replica
/
hanchenye-llvm-project
2
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

[OPENMP] Fix crash on reduction for complex variables. 

reworked codegen for reduction operation for complex types to avoid crash

llvm-svn: 258394
This commit is contained in:
Alexey Bataev 2016-01-21 12:35:58 +00:00
parent ff262fa807
commit 8524d15954
4 changed files with 25 additions and 17 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
clang/lib/CodeGen/CGOpenMPRuntime.cpp
View File

@ -3548,14 +3548,16 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
E = CGF.EmitAnyExpr(EExpr);
CGF.EmitOMPAtomicSimpleUpdateExpr(
X, E, BO, /*IsXLHSInRHSPart=*/true, llvm::Monotonic, Loc,
[&CGF, UpExpr, VD, IPriv](RValue XRValue) {
[&CGF, UpExpr, VD, IPriv, Loc](RValue XRValue) {
CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
PrivateScope.addPrivate(VD, [&CGF, VD, XRValue]() -> Address {
Address LHSTemp = CGF.CreateMemTemp(VD->getType());
CGF.EmitStoreThroughLValue(
XRValue, CGF.MakeAddrLValue(LHSTemp, VD->getType()));
return LHSTemp;
});
PrivateScope.addPrivate(
VD, [&CGF, VD, XRValue, Loc]() -> Address {
Address LHSTemp = CGF.CreateMemTemp(VD->getType());
CGF.emitOMPSimpleStore(
CGF.MakeAddrLValue(LHSTemp, VD->getType()), XRValue,
VD->getType().getNonReferenceType(), Loc);
return LHSTemp;
});
(void)PrivateScope.Privatize();
return CGF.EmitAnyExpr(UpExpr);
});

20
clang/lib/CodeGen/CGStmtOpenMP.cpp
View File

@ -2167,17 +2167,17 @@ static void emitSimpleAtomicStore(CodeGenFunction &CGF, bool IsSeqCst,
}
}
static void emitSimpleStore(CodeGenFunction &CGF, LValue LVal, RValue RVal,
QualType RValTy, SourceLocation Loc) {
switch (CGF.getEvaluationKind(LVal.getType())) {
void CodeGenFunction::emitOMPSimpleStore(LValue LVal, RValue RVal,
QualType RValTy, SourceLocation Loc) {
switch (getEvaluationKind(LVal.getType())) {
case TEK_Scalar:
CGF.EmitStoreThroughLValue(RValue::get(convertToScalarValue(
CGF, RVal, RValTy, LVal.getType(), Loc)),
LVal);
EmitStoreThroughLValue(RValue::get(convertToScalarValue(
*this, RVal, RValTy, LVal.getType(), Loc)),
LVal);
break;
case TEK_Complex:
CGF.EmitStoreOfComplex(
convertToComplexValue(CGF, RVal, RValTy, LVal.getType(), Loc), LVal,
EmitStoreOfComplex(
convertToComplexValue(*this, RVal, RValTy, LVal.getType(), Loc), LVal,
/*isInit=*/false);
break;
case TEK_Aggregate:
@ -2205,7 +2205,7 @@ static void EmitOMPAtomicReadExpr(CodeGenFunction &CGF, bool IsSeqCst,
// list.
if (IsSeqCst)
CGF.CGM.getOpenMPRuntime().emitFlush(CGF, llvm::None, Loc);
emitSimpleStore(CGF, VLValue, Res, X->getType().getNonReferenceType(), Loc);
CGF.emitOMPSimpleStore(VLValue, Res, X->getType().getNonReferenceType(), Loc);
}
static void EmitOMPAtomicWriteExpr(CodeGenFunction &CGF, bool IsSeqCst,
@ -2463,7 +2463,7 @@ static void EmitOMPAtomicCaptureExpr(CodeGenFunction &CGF, bool IsSeqCst,
}
}
// Emit post-update store to 'v' of old/new 'x' value.
emitSimpleStore(CGF, VLValue, NewVVal, NewVValType, Loc);
CGF.emitOMPSimpleStore(VLValue, NewVVal, NewVValType, Loc);
// OpenMP, 2.12.6, atomic Construct
// Any atomic construct with a seq_cst clause forces the atomically
// performed operation to include an implicit flush operation without a

2
clang/lib/CodeGen/CodeGenFunction.h
View File

@ -2212,6 +2212,8 @@ public:
llvm::Function *GenerateOpenMPCapturedStmtFunction(const CapturedStmt &S);
void GenerateOpenMPCapturedVars(const CapturedStmt &S,
SmallVectorImpl<llvm::Value *> &CapturedVars);
void emitOMPSimpleStore(LValue LVal, RValue RVal, QualType RValTy,
SourceLocation Loc);
/// \brief Perform element by element copying of arrays with type \a
/// OriginalType from \a SrcAddr to \a DestAddr using copying procedure
/// generated by \a CopyGen.

4
clang/test/OpenMP/parallel_reduction_codegen.cpp
View File

@ -158,6 +158,7 @@ int main() {
int vec[] = {1, 2};
S<float> s_arr[] = {1, 2};
S<float> var(3), var1;
float _Complex cf;
#pragma omp parallel reduction(+:t_var) reduction(&:var) reduction(&& : var1) reduction(min: t_var1)
{
vec[0] = t_var;
@ -169,6 +170,8 @@ int main() {
vec[0] = t_var;
s_arr[0] = var;
}
#pragma omp parallel reduction(+ : cf)
;
return tmain<int>();
#endif
}
@ -178,6 +181,7 @@ int main() {
// CHECK: call {{.*}} [[S_FLOAT_TY_CONSTR:@.+]]([[S_FLOAT_TY]]* [[TEST]])
// CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 6, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i32]*, float*, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]*, float*)* [[MAIN_MICROTASK:@.+]] to void
// CHECK: call void (%{{.+}}*, i{{[0-9]+}}, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)*, ...) @__kmpc_fork_call(%{{.+}}* @{{.+}}, i{{[0-9]+}} 6, void (i{{[0-9]+}}*, i{{[0-9]+}}*, ...)* bitcast (void (i{{[0-9]+}}*, i{{[0-9]+}}*, [2 x i32]*, float*, [2 x [[S_FLOAT_TY]]]*, [[S_FLOAT_TY]]*, [[S_FLOAT_TY]]*, float*)* [[MAIN_MICROTASK1:@.+]] 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]+}}*, { float, float }*)* [[MAIN_MICROTASK2:@.+]] to void
// CHECK: = call {{.*}}i{{.+}} [[TMAIN_INT:@.+]]()
// CHECK: call {{.*}} [[S_FLOAT_TY_DESTR:@.+]]([[S_FLOAT_TY]]*
// CHECK: ret