ -20,6 +20,62 @@
~S() {}
};
+struct SS {
+ int a;
+ int b : 4;
+ int &c;
+ SS(int &d) : a(0), b(0), c(d) {
+#pragma omp parallel reduction(+: a, b, c)
+#ifdef LAMBDA
+ [&]() {
+ ++this->a, --b, (this)->c /= 1;
+#pragma omp parallel reduction(&: a, b, c)
+ ++(this)->a, --b, this->c /= 1;
+ }();
+#elif defined(BLOCKS)
+ ^{
+ ++a;
+ --this->b;
+ (this)->c /= 1;
+#pragma omp parallel reduction(-: a, b, c)
+ ++(this)->a, --b, this->c /= 1;
+ }();
+#else
+ ++this->a, --b, c /= 1;
+#endif
+ }
+};
+
+template<typename T>
+struct SST {
+ T a;
+ SST() : a(T()) {
+#pragma omp parallel reduction(*: a)
+#ifdef LAMBDA
+ [&]() {
+ [&]() {
+ ++this->a;
+#pragma omp parallel reduction(&& :a)
+ ++(this)->a;
+ }();
+ }();
+#elif defined(BLOCKS)
+ ^{
+ ^{
+ ++a;
+#pragma omp parallel reduction(|: a)
+ ++(this)->a;
+ }();
+ }();
+#else
+ ++(this)->a;
+#endif
+ }
+};
+
+// CHECK: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8
+// LAMBDA: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8
+// BLOCKS: [[SS_TY:%.+]] = type { i{{[0-9]+}}, i8
// CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float }
// CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
// CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8*
@@ -29,6 +85,7 @@
T tmain() {
T t;
S<T> test;
+ SST<T> sst;
T t_var __attribute__((aligned(128))) = T(), t_var1 __attribute__((aligned(128)));
T vec[] = {1, 2};
S<T> s_arr[] = {1, 2};
@@ -41,16 +98,62 @@
return T();
}
+int sivar;
int main() {
+ SS ss(sivar);
#ifdef LAMBDA
// LAMBDA: [[G:@.+]] = global i{{[0-9]+}} 1212,
// LAMBDA-LABEL: @main
- // LAMBDA: call void [[OUTER_LAMBDA:@.+]](
+ // LAMBDA: alloca [[SS_TY]],
+ // LAMBDA: alloca [[CAP_TY:%.+]],
+ // LAMBDA: call{{.*}} void [[OUTER_LAMBDA:@[^(]+]]([[CAP_TY]]*
[&]() {
// LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]](
// LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i32* [[G]])
#pragma omp parallel reduction(+:g)
{
+ // LAMBDA: define {{.+}} @{{.+}}([[SS_TY]]*
+ // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+ // LAMBDA: store i{{[0-9]+}} 0, i{{[0-9]+}}* %
+ // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+ // LAMBDA: store i8
+ // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+ // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+ // LAMBDA-NOT: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+ // LAMBDA: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+ // LAMBDA: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i32*, i32*, i32*)* [[SS_MICROTASK:@.+]] to void
+ // LAMBDA: [[B_REF:%.+]] = getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* %{{.*}}, i32 0, i32 1
+ // LAMBDA: store i8 %{{.+}}, i8* [[B_REF]],
+ // LAMBDA: ret
+
+ // LAMBDA: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32* {{.+}}, i32* {{.+}}, i32* {{.+}})
+ // LAMBDA-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* %
+ // LAMBDA: call{{.*}} void
+ // LAMBDA: ret void
+
+ // LAMBDA: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]*
+ // LAMBDA: [[A_PRIV:%.+]] = alloca i{{[0-9]+}},
+ // LAMBDA: [[B_PRIV:%.+]] = alloca i{{[0-9]+}},
+ // LAMBDA: [[C_PRIV:%.+]] = alloca i{{[0-9]+}},
+ // LAMBDA: store i{{[0-9]+}} -1, i{{[0-9]+}}* [[A_PRIV]],
+ // LAMBDA: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]],
+ // LAMBDA: store i{{[0-9]+}} -1, i{{[0-9]+}}* [[B_PRIV]],
+ // LAMBDA: store i{{[0-9]+}} -1, i{{[0-9]+}}* [[C_PRIV]],
+ // LAMBDA: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]],
+ // LAMBDA: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]],
+ // LAMBDA-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]],
+ // LAMBDA-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1
+ // LAMBDA-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]],
+ // LAMBDA-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]],
+ // LAMBDA-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1
+ // LAMBDA-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]],
+ // LAMBDA-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]],
+ // LAMBDA-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]],
+ // LAMBDA-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1
+ // LAMBDA-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]],
+ // LAMBDA: call i32 @__kmpc_reduce_nowait(
+ // LAMBDA: ret void
+
// LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* noalias %{{.+}}, i32* noalias %{{.+}}, i32* dereferenceable(4) %{{.+}})
// LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
@@ -100,6 +203,7 @@
#elif defined(BLOCKS)
// BLOCKS: [[G:@.+]] = global i{{[0-9]+}} 1212,
// BLOCKS-LABEL: @main
+ // BLOCKS: call
// BLOCKS: call void {{%.+}}(i8
^{
// BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8*
@@ -152,6 +256,47 @@
}
}();
return 0;
+// BLOCKS: define {{.+}} @{{.+}}([[SS_TY]]*
+// BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+// BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* %
+// BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+// BLOCKS: store i8
+// BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+// BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+// BLOCKS-NOT: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+// BLOCKS: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+// BLOCKS: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i32*, i32*, i32*)* [[SS_MICROTASK:@.+]] to void
+// BLOCKS: [[B_REF:%.+]] = getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* %{{.*}}, i32 0, i32 1
+// BLOCKS: store i8 %{{.+}}, i8* [[B_REF]],
+// BLOCKS: ret
+
+// BLOCKS: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32* {{.+}}, i32* {{.+}}, i32* {{.+}})
+// BLOCKS-NOT: getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* %
+// BLOCKS: call{{.*}} void
+// BLOCKS: ret void
+
+// BLOCKS: define internal void @{{.+}}(i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]* %{{.+}}, i32* {{.+}}, i32* {{.+}}, i32* {{.+}})
+// BLOCKS: [[A_PRIV:%.+]] = alloca i{{[0-9]+}},
+// BLOCKS: [[B_PRIV:%.+]] = alloca i{{[0-9]+}},
+// BLOCKS: [[C_PRIV:%.+]] = alloca i{{[0-9]+}},
+// BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[A_PRIV]],
+// BLOCKS: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]],
+// BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[B_PRIV]],
+// BLOCKS: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[C_PRIV]],
+// BLOCKS: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]],
+// BLOCKS: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]],
+// BLOCKS-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]],
+// BLOCKS-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1
+// BLOCKS-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]],
+// BLOCKS-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]],
+// BLOCKS-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1
+// BLOCKS-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]],
+// BLOCKS-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]],
+// BLOCKS-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]],
+// BLOCKS-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1
+// BLOCKS-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]],
+// BLOCKS: call i32 @__kmpc_reduce_nowait(
+// BLOCKS: ret void
#else
S<float> test;
float t_var = 0, t_var1;
@@ -472,6 +617,43 @@
// CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]*
// CHECK: ret
//
+// CHECK: define {{.+}} @{{.+}}([[SS_TY]]*
+// CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+// CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* %
+// CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+// CHECK: store i8
+// CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+// CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 0
+// CHECK-NOT: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 1
+// CHECK: getelementptr inbounds [[SS_TY]], [[SS_TY]]* %{{.+}}, i32 0, i32 2
+// CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 4, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[SS_TY]]*, i{{[0-9]+}}*, i{{[0-9]+}}*, i{{[0-9]+}}*)* [[SS_MICROTASK:@.+]] to void
+// CHECK: [[B_REF:%.+]] = getelementptr {{.*}}[[SS_TY]], [[SS_TY]]* %{{.*}}, i32 0, i32 1
+// CHECK: store i8 %{{.+}}, i8* [[B_REF]],
+// CHECK: ret
+
+// CHECK: define internal void [[SS_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}}, [[SS_TY]]*
+// CHECK: [[A_PRIV:%.+]] = alloca i{{[0-9]+}},
+// CHECK: [[B_PRIV:%.+]] = alloca i{{[0-9]+}},
+// CHECK: [[C_PRIV:%.+]] = alloca i{{[0-9]+}},
+// CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[A_PRIV]],
+// CHECK: store i{{[0-9]+}}* [[A_PRIV]], i{{[0-9]+}}** [[REFA:%.+]],
+// CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[B_PRIV]],
+// CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[C_PRIV]],
+// CHECK: store i{{[0-9]+}}* [[C_PRIV]], i{{[0-9]+}}** [[REFC:%.+]],
+// CHECK: [[A_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFA]],
+// CHECK-NEXT: [[A_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[A_PRIV]],
+// CHECK-NEXT: [[INC:%.+]] = add nsw i{{[0-9]+}} [[A_VAL]], 1
+// CHECK-NEXT: store i{{[0-9]+}} [[INC]], i{{[0-9]+}}* [[A_PRIV]],
+// CHECK-NEXT: [[B_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[B_PRIV]],
+// CHECK-NEXT: [[DEC:%.+]] = add nsw i{{[0-9]+}} [[B_VAL]], -1
+// CHECK-NEXT: store i{{[0-9]+}} [[DEC]], i{{[0-9]+}}* [[B_PRIV]],
+// CHECK-NEXT: [[C_PRIV:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[REFC]],
+// CHECK-NEXT: [[C_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[C_PRIV]],
+// CHECK-NEXT: [[DIV:%.+]] = sdiv i{{[0-9]+}} [[C_VAL]], 1
+// CHECK-NEXT: store i{{[0-9]+}} [[DIV]], i{{[0-9]+}}* [[C_PRIV]],
+// CHECK: call i32 @__kmpc_reduce_nowait(
+// CHECK: ret void
+
// CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* noalias [[GTID_ADDR:%.+]], i{{[0-9]+}}* noalias %{{.+}},
// CHECK: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}}, align 128
// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]], align 128