From a89adf22db0db5aaa561b20a5ae1845190c851a7 Mon Sep 17 00:00:00 2001 From: Alexey Bataev Date: Mon, 27 Apr 2015 05:04:13 +0000 Subject: [PATCH] [OPENMP] Codegen for 'reduction' clause in 'sections' directive. Emit a code for reduction clause. Next code should be emitted for reductions: static kmp_critical_name lock = { 0 }; void reduce_func(void *lhs[], void *rhs[]) { *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); ... *(Type-1*)lhs[-1] = ReductionOperation-1(*(Type-1*)lhs[-1], *(Type-1*)rhs[-1]); } ... void *RedList[] = {&[0], ..., &[-1]}; switch (__kmpc_reduce{_nowait}(, , , sizeof(RedList), RedList, reduce_func, &)) { case 1: [0] = ReductionOperation0(*[0], *[0]); ... [-1] = ReductionOperation-1(*[-1], *[-1]); __kmpc_end_reduce{_nowait}(, , &); break; case 2: Atomic([0] = ReductionOperation0(*[0], *[0])); ... Atomic([-1] = ReductionOperation-1(*[-1], *[-1])); break; default:; } Reduction variables are a kind of a private variables, they have private copies, but initial values are chosen in accordance with the reduction operation. If sections directive has only single section, then original shared variables are used instead with barrier at the end of the directive. Differential Revision: http://reviews.llvm.org/D9242 llvm-svn: 235835 --- clang/lib/CodeGen/CGStmtOpenMP.cpp | 23 +- .../OpenMP/sections_reduction_codegen.cpp | 472 ++++++++++++++++++ 2 files changed, 489 insertions(+), 6 deletions(-) create mode 100644 clang/test/OpenMP/sections_reduction_codegen.cpp diff --git a/clang/lib/CodeGen/CGStmtOpenMP.cpp b/clang/lib/CodeGen/CGStmtOpenMP.cpp index 62fe4bcd1f24..22f6c3432a0e 100644 --- a/clang/lib/CodeGen/CGStmtOpenMP.cpp +++ b/clang/lib/CodeGen/CGStmtOpenMP.cpp @@ -1175,6 +1175,7 @@ static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, } CGF.EmitOMPPrivateClause(S, LoopScope); HasLastprivates = CGF.EmitOMPLastprivateClauseInit(S, LoopScope); + CGF.EmitOMPReductionClauseInit(S, LoopScope); (void)LoopScope.Privatize(); // Emit static non-chunked loop. @@ -1194,6 +1195,7 @@ static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, [](CodeGenFunction &) {}); // Tell the runtime we are done. CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart()); + CGF.EmitOMPReductionClauseFinal(S); // Emit final copy of the lastprivate variables if IsLastIter != 0. if (HasLastprivates) @@ -1217,6 +1219,14 @@ static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, // If only one section is found - no need to generate loop, emit as a single // region. bool HasFirstprivates; + // No need to generate reductions for sections with single section region, we + // can use original shared variables for all operations. + auto ReductionFilter = [](const OMPClause *C) -> bool { + return C->getClauseKind() == OMPC_reduction; + }; + OMPExecutableDirective::filtered_clause_iterator I( + S.clauses(), ReductionFilter); + bool HasReductions = I; // No need to generate lastprivates for sections with single section region, // we can use original shared variable for all calculations with barrier at // the end of the sections. @@ -1224,8 +1234,8 @@ static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, return C->getClauseKind() == OMPC_lastprivate; }; OMPExecutableDirective::filtered_clause_iterator - I(S.clauses(), LastprivateFilter); - bool HasLastprivates = I; + I1(S.clauses(), LastprivateFilter); + bool HasLastprivates = I1; auto &&CodeGen = [Stmt, &S, &HasFirstprivates](CodeGenFunction &CGF) { CodeGenFunction::OMPPrivateScope SingleScope(CGF); HasFirstprivates = CGF.EmitOMPFirstprivateClause(S, SingleScope); @@ -1238,10 +1248,11 @@ static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, CGF.CGM.getOpenMPRuntime().emitSingleRegion(CGF, CodeGen, S.getLocStart(), llvm::None, llvm::None, llvm::None, llvm::None); - // Emit barrier for firstprivates or lastprivates only if 'sections' directive - // has 'nowait' clause. Otherwise the barrier will be generated by the codegen - // for the directive. - if ((HasFirstprivates || HasLastprivates) && S.getSingleClause(OMPC_nowait)) { + // Emit barrier for firstprivates, lastprivates or reductions only if + // 'sections' directive has 'nowait' clause. Otherwise the barrier will be + // generated by the codegen for the directive. + if ((HasFirstprivates || HasLastprivates || HasReductions) && + S.getSingleClause(OMPC_nowait)) { // Emit implicit barrier to synchronize threads and avoid data races on // initialization of firstprivate variables. CGF.CGM.getOpenMPRuntime().emitBarrierCall(CGF, S.getLocStart(), diff --git a/clang/test/OpenMP/sections_reduction_codegen.cpp b/clang/test/OpenMP/sections_reduction_codegen.cpp new file mode 100644 index 000000000000..4623f54d556c --- /dev/null +++ b/clang/test/OpenMP/sections_reduction_codegen.cpp @@ -0,0 +1,472 @@ +// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck %s +// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-apple-darwin10 -emit-pch -o %t %s +// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-apple-darwin10 -std=c++11 -include-pch %t -verify %s -emit-llvm -o - | FileCheck %s +// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -std=c++11 -DLAMBDA -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s +// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -fblocks -DBLOCKS -triple x86_64-apple-darwin10 -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s +// expected-no-diagnostics +#ifndef HEADER +#define HEADER + +volatile double g; + +template +struct S { + T f; + S(T a) : f(a + g) {} + S() : f(g) {} + operator T() { return T(); } + S &operator&(const S &) { return *this; } + ~S() {} +}; + +// CHECK-DAG: [[S_FLOAT_TY:%.+]] = type { float } +// CHECK-DAG: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} } +// CHECK-DAG: [[CAP_MAIN_TY:%.+]] = type { float*, [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]*, float*, [2 x i{{[0-9]+}}]*, [2 x [[S_FLOAT_TY]]]* } +// CHECK-DAG: [[CAP_TMAIN_TY:%.+]] = type { i{{[0-9]+}}*, [[S_INT_TY]]*, [[S_INT_TY]]*, i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_INT_TY]]]* } +// CHECK-DAG: [[ATOMIC_REDUCE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* +// CHECK-DAG: [[IMPLICIT_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 66, i32 0, i32 0, i8* +// CHECK-DAG: [[SINGLE_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 322, i32 0, i32 0, i8* +// CHECK-DAG: [[REDUCTION_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 18, i32 0, i32 0, i8* +// CHECK-DAG: [[REDUCTION_LOCK:@.+]] = common global [8 x i32] zeroinitializer + +template +T tmain() { + T t; + S test; + T t_var = T(), t_var1; + T vec[] = {1, 2}; + S s_arr[] = {1, 2}; + S var(3), var1; +#pragma omp parallel +#pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) nowait + { + vec[0] = t_var; +#pragma omp section + s_arr[0] = var; + } + return T(); +} + +int main() { +#ifdef LAMBDA + // LAMBDA: [[G:@.+]] = global double + // LAMBDA-LABEL: @main + // LAMBDA: call void [[OUTER_LAMBDA:@.+]]( + [&]() { + // LAMBDA: define{{.*}} internal{{.*}} void [[OUTER_LAMBDA]]( + // LAMBDA: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i8* %{{.+}}) +#pragma omp parallel +#pragma omp sections reduction(+:g) + { + // LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* %{{.+}}) + // LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double, + + // Reduction list for runtime. + // LAMBDA: [[RED_LIST:%.+]] = alloca [1 x i8*], + + // LAMBDA: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] + // LAMBDA: call void @__kmpc_for_static_init_4( + g = 1; + // LAMBDA: store volatile double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], + // LAMBDA: [[G_PRIVATE_ADDR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG:%.+]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 + // LAMBDA: store double* [[G_PRIVATE_ADDR]], double** [[G_PRIVATE_ADDR_REF]] + // LAMBDA: call void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]]) + // LAMBDA: call void @__kmpc_for_static_fini( + + // LAMBDA: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i32 0, i32 0 + // LAMBDA: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* + // LAMBDA: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], + // LAMBDA: call i32 @__kmpc_reduce( + // LAMBDA: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ + // LAMBDA: i32 1, label %[[CASE1:.+]] + // LAMBDA: i32 2, label %[[CASE2:.+]] + // LAMBDA: [[CASE1]] + // LAMBDA: [[G_VAL:%.+]] = load double, double* [[G]] + // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] + // LAMBDA: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] + // LAMBDA: store double [[ADD]], double* [[G]] + // LAMBDA: call void @__kmpc_end_reduce( + // LAMBDA: br label %[[REDUCTION_DONE]] + // LAMBDA: [[CASE2]] + // LAMBDA: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] + // LAMBDA: fadd double + // LAMBDA: cmpxchg i64* + // LAMBDA: br label %[[REDUCTION_DONE]] + // LAMBDA: [[REDUCTION_DONE]] + // LAMBDA: ret void +#pragma omp section + [&]() { + // LAMBDA: define {{.+}} void [[INNER_LAMBDA]](%{{.+}}* [[ARG_PTR:%.+]]) + // LAMBDA: store %{{.+}}* [[ARG_PTR]], %{{.+}}** [[ARG_PTR_REF:%.+]], + g = 2; + // LAMBDA: [[ARG_PTR:%.+]] = load %{{.+}}*, %{{.+}}** [[ARG_PTR_REF]] + // LAMBDA: [[G_PTR_REF:%.+]] = getelementptr inbounds %{{.+}}, %{{.+}}* [[ARG_PTR]], i{{[0-9]+}} 0, i{{[0-9]+}} 0 + // LAMBDA: [[G_REF:%.+]] = load double*, double** [[G_PTR_REF]] + // LAMBDA: store volatile double 2.0{{.+}}, double* [[G_REF]] + }(); + } + }(); + return 0; +#elif defined(BLOCKS) + // BLOCKS: [[G:@.+]] = global double + // BLOCKS-LABEL: @main + // BLOCKS: call void {{%.+}}(i8 + ^{ + // BLOCKS: define{{.*}} internal{{.*}} void {{.+}}(i8* + // BLOCKS: call void {{.+}} @__kmpc_fork_call({{.+}}, i32 1, {{.+}}* [[OMP_REGION:@.+]] to {{.+}}, i8* %{{.+}}) +#pragma omp parallel +#pragma omp sections reduction(-:g) + { + // BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* %{{.+}}) + // BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double, + + // Reduction list for runtime. + // BLOCKS: [[RED_LIST:%.+]] = alloca [1 x i8*], + + // BLOCKS: store double 0.0{{.+}}, double* [[G_PRIVATE_ADDR]] + g = 1; + // BLOCKS: call void @__kmpc_for_static_init_4( + // BLOCKS: store volatile double 1.0{{.+}}, double* [[G_PRIVATE_ADDR]], + // BLOCKS-NOT: [[G]]{{[[^:word:]]}} + // BLOCKS: double* [[G_PRIVATE_ADDR]] + // BLOCKS-NOT: [[G]]{{[[^:word:]]}} + // BLOCKS: call void {{%.+}}(i8 + // BLOCKS: call void @__kmpc_for_static_fini( + + // BLOCKS: [[G_PRIV_REF:%.+]] = getelementptr inbounds [1 x i8*], [1 x i8*]* [[RED_LIST]], i32 0, i32 0 + // BLOCKS: [[BITCAST:%.+]] = bitcast double* [[G_PRIVATE_ADDR]] to i8* + // BLOCKS: store i8* [[BITCAST]], i8** [[G_PRIV_REF]], + // BLOCKS: call i32 @__kmpc_reduce( + // BLOCKS: switch i32 %{{.+}}, label %[[REDUCTION_DONE:.+]] [ + // BLOCKS: i32 1, label %[[CASE1:.+]] + // BLOCKS: i32 2, label %[[CASE2:.+]] + // BLOCKS: [[CASE1]] + // BLOCKS: [[G_VAL:%.+]] = load double, double* [[G]] + // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] + // BLOCKS: [[ADD:%.+]] = fadd double [[G_VAL]], [[G_PRIV_VAL]] + // BLOCKS: store double [[ADD]], double* [[G]] + // BLOCKS: call void @__kmpc_end_reduce( + // BLOCKS: br label %[[REDUCTION_DONE]] + // BLOCKS: [[CASE2]] + // BLOCKS: [[G_PRIV_VAL:%.+]] = load double, double* [[G_PRIVATE_ADDR]] + // BLOCKS: fadd double + // BLOCKS: cmpxchg i64* + // BLOCKS: br label %[[REDUCTION_DONE]] + // BLOCKS: [[REDUCTION_DONE]] + // BLOCKS: ret void +#pragma omp section + ^{ + // BLOCKS: define {{.+}} void {{@.+}}(i8* + g = 2; + // BLOCKS-NOT: [[G]]{{[[^:word:]]}} + // BLOCKS: store volatile double 2.0{{.+}}, double* + // BLOCKS-NOT: [[G]]{{[[^:word:]]}} + // BLOCKS: ret + }(); + } + }(); + return 0; +#else + S test; + float t_var = 0, t_var1; + int vec[] = {1, 2}; + S s_arr[] = {1, 2}; + S var(3), var1; +#pragma omp parallel +#pragma omp sections reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1) + { + { + vec[0] = t_var; + s_arr[0] = var; + vec[1] = t_var1; + s_arr[1] = var1; + } + } + return tmain(); +#endif +} + +// CHECK: define {{.*}}i{{[0-9]+}} @main() +// CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]], +// CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]]) +// CHECK: %{{.+}} = bitcast [[CAP_MAIN_TY]]* +// CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[CAP_MAIN_TY]]*)* [[MAIN_MICROTASK:@.+]] to void +// CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]() +// CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]* +// CHECK: ret +// +// CHECK: define internal void [[MAIN_MICROTASK]](i{{[0-9]+}}* [[GTID_ADDR:%.+]], i{{[0-9]+}}* %{{.+}}, [[CAP_MAIN_TY]]* %{{.+}}) +// CHECK-NOT: alloca float, +// CHECK-NOT: alloca [[S_FLOAT_TY]], +// CHECK-NOT: alloca [[S_FLOAT_TY]], +// CHECK-NOT: alloca float, + +// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], + +// CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] +// CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] +// CHECK: call i32 @__kmpc_single( + +// CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0 +// CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 1 +// CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 2 +// CHECK-DAG: getelementptr inbounds [[CAP_MAIN_TY]], [[CAP_MAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 3 + +// CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]]) +// CHECK-NOT: call {{.*}} [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* + +// CHECK: call void @__kmpc_end_single( + +// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[SINGLE_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) +// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) + +// CHECK: ret void + +// CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]() +// CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]], +// CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]]) +// CHECK: %{{.+}} = bitcast [[CAP_TMAIN_TY]]* +// CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 1, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [[CAP_TMAIN_TY]]*)* [[TMAIN_MICROTASK:@.+]] to void +// CHECK: call {{.*}} [[S_INT_TY_DESTR:@.+]]([[S_INT_TY]]* +// CHECK: ret +// +// CHECK: define internal void [[TMAIN_MICROTASK]](i{{[0-9]+}}* [[GTID_ADDR:%.+]], i{{[0-9]+}}* %{{.+}}, [[CAP_TMAIN_TY]]* %{{.+}}) +// CHECK: alloca i{{[0-9]+}}, +// CHECK: alloca i{{[0-9]+}}, +// CHECK: alloca i{{[0-9]+}}, +// CHECK: alloca i{{[0-9]+}}, +// CHECK: alloca i{{[0-9]+}}, +// 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]+}}, + +// Reduction list for runtime. +// CHECK: [[RED_LIST:%.+]] = alloca [4 x i8*], + +// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_ADDR:%.+]], + +// CHECK: [[T_VAR_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} +// CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[T_VAR_PTR_REF]], +// For + reduction operation initial value of private variable is 0. +// CHECK: store i{{[0-9]+}} 0, i{{[0-9]+}}* [[T_VAR_PRIV]], + +// CHECK: [[VAR_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} +// CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[VAR_PTR_REF:%.+]], +// For & reduction operation initial value of private variable is ones in all bits. +// CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR_PRIV]]) + +// CHECK: [[VAR1_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} +// CHECK: [[VAR1_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[VAR_PTR_REF:%.+]], +// For && reduction operation initial value of private variable is 1.0. +// CHECK: call {{.*}} [[S_INT_TY_CONSTR:@.+]]([[S_INT_TY]]* [[VAR1_PRIV]]) + +// CHECK: [[T_VAR1_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} {{[0-9]+}} +// CHECK: [[T_VAR1_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[T_VAR1_PTR_REF]], +// For min reduction operation initial value of private variable is largest repesentable value. +// CHECK: store i{{[0-9]+}} 2147483647, i{{[0-9]+}}* [[T_VAR1_PRIV]], + +// CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_ADDR]] +// CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]] +// CHECK: call void @__kmpc_for_static_init_4( +// Skip checks for internal operations. +// CHECK: call void @__kmpc_for_static_fini( + +// void *RedList[] = {[0], ..., [-1]}; + +// CHECK: [[T_VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 0 +// CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR_PRIV]] to i8* +// CHECK: store i8* [[BITCAST]], i8** [[T_VAR_PRIV_REF]], +// CHECK: [[VAR_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 1 +// CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR_PRIV]] to i8* +// CHECK: store i8* [[BITCAST]], i8** [[VAR_PRIV_REF]], +// CHECK: [[VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 2 +// CHECK: [[BITCAST:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_PRIV]] to i8* +// CHECK: store i8* [[BITCAST]], i8** [[VAR1_PRIV_REF]], +// CHECK: [[T_VAR1_PRIV_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST]], i32 0, i32 3 +// CHECK: [[BITCAST:%.+]] = bitcast i{{[0-9]+}}* [[T_VAR1_PRIV]] to i8* +// CHECK: store i8* [[BITCAST]], i8** [[T_VAR1_PRIV_REF]], + +// res = __kmpc_reduce_nowait(, , , sizeof(RedList), RedList, reduce_func, &); + +// CHECK: [[BITCAST:%.+]] = bitcast [4 x i8*]* [[RED_LIST]] to i8* +// CHECK: [[RES:%.+]] = call i32 @__kmpc_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], i32 4, i64 32, i8* [[BITCAST]], void (i8*, i8*)* [[REDUCTION_FUNC:@.+]], [8 x i32]* [[REDUCTION_LOCK]]) + +// switch(res) +// CHECK: switch i32 [[RES]], label %[[RED_DONE:.+]] [ +// CHECK: i32 1, label %[[CASE1:.+]] +// CHECK: i32 2, label %[[CASE2:.+]] +// CHECK: ] + +// case 1: +// t_var += t_var_reduction; +// CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_REF]], +// CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]], +// CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_VAL]], [[T_VAR_PRIV_VAL]] +// CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_REF]], + +// var = var.operator &(var_reduction); +// CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) + +// var1 = var1.operator &&(var1_reduction); +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) +// CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] +// CHECK: [[TRUE]] +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) +// CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] +// CHECK: [[TRUE2]] +// CHECK: br label %[[END2:.+]] +// CHECK: [[FALSE2]] +// CHECK: br label %[[END2]] +// CHECK: [[END2]] +// CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_REF]], %[[TRUE2]] ], [ [[VAR1_PRIV]], %[[FALSE2]] ] +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) + +// t_var1 = min(t_var1, t_var1_reduction); +// CHECK: [[T_VAR1_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_REF]], +// CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]], +// CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_VAL]], [[T_VAR1_PRIV_VAL]] +// CHECK: [[UP:%.+]] = zext i1 [[CMP]] to i{{[0-9]+}} +// CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_REF]], + +// __kmpc_end_reduce_nowait(, , &); +// CHECK: call void @__kmpc_end_reduce_nowait(%{{.+}}* [[REDUCTION_LOC]], i32 [[GTID]], [8 x i32]* [[REDUCTION_LOCK]]) + +// break; +// CHECK: br label %[[RED_DONE]] + +// case 2: +// t_var += t_var_reduction; +// CHECK: [[T_VAR_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]] +// CHECK: atomicrmw add i32* [[T_VAR_REF]], i32 [[T_VAR_PRIV_VAL]] monotonic + +// var = var.operator &(var_reduction); +// CHECK: call void @__kmpc_critical( +// CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* dereferenceable(4) [[VAR_PRIV]]) +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_REF]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) +// CHECK: call void @__kmpc_end_critical( + +// var1 = var1.operator &&(var1_reduction); +// CHECK: call void @__kmpc_critical( +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_REF]]) +// CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] +// CHECK: [[TRUE]] +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_PRIV]]) +// CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] +// CHECK: [[TRUE2]] +// CHECK: br label %[[END2:.+]] +// CHECK: [[FALSE2]] +// CHECK: br label %[[END2]] +// CHECK: [[END2]] +// CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_REF]], %[[TRUE2]] ], [ [[VAR1_PRIV]], %[[FALSE2]] ] +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_REF]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) +// CHECK: call void @__kmpc_end_critical( + +// t_var1 = min(t_var1, t_var1_reduction); +// CHECK: [[T_VAR1_PRIV_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_PRIV]] +// CHECK: atomicrmw min i32* [[T_VAR1_REF]], i32 [[T_VAR1_PRIV_VAL]] monotonic + +// break; +// CHECK: br label %[[RED_DONE]] +// CHECK: [[RED_DONE]] +// CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]]) +// CHECK-DAG: call {{.*}} [[S_INT_TY_DESTR]]([[S_INT_TY]]* +// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]]) +// CHECK: ret void + +// void reduce_func(void *lhs[], void *rhs[]) { +// *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]); +// ... +// *(Type-1*)lhs[-1] = ReductionOperation-1(*(Type-1*)lhs[-1], +// *(Type-1*)rhs[-1]); +// } +// CHECK: define internal void [[REDUCTION_FUNC]](i8*, i8*) +// t_var_lhs = (i{{[0-9]+}}*)lhs[0]; +// CHECK: [[T_VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS:%.+]], i32 0, i32 0 +// CHECK: [[T_VAR_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR_RHS_REF]], +// CHECK: [[T_VAR_RHS:%.+]] = bitcast i8* [[T_VAR_RHS_VOID]] to i{{[0-9]+}}* +// t_var_rhs = (i{{[0-9]+}}*)rhs[0]; +// CHECK: [[T_VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS:%.+]], i32 0, i32 0 +// CHECK: [[T_VAR_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR_LHS_REF]], +// CHECK: [[T_VAR_LHS:%.+]] = bitcast i8* [[T_VAR_LHS_VOID]] to i{{[0-9]+}}* + +// var_lhs = (S*)lhs[1]; +// CHECK: [[VAR_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 1 +// CHECK: [[VAR_RHS_VOID:%.+]] = load i8*, i8** [[VAR_RHS_REF]], +// CHECK: [[VAR_RHS:%.+]] = bitcast i8* [[VAR_RHS_VOID]] to [[S_INT_TY]]* +// var_rhs = (S*)rhs[1]; +// CHECK: [[VAR_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 1 +// CHECK: [[VAR_LHS_VOID:%.+]] = load i8*, i8** [[VAR_LHS_REF]], +// CHECK: [[VAR_LHS:%.+]] = bitcast i8* [[VAR_LHS_VOID]] to [[S_INT_TY]]* + +// var1_lhs = (S*)lhs[2]; +// CHECK: [[VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 2 +// CHECK: [[VAR1_RHS_VOID:%.+]] = load i8*, i8** [[VAR1_RHS_REF]], +// CHECK: [[VAR1_RHS:%.+]] = bitcast i8* [[VAR1_RHS_VOID]] to [[S_INT_TY]]* +// var1_rhs = (S*)rhs[2]; +// CHECK: [[VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 2 +// CHECK: [[VAR1_LHS_VOID:%.+]] = load i8*, i8** [[VAR1_LHS_REF]], +// CHECK: [[VAR1_LHS:%.+]] = bitcast i8* [[VAR1_LHS_VOID]] to [[S_INT_TY]]* + +// t_var1_lhs = (i{{[0-9]+}}*)lhs[3]; +// CHECK: [[T_VAR1_RHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_RHS]], i32 0, i32 3 +// CHECK: [[T_VAR1_RHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_RHS_REF]], +// CHECK: [[T_VAR1_RHS:%.+]] = bitcast i8* [[T_VAR1_RHS_VOID]] to i{{[0-9]+}}* +// t_var1_rhs = (i{{[0-9]+}}*)rhs[3]; +// CHECK: [[T_VAR1_LHS_REF:%.+]] = getelementptr inbounds [4 x i8*], [4 x i8*]* [[RED_LIST_LHS]], i32 0, i32 3 +// CHECK: [[T_VAR1_LHS_VOID:%.+]] = load i8*, i8** [[T_VAR1_LHS_REF]], +// CHECK: [[T_VAR1_LHS:%.+]] = bitcast i8* [[T_VAR1_LHS_VOID]] to i{{[0-9]+}}* + +// t_var_lhs += t_var_rhs; +// CHECK: [[T_VAR_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_LHS]], +// CHECK: [[T_VAR_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_RHS]], +// CHECK: [[UP:%.+]] = add nsw i{{[0-9]+}} [[T_VAR_LHS_VAL]], [[T_VAR_RHS_VAL]] +// CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR_LHS]], + +// var_lhs = var_lhs.operator &(var_rhs); +// CHECK: [[UP:%.+]] = call dereferenceable(4) [[S_INT_TY]]* @{{.+}}([[S_INT_TY]]* [[VAR_LHS]], [[S_INT_TY]]* dereferenceable(4) [[VAR_RHS]]) +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR_LHS]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[UP]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) + +// var1_lhs = var1_lhs.operator &&(var1_rhs); +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_LHS]]) +// CHECK: [[VAR1_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_BOOL]], label %[[TRUE:.+]], label %[[FALSE:.+]] +// CHECK: [[TRUE]] +// CHECK: [[TO_INT:%.+]] = call i{{[0-9]+}} @{{.+}}([[S_INT_TY]]* [[VAR1_RHS]]) +// CHECK: [[VAR1_REDUCTION_BOOL:%.+]] = icmp ne i{{[0-9]+}} [[TO_INT]], 0 +// CHECK: br i1 [[VAR1_REDUCTION_BOOL]], label %[[TRUE2:.+]], label %[[FALSE2:.+]] +// CHECK: [[TRUE2]] +// CHECK: br label %[[END2:.+]] +// CHECK: [[FALSE2]] +// CHECK: br label %[[END2]] +// CHECK: [[END2]] +// CHECK: [[COND_LVALUE:%.+]] = phi [[S_INT_TY]]* [ [[VAR1_LHS]], %[[TRUE2]] ], [ [[VAR1_RHS]], %[[FALSE2]] ] +// CHECK: [[BC1:%.+]] = bitcast [[S_INT_TY]]* [[VAR1_LHS]] to i8* +// CHECK: [[BC2:%.+]] = bitcast [[S_INT_TY]]* [[COND_LVALUE]] to i8* +// CHECK: call void @llvm.memcpy.p0i8.p0i8.i64(i8* [[BC1]], i8* [[BC2]], i64 4, i32 4, i1 false) + +// t_var1_lhs = min(t_var1_lhs, t_var1_rhs); +// CHECK: [[T_VAR1_LHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_LHS]], +// CHECK: [[T_VAR1_RHS_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR1_RHS]], +// CHECK: [[CMP:%.+]] = icmp slt i{{[0-9]+}} [[T_VAR1_LHS_VAL]], [[T_VAR1_RHS_VAL]] +// CHECK: [[UP:%.+]] = zext i1 [[CMP]] to i{{[0-9]+}} +// CHECK: store i{{[0-9]+}} [[UP]], i{{[0-9]+}}* [[T_VAR1_LHS]], +// CHECK: ret void + +#endif +