This is an archive of the discontinued LLVM Phabricator instance.

Differential D9240

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

Authored by ABataev on Apr 23 2015, 11:01 PM.

Details

Reviewers
rjmccall
hfinkel
Commits
rG9efc03b6f766: [OPENMP] Codegen for 'lastprivate' clause in 'sections' directive.
rC235834: [OPENMP] Codegen for 'lastprivate' clause in 'sections' directive.
rL235834: [OPENMP] Codegen for 'lastprivate' clause in 'sections' directive.
Summary
#pragma omp sections lastprivate(<var>)
<BODY>;

This construct is translated into something like:

<last_iter> = alloca i32
<init for lastprivates>;
<last_iter> = 0
; No initializer for simple variables or a default constructor is called for objects.
; For arrays perform element by element initialization by the call of the default constructor.
...
OMP_FOR_START(...,<last_iter>, ..); sets <last_iter> to 1 if this is the last iteration.
<BODY>
...
OMP_FOR_END
if (<last_iter> != 0) {
<final copy for lastprivate>; Update original variable with the lastprivate value.
}
call __kmpc_cancel_barrier() ; an implicit barrier to avoid possible data race.

If there is only one section, there is no special code generation, original shared variables are used + barrier is emitted at the end of the directive.

Diff Detail

Repository
rL LLVM

Event Timeline

ABataev updated this revision to Diff 24355.Apr 23 2015, 11:01 PM
ABataev retitled this revision from to [OPENMP] Codegen for 'lastprivate' 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).
ABataev updated this revision to Diff 24356.Apr 23 2015, 11:26 PM

Small fix in barrier generation

ABataev updated this revision to Diff 24357.Apr 24 2015, 12:11 AM
ABataev updated this object.

Fixed codegen for sections regions with single section.

rjmccall edited edge metadata.Apr 24 2015, 10:48 AM

LGTM, but I do have a general request that you clean up iterating over clauses. You can do that as a separate patch, but when you do, please go over all the existing uses of filtered_clause_iterator and try to take advantage of it.

lib/CodeGen/CGStmtOpenMP.cpp
1227 ↗(On Diff #24357)

You construct iterators like this so often — and almost always with a single clause kind as the filter — that I think you should probably just add specific API for it on OMPExecutableDirective. Something like

template <Fn> filtered_clause_iterator<Fn> getFilteredClauses(Fn &&fn) const {
  return filtered_clause_iterator<Fn>(clauses(), std::move(fn));
}

struct ClauseKindFilter {
  OpenMPClauseKind Kind;
  bool operator()(const OMPClause *clause) const { return clause->getClauseKind() == Kind; }
};
filtered_clause_iterator<ClauseKindFilter> getClausesOfKind(OpenMPClauseKind kind) const {
  return getFilteredClauses(ClauseKindFilter{ kind });
}

Also, the operator bool on filtered_clause_iterator should be explicit.

The upshot is that this entire block should turn into:

bool HasLastprivates = !S.getClausesOfKind(OMPC_lastprivate).empty();
ABataev added inline comments.Apr 26 2015, 9:18 PM
lib/CodeGen/CGStmtOpenMP.cpp
1227 ↗(On Diff #24357)

John, I already thought about it. I will commit a fix for it later today.

Closed by commit rL235834: [OPENMP] Codegen for 'lastprivate' clause in 'sections' directive. (authored by ABataev). · Explain WhyApr 26 2015, 9:37 PM
This revision was automatically updated to reflect the committed changes.

Revision Contents

PathSize
cfe/
trunk/
lib/
CodeGen/
36 lines
test/
OpenMP/
332 lines

Diff 24452

cfe/trunk/lib/CodeGen/CGStmtOpenMP.cpp

Show First 20 Lines • Show All 1,106 Lines • ▼ Show 20 Linesstatic LValue createSectionLVal(CodeGenFunction &CGF, QualType Ty,
return LVal; return LVal;
} }
static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF, static OpenMPDirectiveKind emitSections(CodeGenFunction &CGF,
const OMPExecutableDirective &S) { const OMPExecutableDirective &S) {
auto *Stmt = cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt(); auto *Stmt = cast<CapturedStmt>(S.getAssociatedStmt())->getCapturedStmt();
auto *CS = dyn_cast<CompoundStmt>(Stmt); auto *CS = dyn_cast<CompoundStmt>(Stmt);
if (CS && CS->size() > 1) { if (CS && CS->size() > 1) {
auto &&CodeGen = [&S, CS](CodeGenFunction &CGF) { bool HasLastprivates = false;
auto &&CodeGen = [&S, CS, &HasLastprivates](CodeGenFunction &CGF) {
auto &C = CGF.CGM.getContext(); auto &C = CGF.CGM.getContext();
auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1); auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
// Emit helper vars inits. // Emit helper vars inits.
LValue LB = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.lb.", LValue LB = createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.lb.",
CGF.Builder.getInt32(0)); CGF.Builder.getInt32(0));
auto *GlobalUBVal = CGF.Builder.getInt32(CS->size() - 1); auto *GlobalUBVal = CGF.Builder.getInt32(CS->size() - 1);
LValue UB = LValue UB =
createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.ub.", GlobalUBVal); createSectionLVal(CGF, KmpInt32Ty, ".omp.sections.ub.", GlobalUBVal);
▲ Show 20 Lines • Show All 44 Lines • ▼ Show 20 Lines auto &&CodeGen = [&S, CS, &HasLastprivates](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); CGF.EmitOMPPrivateClause(S, LoopScope);
HasLastprivates = CGF.EmitOMPLastprivateClauseInit(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);
auto *UBVal = CGF.EmitLoadOfScalar(UB, S.getLocStart()); auto *UBVal = CGF.EmitLoadOfScalar(UB, S.getLocStart());
auto *MinUBGlobalUB = CGF.Builder.CreateSelect( auto *MinUBGlobalUB = CGF.Builder.CreateSelect(
CGF.Builder.CreateICmpSLT(UBVal, GlobalUBVal), UBVal, GlobalUBVal); CGF.Builder.CreateICmpSLT(UBVal, GlobalUBVal), UBVal, GlobalUBVal);
CGF.EmitStoreOfScalar(MinUBGlobalUB, UB); CGF.EmitStoreOfScalar(MinUBGlobalUB, UB);
// IV = LB; // IV = LB;
CGF.EmitStoreOfScalar(CGF.EmitLoadOfScalar(LB, S.getLocStart()), IV); CGF.EmitStoreOfScalar(CGF.EmitLoadOfScalar(LB, S.getLocStart()), IV);
// while (idx <= UB) { BODY; ++idx; } // while (idx <= UB) { BODY; ++idx; }
CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen, CGF.EmitOMPInnerLoop(S, /*RequiresCleanup=*/false, &Cond, &Inc, BodyGen,
[](CodeGenFunction &) {}); [](CodeGenFunction &) {});
// Tell the runtime we are done. // Tell the runtime we are done.
CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart()); CGF.CGM.getOpenMPRuntime().emitForStaticFinish(CGF, S.getLocStart());
// Emit final copy of the lastprivate variables if IsLastIter != 0.
if (HasLastprivates)
CGF.EmitOMPLastprivateClauseFinal(
S, CGF.Builder.CreateIsNotNull(
CGF.EmitLoadOfScalar(IL, S.getLocStart())));
}; };
CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, CodeGen); CGF.CGM.getOpenMPRuntime().emitInlinedDirective(CGF, CodeGen);
// Emit barrier for lastprivates only if 'sections' directive has 'nowait'
// clause. Otherwise the barrier will be generated by the codegen for the
// directive.
if (HasLastprivates && 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(),
OMPD_unknown);
}
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;
// 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.
auto LastprivateFilter = [](const OMPClause *C) -> bool {
return C->getClauseKind() == OMPC_lastprivate;
};
OMPExecutableDirective::filtered_clause_iterator<decltype(LastprivateFilter)>
I(S.clauses(), LastprivateFilter);
bool HasLastprivates = I;
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); 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);
// Emit barrier for firstprivates only if 'sections' directive has 'nowait' // Emit barrier for firstprivates or lastprivates only if 'sections' directive
// clause. Otherwise the barrier will be generated by the codegen for the // has 'nowait' clause. Otherwise the barrier will be generated by the codegen
// directive. // for the directive.
if (HasFirstprivates && S.getSingleClause(OMPC_nowait)) { if ((HasFirstprivates || HasLastprivates) && S.getSingleClause(OMPC_nowait)) {
// 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);
} }
return OMPD_single; return OMPD_single;
} }
▲ Show 20 Lines • Show All 635 Lines • Show Last 20 Lines

cfe/trunk/test/OpenMP/sections_lastprivate_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-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
template <class T>
struct S {
T f;
S(T a) : f(a) {}
S() : f() {}
S<T> &operator=(const S<T> &);
operator T() { return T(); }
~S() {}
};
volatile int g = 1212;
// 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 { i32 }
// CHECK: [[CAP_TMAIN_TY:%.+]] = type { i{{[0-9]+}}*, [2 x i{{[0-9]+}}]*, [2 x [[S_INT_TY]]]*, [[S_INT_TY]]* }
// 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: [[SECTIONS_BARRIER_LOC:@.+]] = private unnamed_addr constant %{{.+}} { i32 0, i32 194, i32 0, i32 0, i8*
// CHECK-DAG: [[X:@.+]] = global double 0.0
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 lastprivate(t_var, vec, s_arr, var)
{
vec[0] = t_var;
#pragma omp section
s_arr[0] = var;
}
return T();
}
namespace A {
double x;
}
namespace B {
using A::x;
}
int main() {
#ifdef LAMBDA
// LAMBDA: [[G:@.+]] = global i{{[0-9]+}} 1212,
// 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 lastprivate(g)
{
// LAMBDA: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* [[ARG:%.+]])
// LAMBDA: alloca i{{[0-9]+}},
// LAMBDA: alloca i{{[0-9]+}},
// LAMBDA: alloca i{{[0-9]+}},
// LAMBDA: alloca i{{[0-9]+}},
// LAMBDA: alloca i{{[0-9]+}},
// LAMBDA: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
// LAMBDA: store %{{.+}}* [[ARG]], %{{.+}}** [[ARG_REF:%.+]],
// 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)
// LAMBDA: store volatile i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
// 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]])
// LAMBDA: call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 [[GTID]])
g = 1;
// Check for final copying of private values back to original vars.
// LAMBDA: [[IS_LAST_VAL:%.+]] = load i32, i32* [[IS_LAST_ADDR]],
// LAMBDA: [[IS_LAST_ITER:%.+]] = icmp ne i32 [[IS_LAST_VAL]], 0
// LAMBDA: br i1 [[IS_LAST_ITER:%.+]], label %[[LAST_THEN:.+]], label %[[LAST_DONE:.+]]
// LAMBDA: [[LAST_THEN]]
// Actual copying.
// original g=private_g;
// LAMBDA: [[G_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
// LAMBDA: store volatile i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G]],
// LAMBDA: br label %[[LAST_DONE]]
// LAMBDA: [[LAST_DONE]]
// LAMBDA: call i32 @__kmpc_cancel_barrier(%{{.+}}* @{{.+}}, i{{[0-9]+}} [[GTID]])
#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 i{{[0-9]+}}*, i{{[0-9]+}}** [[G_PTR_REF]]
// LAMBDA: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}* [[G_REF]]
}();
}
}();
return 0;
#elif defined(BLOCKS)
// BLOCKS: [[G:@.+]] = global i{{[0-9]+}} 1212,
// 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 lastprivate(g)
{
// BLOCKS: define{{.*}} internal{{.*}} void [[OMP_REGION]](i32* %{{.+}}, i32* %{{.+}}, %{{.+}}* [[ARG:%.+]])
// BLOCKS: alloca i{{[0-9]+}},
// BLOCKS: alloca i{{[0-9]+}},
// BLOCKS: alloca i{{[0-9]+}},
// BLOCKS: alloca i{{[0-9]+}},
// BLOCKS: alloca i{{[0-9]+}},
// BLOCKS: [[G_PRIVATE_ADDR:%.+]] = alloca i{{[0-9]+}},
// BLOCKS: store %{{.+}}* [[ARG]], %{{.+}}** [[ARG_REF:%.+]],
// 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)
// BLOCKS: store volatile i{{[0-9]+}} 1, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// BLOCKS: i{{[0-9]+}}* [[G_PRIVATE_ADDR]]
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// BLOCKS: call void {{%.+}}(i8
// BLOCKS: call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 [[GTID]])
g = 1;
// Check for final copying of private values back to original vars.
// BLOCKS: [[IS_LAST_VAL:%.+]] = load i32, i32* [[IS_LAST_ADDR]],
// BLOCKS: [[IS_LAST_ITER:%.+]] = icmp ne i32 [[IS_LAST_VAL]], 0
// BLOCKS: br i1 [[IS_LAST_ITER:%.+]], label %[[LAST_THEN:.+]], label %[[LAST_DONE:.+]]
// BLOCKS: [[LAST_THEN]]
// Actual copying.
// original g=private_g;
// BLOCKS: [[G_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[G_PRIVATE_ADDR]],
// BLOCKS: store volatile i{{[0-9]+}} [[G_VAL]], i{{[0-9]+}}* [[G]],
// BLOCKS: br label %[[LAST_DONE]]
// BLOCKS: [[LAST_DONE]]
// BLOCKS: call i32 @__kmpc_cancel_barrier(%{{.+}}* @{{.+}}, i{{[0-9]+}} [[GTID]])
#pragma omp section
^{
// BLOCKS: define {{.+}} void {{@.+}}(i8*
g = 2;
// BLOCKS-NOT: [[G]]{{[[^:word:]]}}
// BLOCKS: store volatile i{{[0-9]+}} 2, i{{[0-9]+}}*
// 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 lastprivate(t_var, vec, s_arr, var)
{
{
vec[0] = t_var;
s_arr[0] = var;
}
}
#pragma omp parallel
#pragma omp sections lastprivate(A::x, B::x)
{
A::x++;
#pragma omp section
;
}
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 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]+}}*, %{{.+}}*)* [[MAIN_MICROTASK1:@.+]] to void
// CHECK: = call {{.+}} [[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-NOT: alloca i{{[0-9]+}},
// CHECK-NOT: alloca [2 x i{{[0-9]+}}],
// CHECK-NOT: alloca [2 x [[S_FLOAT_TY]]],
// CHECK-NOT: alloca [[S_FLOAT_TY]],
// CHECK: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
// CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_REF]]
// 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
// <Skip loop body>
// CHECK-NOT: call void [[S_FLOAT_TY_DESTR]]([[S_FLOAT_TY]]* [[VAR_PRIV]])
// CHECK-NOT: call void [[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 internal void [[MAIN_MICROTASK1]](i{{[0-9]+}}* [[GTID_ADDR:%.+]], i{{[0-9]+}}* %{{.+}}, %{{.+}}* %{{.+}})
// CHECK: [[X_PRIV:%.+]] = alloca double,
// CHECK-NOT: alloca double
// Check for default initialization.
// CHECK-NOT: [[X_PRIV]]
// CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_REF]]
// CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
// CHECK: call void @__kmpc_for_static_init_4(%{{.+}}* @{{.+}}, i32 [[GTID]], i32 34, i32* [[IS_LAST_ADDR:%.+]], i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
// <Skip loop body>
// CHECK: call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 [[GTID]])
// Check for final copying of private values back to original vars.
// CHECK: [[IS_LAST_VAL:%.+]] = load i32, i32* [[IS_LAST_ADDR]],
// CHECK: [[IS_LAST_ITER:%.+]] = icmp ne i32 [[IS_LAST_VAL]], 0
// CHECK: br i1 [[IS_LAST_ITER:%.+]], label %[[LAST_THEN:.+]], label %[[LAST_DONE:.+]]
// CHECK: [[LAST_THEN]]
// Actual copying.
// original x=private_x;
// CHECK: [[X_VAL:%.+]] = load double, double* [[X_PRIV]],
// CHECK: store double [[X_VAL]], double* [[X]],
// CHECK-NEXT: br label %[[LAST_DONE]]
// CHECK: [[LAST_DONE]]
// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[SECTIONS_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_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 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: [[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: store i{{[0-9]+}}* [[GTID_ADDR]], i{{[0-9]+}}** [[GTID_ADDR_REF:%.+]]
// Check for default initialization.
// CHECK: [[T_VAR_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CHECK: [[T_VAR_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[T_VAR_PTR_REF]],
// CHECK-NOT: [[T_VAR_PRIV]]
// CHECK: [[VEC_PTR_REF:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 1
// CHECK: [[VEC_REF:%.+]] = load [2 x i{{[0-9]+}}]*, [2 x i{{[0-9]+}}]** [[VEC_PTR_REF:%.+]],
// CHECK-NOT: [[VEC_PRIV]]
// CHECK: [[S_ARR_REF_PTR:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 2
// CHECK: [[S_ARR_REF:%.+]] = load [2 x [[S_INT_TY]]]*, [2 x [[S_INT_TY]]]** [[S_ARR_REF_PTR]],
// CHECK: [[S_ARR_PRIV_ITEM:%.+]] = phi [[S_INT_TY]]*
// CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR]]([[S_INT_TY]]* [[S_ARR_PRIV_ITEM]])
// CHECK: [[VAR_REF_PTR:%.+]] = getelementptr inbounds [[CAP_TMAIN_TY]], [[CAP_TMAIN_TY]]* %{{.+}}, i{{[0-9]+}} 0, i{{[0-9]+}} 3
// CHECK: [[VAR_REF:%.+]] = load [[S_INT_TY]]*, [[S_INT_TY]]** [[VAR_REF_PTR]],
// CHECK: call {{.*}} [[S_INT_TY_DEF_CONSTR]]([[S_INT_TY]]* [[VAR_PRIV]])
// CHECK: call {{.+}} @__kmpc_for_static_init_4(%{{.+}}* @{{.+}}, i32 %{{.+}}, i32 34, i32* [[IS_LAST_ADDR:%.+]], i32* %{{.+}}, i32* %{{.+}}, i32* %{{.+}}, i32 1, i32 1)
// <Skip loop body>
// CHECK: call void @__kmpc_for_static_fini(%{{.+}}* @{{.+}}, i32 %{{.+}})
// Check for final copying of private values back to original vars.
// CHECK: [[IS_LAST_VAL:%.+]] = load i32, i32* [[IS_LAST_ADDR]],
// CHECK: [[IS_LAST_ITER:%.+]] = icmp ne i32 [[IS_LAST_VAL]], 0
// CHECK: br i1 [[IS_LAST_ITER:%.+]], label %[[LAST_THEN:.+]], label %[[LAST_DONE:.+]]
// CHECK: [[LAST_THEN]]
// Actual copying.
// original t_var=private_t_var;
// CHECK: [[T_VAR_VAL:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[T_VAR_PRIV]],
// CHECK: store i{{[0-9]+}} [[T_VAR_VAL]], i{{[0-9]+}}* [[T_VAR_REF]],
// original vec[]=private_vec[];
// CHECK: [[VEC_DEST:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_REF]] to i8*
// CHECK: [[VEC_SRC:%.+]] = bitcast [2 x i{{[0-9]+}}]* [[VEC_PRIV]] to i8*
// CHECK: call void @llvm.memcpy.{{.+}}(i8* [[VEC_DEST]], i8* [[VEC_SRC]],
// original s_arr[]=private_s_arr[];
// CHECK: [[S_ARR_BEGIN:%.+]] = getelementptr inbounds [2 x [[S_INT_TY]]], [2 x [[S_INT_TY]]]* [[S_ARR_REF]], i{{[0-9]+}} 0, i{{[0-9]+}} 0
// CHECK: [[S_ARR_PRIV_BEGIN:%.+]] = bitcast [2 x [[S_INT_TY]]]* [[S_ARR_PRIV]] to [[S_INT_TY]]*
// CHECK: [[S_ARR_END:%.+]] = getelementptr [[S_INT_TY]], [[S_INT_TY]]* [[S_ARR_BEGIN]], i{{[0-9]+}} 2
// CHECK: [[IS_EMPTY:%.+]] = icmp eq [[S_INT_TY]]* [[S_ARR_BEGIN]], [[S_ARR_END]]
// CHECK: br i1 [[IS_EMPTY]], label %[[S_ARR_BODY_DONE:.+]], label %[[S_ARR_BODY:.+]]
// CHECK: [[S_ARR_BODY]]
// CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN:@.+]]([[S_INT_TY]]* {{.+}}, [[S_INT_TY]]* {{.+}})
// CHECK: br i1 {{.+}}, label %[[S_ARR_BODY_DONE]], label %[[S_ARR_BODY]]
// CHECK: [[S_ARR_BODY_DONE]]
// original var=private_var;
// CHECK: call {{.*}} [[S_INT_TY_COPY_ASSIGN:@.+]]([[S_INT_TY]]* [[VAR_REF]], [[S_INT_TY]]* {{.*}} [[VAR_PRIV]])
// CHECK: br label %[[LAST_DONE]]
// CHECK: [[LAST_DONE]]
// 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: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_REF]]
// CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[SECTIONS_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
// CHECK: [[GTID_REF:%.+]] = load i{{[0-9]+}}*, i{{[0-9]+}}** [[GTID_ADDR_REF]]
// CHECK: [[GTID:%.+]] = load i{{[0-9]+}}, i{{[0-9]+}}* [[GTID_REF]]
// CHECK: call i32 @__kmpc_cancel_barrier(%{{.+}}* [[IMPLICIT_BARRIER_LOC]], i{{[0-9]+}} [[GTID]])
// CHECK: ret void
#endif