This is an archive of the discontinued LLVM Phabricator instance.

Differential D9239

[OPENMP] Codegen for 'private' clause in 'sections' directive.
ClosedPublic

Authored by ABataev on Apr 23 2015, 10:11 PM.

Details

Reviewers
rjmccall
hfinkel
Commits
rG7387083d95d5: [OPENMP] Codegen for 'private' clause in 'sections' directive.
rC235833: [OPENMP] Codegen for 'private' clause in 'sections' directive.
rL235833: [OPENMP] Codegen for 'private' clause in 'sections' directive.
Summary

If there are 2 or more sections in a 'section' directive the following code is generated:

<default init for privates>
@__kmpc_for_static_init_4();
<BODY for sections directive>
@__kmpc_for_static_fini()

If there is only one section, the following code is generated:

if (@__kmpc_single()) {
<default init for privates>
@__kmpc_end_single();
}

Diff Detail

Repository
rL LLVM

Event Timeline

ABataev updated this revision to Diff 24354.Apr 23 2015, 10:11 PM
ABataev retitled this revision from to [OPENMP] Codegen for 'private' clause in 'sections' directive..
ABataev updated this object.
ABataev edited the test plan for this revision. (Show Details)
ABataev added reviewers: rjmccall, hfinkel.
ABataev added subscribers: • fraggamuffin, • ejstotzer, Unknown Object (MLST).
rjmccall edited edge metadata.Apr 24 2015, 10:30 AM

LGTM.

Closed by commit rL235833: [OPENMP] Codegen for 'private' clause in 'sections' directive. (authored by ABataev). · Explain WhyApr 26 2015, 9:15 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
cfe/
trunk/
lib/
CodeGen/
2 lines
test/
OpenMP/
192 lines

Diff 24451

cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp

Show First 20 Lines • Show All 1,166 Lines • ▼ Show 20 Lines auto &&CodeGen = [&S, CS](CodeGenFunction &CGF) {
CodeGenFunction::OMPPrivateScope LoopScope(CGF); CodeGenFunction::OMPPrivateScope LoopScope(CGF);
if (CGF.EmitOMPFirstprivateClause(S, LoopScope)) { if (CGF.EmitOMPFirstprivateClause(S, LoopScope)) {
// Emit implicit barrier to synchronize threads and avoid data races on // Emit implicit barrier to synchronize threads and avoid data races on
// initialization of firstprivate variables. // initialization of firstprivate variables.
CGF.CGM.getOpenMPRuntime().emitBarrierCall(CGF, S.getLocStart(), CGF.CGM.getOpenMPRuntime().emitBarrierCall(CGF, S.getLocStart(),
OMPD_unknown); OMPD_unknown);
} }
CGF.EmitOMPPrivateClause(S, LoopScope);
(void)LoopScope.Privatize(); (void)LoopScope.Privatize();
// Emit static non-chunked loop. // Emit static non-chunked loop.
CGF.CGM.getOpenMPRuntime().emitForInit( CGF.CGM.getOpenMPRuntime().emitForInit(
CGF, S.getLocStart(), OMPC_SCHEDULE_static, /*IVSize=*/32, CGF, S.getLocStart(), OMPC_SCHEDULE_static, /*IVSize=*/32,
/*IVSigned=*/true, IL.getAddress(), LB.getAddress(), UB.getAddress(), /*IVSigned=*/true, IL.getAddress(), LB.getAddress(), UB.getAddress(),
ST.getAddress()); ST.getAddress());
// UB = min(UB, GlobalUB); // UB = min(UB, GlobalUB);
Show All 14 Lines if (CS && CS->size() > 1) {
return OMPD_sections; return OMPD_sections;
} }
// If only one section is found - no need to generate loop, emit as a single // If only one section is found - no need to generate loop, emit as a single
// region. // region.
bool HasFirstprivates; bool HasFirstprivates;
auto &&CodeGen = [Stmt, &S, &HasFirstprivates](CodeGenFunction &CGF) { auto &&CodeGen = [Stmt, &S, &HasFirstprivates](CodeGenFunction &CGF) {
CodeGenFunction::OMPPrivateScope SingleScope(CGF); CodeGenFunction::OMPPrivateScope SingleScope(CGF);
HasFirstprivates = CGF.EmitOMPFirstprivateClause(S, SingleScope); HasFirstprivates = CGF.EmitOMPFirstprivateClause(S, SingleScope);
CGF.EmitOMPPrivateClause(S, SingleScope);
(void)SingleScope.Privatize(); (void)SingleScope.Privatize();
CGF.EmitStmt(Stmt); CGF.EmitStmt(Stmt);
CGF.EnsureInsertPoint(); CGF.EnsureInsertPoint();
}; };
CGF.CGM.getOpenMPRuntime().emitSingleRegion(CGF, CodeGen, S.getLocStart(), CGF.CGM.getOpenMPRuntime().emitSingleRegion(CGF, CodeGen, S.getLocStart(),
llvm::None, llvm::None, llvm::None, llvm::None,
llvm::None, llvm::None); llvm::None, llvm::None);
▲ Show 20 Lines • Show All 647 Lines • Show Last 20 Lines

cfe/trunk/test/OpenMP/sections_private_codegen.cpp

PropertyOld ValueNew Value
svn:eol-stylenullnative
svn:keywordsnullAuthor Date Id Rev URL
svn:mime-typenulltext/plain
// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -emit-llvm %s -o - | FileCheck %s
// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -std=c++11 -triple x86_64-unknown-unknown -emit-pch -o %t %s
// RUN: %clang_cc1 -fopenmp=libiomp5 -x c++ -triple x86_64-unknown-unknown -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 %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=LAMBDA %s
// RUN: %clang_cc1 -verify -fopenmp=libiomp5 -x c++ -fblocks -DBLOCKS -triple %itanium_abi_triple -emit-llvm %s -o - | FileCheck -check-prefix=BLOCKS %s
// expected-no-diagnostics
#ifndef HEADER
#define HEADER
template <class T>
struct S {
T f;
S(T a) : f(a) {}
S() : f() {}
operator T() { return T(); }
~S() {}
};
volatile double g;
// CHECK: [[S_FLOAT_TY:%.+]] = type { float }
// CHECK: [[CAP_MAIN_TY:%.+]] = type { i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]* }
// CHECK: [[S_INT_TY:%.+]] = type { i{{[0-9]+}} }
// CHECK: [[CAP_TMAIN_TY:%.+]] = type { i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]* }
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);
#pragma omp parallel
#pragma omp sections private(t_var, vec, s_arr, s_arr, var, var)
{
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{{( x86_thiscallcc)?}} 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 private(g)
{
// LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* [[ARG:%.+]])
// LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca double,
// LAMBDA: store %{{.+}}* [[ARG]], %{{.+}}** [[ARG_REF:%.+]],
g = 1;
// LAMBDA: call void @__kmpc_for_static_init_4(
// 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{{( x86_thiscallcc)?}} void [[INNER_LAMBDA:@.+]](%{{.+}}* [[ARG]])
// LAMBDA: call void @__kmpc_for_static_fini(
#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 private(g)
{
// BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* [[ARG:%.+]])
// BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca double,
// BLOCKS: store %{{.+}}* [[ARG]], %{{.+}}** [[ARG_REF:%.+]],
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(
#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<float> test;
int t_var = 0;
int vec[] = {1, 2};
S<float> s_arr[] = {1, 2};
S<float> var(3);
#pragma omp parallel
#pragma omp sections private(t_var, vec, s_arr, s_arr, var, var)
{
{
vec[0] = t_var;
s_arr[0] = var;
}
}
return tmain<int>();
#endif
}
// CHECK: define i{{[0-9]+}} @main()
// CHECK: [[TEST:%.+]] = alloca [[S_FLOAT_TY]],
// CHECK: call {{.*}} [[S_FLOAT_TY_DEF_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 void [[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: [[T_VAR_PRIV:%.+]] = alloca i{{[0-9]+}},
// CHECK: [[VEC_PRIV:%.+]] = alloca [2 x i{{[0-9]+}}],
// CHECK: [[S_ARR_PRIV:%.+]] = alloca [2 x [[S_FLOAT_TY]]],
// CHECK-NOT: alloca [2 x [[S_FLOAT_TY]]],
// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_FLOAT_TY]],
// CHECK-NOT: alloca [[S_FLOAT_TY]],
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
// CHECK: call i32 @__kmpc_single(
// CHECK-NOT: [[T_VAR_PRIV]]
// CHECK-NOT: [[VEC_PRIV]]
// CHECK: {{.+}}:
// CHECK: [[S_ARR_PRIV_ITEM:%.+]] = phi [[S_FLOAT_TY]]*
// CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR]]([[S_FLOAT_TY]]* [[S_ARR_PRIV_ITEM]])
// CHECK-NOT: [[T_VAR_PRIV]]
// CHECK-NOT: [[VEC_PRIV]]
// CHECK: call {{.*}} [[S_FLOAT_TY_DEF_CONSTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
// CHECK-DAG: call void [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
// CHECK-DAG: call void [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]*
// CHECK: call void @__kmpc_end_single(
// CHECK: ret void
// CHECK: define {{.*}} i{{[0-9]+}} [[TMAIN_INT]]()
// CHECK: [[TEST:%.+]] = alloca [[S_INT_TY]],
// CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR:@.+]]([[S_INT_TY]]* [[TEST]])
// 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 void [[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 i32,
// CHECK: alloca i32,
// CHECK: alloca i32,
// CHECK: alloca i32,
// CHECK: alloca i32,
// 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-NOT: alloca [2 x [[S_INT_TY]]],
// CHECK: [[VAR_PRIV:%.+]] = alloca [[S_INT_TY]],
// CHECK-NOT: alloca [[S_INT_TY]],
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
// CHECK-NOT: [[T_VAR_PRIV]]
// CHECK-NOT: [[VEC_PRIV]]
// CHECK: {{.+}}:
// CHECK: [[S_ARR_PRIV_ITEM:%.+]] = phi [[S_INT_TY]]*
// CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]])
// CHECK-NOT: [[T_VAR_PRIV]]
// CHECK-NOT: [[VEC_PRIV]]
// CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR]]([[S_INT_TY]]* [[VAR_PRIV]])
// CHECK: call void @__kmpc_for_static_init_4(
// CHECK: call void @__kmpc_for_static_fini(
// CHECK-DAG: call void [[S_INT_TY_DESTR]]([[S_INT_TY]]* [[VAR_PRIV]])
// CHECK-DAG: call void [[S_INT_TY_DESTR]]([[S_INT_TY]]*
// CHECK: ret void
#endif