summaryrefslogtreecommitdiff
path: root/lib/CodeGen/CGOpenMPRuntime.cpp
diff options
context:
space:
mode:
authorDimitry Andric <dim@FreeBSD.org>2018-07-28 11:06:01 +0000
committerDimitry Andric <dim@FreeBSD.org>2018-07-28 11:06:01 +0000
commit486754660bb926339aefcf012a3f848592babb8b (patch)
treeecdbc446c9876f4f120f701c243373cd3cb43db3 /lib/CodeGen/CGOpenMPRuntime.cpp
parent55e6d896ad333f07bb3b1ba487df214fc268a4ab (diff)
vendor/clang/clang-trunk-r338150
Notes
Diffstat (limited to 'lib/CodeGen/CGOpenMPRuntime.cpp')
-rw-r--r--lib/CodeGen/CGOpenMPRuntime.cpp4158
1 files changed, 2660 insertions, 1498 deletions
diff --git a/lib/CodeGen/CGOpenMPRuntime.cpp b/lib/CodeGen/CGOpenMPRuntime.cpp
index fa38ee80bf41..3730b9af12fa 100644
--- a/lib/CodeGen/CGOpenMPRuntime.cpp
+++ b/lib/CodeGen/CGOpenMPRuntime.cpp
@@ -14,12 +14,13 @@
#include "CGCXXABI.h"
#include "CGCleanup.h"
#include "CGOpenMPRuntime.h"
+#include "CGRecordLayout.h"
#include "CodeGenFunction.h"
#include "clang/CodeGen/ConstantInitBuilder.h"
#include "clang/AST/Decl.h"
#include "clang/AST/StmtOpenMP.h"
+#include "clang/Basic/BitmaskEnum.h"
#include "llvm/ADT/ArrayRef.h"
-#include "llvm/ADT/BitmaskEnum.h"
#include "llvm/Bitcode/BitcodeReader.h"
#include "llvm/IR/CallSite.h"
#include "llvm/IR/DerivedTypes.h"
@@ -33,20 +34,20 @@ using namespace clang;
using namespace CodeGen;
namespace {
-/// \brief Base class for handling code generation inside OpenMP regions.
+/// Base class for handling code generation inside OpenMP regions.
class CGOpenMPRegionInfo : public CodeGenFunction::CGCapturedStmtInfo {
public:
- /// \brief Kinds of OpenMP regions used in codegen.
+ /// Kinds of OpenMP regions used in codegen.
enum CGOpenMPRegionKind {
- /// \brief Region with outlined function for standalone 'parallel'
+ /// Region with outlined function for standalone 'parallel'
/// directive.
ParallelOutlinedRegion,
- /// \brief Region with outlined function for standalone 'task' directive.
+ /// Region with outlined function for standalone 'task' directive.
TaskOutlinedRegion,
- /// \brief Region for constructs that do not require function outlining,
+ /// Region for constructs that do not require function outlining,
/// like 'for', 'sections', 'atomic' etc. directives.
InlinedRegion,
- /// \brief Region with outlined function for standalone 'target' directive.
+ /// Region with outlined function for standalone 'target' directive.
TargetRegion,
};
@@ -63,14 +64,14 @@ public:
: CGCapturedStmtInfo(CR_OpenMP), RegionKind(RegionKind), CodeGen(CodeGen),
Kind(Kind), HasCancel(HasCancel) {}
- /// \brief Get a variable or parameter for storing global thread id
+ /// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
virtual const VarDecl *getThreadIDVariable() const = 0;
- /// \brief Emit the captured statement body.
+ /// Emit the captured statement body.
void EmitBody(CodeGenFunction &CGF, const Stmt *S) override;
- /// \brief Get an LValue for the current ThreadID variable.
+ /// Get an LValue for the current ThreadID variable.
/// \return LValue for thread id variable. This LValue always has type int32*.
virtual LValue getThreadIDVariableLValue(CodeGenFunction &CGF);
@@ -95,7 +96,7 @@ protected:
bool HasCancel;
};
-/// \brief API for captured statement code generation in OpenMP constructs.
+/// API for captured statement code generation in OpenMP constructs.
class CGOpenMPOutlinedRegionInfo final : public CGOpenMPRegionInfo {
public:
CGOpenMPOutlinedRegionInfo(const CapturedStmt &CS, const VarDecl *ThreadIDVar,
@@ -108,11 +109,11 @@ public:
assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.");
}
- /// \brief Get a variable or parameter for storing global thread id
+ /// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }
- /// \brief Get the name of the capture helper.
+ /// Get the name of the capture helper.
StringRef getHelperName() const override { return HelperName; }
static bool classof(const CGCapturedStmtInfo *Info) {
@@ -122,13 +123,13 @@ public:
}
private:
- /// \brief A variable or parameter storing global thread id for OpenMP
+ /// A variable or parameter storing global thread id for OpenMP
/// constructs.
const VarDecl *ThreadIDVar;
StringRef HelperName;
};
-/// \brief API for captured statement code generation in OpenMP constructs.
+/// API for captured statement code generation in OpenMP constructs.
class CGOpenMPTaskOutlinedRegionInfo final : public CGOpenMPRegionInfo {
public:
class UntiedTaskActionTy final : public PrePostActionTy {
@@ -144,11 +145,12 @@ public:
void Enter(CodeGenFunction &CGF) override {
if (Untied) {
// Emit task switching point.
- auto PartIdLVal = CGF.EmitLoadOfPointerLValue(
+ LValue PartIdLVal = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(PartIDVar),
PartIDVar->getType()->castAs<PointerType>());
- auto *Res = CGF.EmitLoadOfScalar(PartIdLVal, SourceLocation());
- auto *DoneBB = CGF.createBasicBlock(".untied.done.");
+ llvm::Value *Res =
+ CGF.EmitLoadOfScalar(PartIdLVal, PartIDVar->getLocation());
+ llvm::BasicBlock *DoneBB = CGF.createBasicBlock(".untied.done.");
UntiedSwitch = CGF.Builder.CreateSwitch(Res, DoneBB);
CGF.EmitBlock(DoneBB);
CGF.EmitBranchThroughCleanup(CGF.ReturnBlock);
@@ -160,7 +162,7 @@ public:
}
void emitUntiedSwitch(CodeGenFunction &CGF) const {
if (Untied) {
- auto PartIdLVal = CGF.EmitLoadOfPointerLValue(
+ LValue PartIdLVal = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(PartIDVar),
PartIDVar->getType()->castAs<PointerType>());
CGF.EmitStoreOfScalar(CGF.Builder.getInt32(UntiedSwitch->getNumCases()),
@@ -188,14 +190,14 @@ public:
assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.");
}
- /// \brief Get a variable or parameter for storing global thread id
+ /// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }
- /// \brief Get an LValue for the current ThreadID variable.
+ /// Get an LValue for the current ThreadID variable.
LValue getThreadIDVariableLValue(CodeGenFunction &CGF) override;
- /// \brief Get the name of the capture helper.
+ /// Get the name of the capture helper.
StringRef getHelperName() const override { return ".omp_outlined."; }
void emitUntiedSwitch(CodeGenFunction &CGF) override {
@@ -209,14 +211,14 @@ public:
}
private:
- /// \brief A variable or parameter storing global thread id for OpenMP
+ /// A variable or parameter storing global thread id for OpenMP
/// constructs.
const VarDecl *ThreadIDVar;
/// Action for emitting code for untied tasks.
const UntiedTaskActionTy &Action;
};
-/// \brief API for inlined captured statement code generation in OpenMP
+/// API for inlined captured statement code generation in OpenMP
/// constructs.
class CGOpenMPInlinedRegionInfo : public CGOpenMPRegionInfo {
public:
@@ -227,7 +229,7 @@ public:
OldCSI(OldCSI),
OuterRegionInfo(dyn_cast_or_null<CGOpenMPRegionInfo>(OldCSI)) {}
- // \brief Retrieve the value of the context parameter.
+ // Retrieve the value of the context parameter.
llvm::Value *getContextValue() const override {
if (OuterRegionInfo)
return OuterRegionInfo->getContextValue();
@@ -242,7 +244,7 @@ public:
llvm_unreachable("No context value for inlined OpenMP region");
}
- /// \brief Lookup the captured field decl for a variable.
+ /// Lookup the captured field decl for a variable.
const FieldDecl *lookup(const VarDecl *VD) const override {
if (OuterRegionInfo)
return OuterRegionInfo->lookup(VD);
@@ -257,7 +259,7 @@ public:
return nullptr;
}
- /// \brief Get a variable or parameter for storing global thread id
+ /// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override {
if (OuterRegionInfo)
@@ -265,14 +267,14 @@ public:
return nullptr;
}
- /// \brief Get an LValue for the current ThreadID variable.
+ /// Get an LValue for the current ThreadID variable.
LValue getThreadIDVariableLValue(CodeGenFunction &CGF) override {
if (OuterRegionInfo)
return OuterRegionInfo->getThreadIDVariableLValue(CGF);
llvm_unreachable("No LValue for inlined OpenMP construct");
}
- /// \brief Get the name of the capture helper.
+ /// Get the name of the capture helper.
StringRef getHelperName() const override {
if (auto *OuterRegionInfo = getOldCSI())
return OuterRegionInfo->getHelperName();
@@ -294,12 +296,12 @@ public:
~CGOpenMPInlinedRegionInfo() override = default;
private:
- /// \brief CodeGen info about outer OpenMP region.
+ /// CodeGen info about outer OpenMP region.
CodeGenFunction::CGCapturedStmtInfo *OldCSI;
CGOpenMPRegionInfo *OuterRegionInfo;
};
-/// \brief API for captured statement code generation in OpenMP target
+/// API for captured statement code generation in OpenMP target
/// constructs. For this captures, implicit parameters are used instead of the
/// captured fields. The name of the target region has to be unique in a given
/// application so it is provided by the client, because only the client has
@@ -312,11 +314,11 @@ public:
/*HasCancel=*/false),
HelperName(HelperName) {}
- /// \brief This is unused for target regions because each starts executing
+ /// This is unused for target regions because each starts executing
/// with a single thread.
const VarDecl *getThreadIDVariable() const override { return nullptr; }
- /// \brief Get the name of the capture helper.
+ /// Get the name of the capture helper.
StringRef getHelperName() const override { return HelperName; }
static bool classof(const CGCapturedStmtInfo *Info) {
@@ -331,7 +333,7 @@ private:
static void EmptyCodeGen(CodeGenFunction &, PrePostActionTy &) {
llvm_unreachable("No codegen for expressions");
}
-/// \brief API for generation of expressions captured in a innermost OpenMP
+/// API for generation of expressions captured in a innermost OpenMP
/// region.
class CGOpenMPInnerExprInfo final : public CGOpenMPInlinedRegionInfo {
public:
@@ -343,7 +345,7 @@ public:
// Make sure the globals captured in the provided statement are local by
// using the privatization logic. We assume the same variable is not
// captured more than once.
- for (auto &C : CS.captures()) {
+ for (const auto &C : CS.captures()) {
if (!C.capturesVariable() && !C.capturesVariableByCopy())
continue;
@@ -354,33 +356,32 @@ public:
DeclRefExpr DRE(const_cast<VarDecl *>(VD),
/*RefersToEnclosingVariableOrCapture=*/false,
VD->getType().getNonReferenceType(), VK_LValue,
- SourceLocation());
- PrivScope.addPrivate(VD, [&CGF, &DRE]() -> Address {
- return CGF.EmitLValue(&DRE).getAddress();
- });
+ C.getLocation());
+ PrivScope.addPrivate(
+ VD, [&CGF, &DRE]() { return CGF.EmitLValue(&DRE).getAddress(); });
}
(void)PrivScope.Privatize();
}
- /// \brief Lookup the captured field decl for a variable.
+ /// Lookup the captured field decl for a variable.
const FieldDecl *lookup(const VarDecl *VD) const override {
- if (auto *FD = CGOpenMPInlinedRegionInfo::lookup(VD))
+ if (const FieldDecl *FD = CGOpenMPInlinedRegionInfo::lookup(VD))
return FD;
return nullptr;
}
- /// \brief Emit the captured statement body.
+ /// Emit the captured statement body.
void EmitBody(CodeGenFunction &CGF, const Stmt *S) override {
llvm_unreachable("No body for expressions");
}
- /// \brief Get a variable or parameter for storing global thread id
+ /// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override {
llvm_unreachable("No thread id for expressions");
}
- /// \brief Get the name of the capture helper.
+ /// Get the name of the capture helper.
StringRef getHelperName() const override {
llvm_unreachable("No helper name for expressions");
}
@@ -392,14 +393,15 @@ private:
CodeGenFunction::OMPPrivateScope PrivScope;
};
-/// \brief RAII for emitting code of OpenMP constructs.
+/// RAII for emitting code of OpenMP constructs.
class InlinedOpenMPRegionRAII {
CodeGenFunction &CGF;
llvm::DenseMap<const VarDecl *, FieldDecl *> LambdaCaptureFields;
FieldDecl *LambdaThisCaptureField = nullptr;
+ const CodeGen::CGBlockInfo *BlockInfo = nullptr;
public:
- /// \brief Constructs region for combined constructs.
+ /// Constructs region for combined constructs.
/// \param CodeGen Code generation sequence for combined directives. Includes
/// a list of functions used for code generation of implicitly inlined
/// regions.
@@ -412,6 +414,8 @@ public:
std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
LambdaThisCaptureField = CGF.LambdaThisCaptureField;
CGF.LambdaThisCaptureField = nullptr;
+ BlockInfo = CGF.BlockInfo;
+ CGF.BlockInfo = nullptr;
}
~InlinedOpenMPRegionRAII() {
@@ -422,28 +426,29 @@ public:
CGF.CapturedStmtInfo = OldCSI;
std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
CGF.LambdaThisCaptureField = LambdaThisCaptureField;
+ CGF.BlockInfo = BlockInfo;
}
};
-/// \brief Values for bit flags used in the ident_t to describe the fields.
+/// Values for bit flags used in the ident_t to describe the fields.
/// All enumeric elements are named and described in accordance with the code
/// from http://llvm.org/svn/llvm-project/openmp/trunk/runtime/src/kmp.h
enum OpenMPLocationFlags : unsigned {
- /// \brief Use trampoline for internal microtask.
+ /// Use trampoline for internal microtask.
OMP_IDENT_IMD = 0x01,
- /// \brief Use c-style ident structure.
+ /// Use c-style ident structure.
OMP_IDENT_KMPC = 0x02,
- /// \brief Atomic reduction option for kmpc_reduce.
+ /// Atomic reduction option for kmpc_reduce.
OMP_ATOMIC_REDUCE = 0x10,
- /// \brief Explicit 'barrier' directive.
+ /// Explicit 'barrier' directive.
OMP_IDENT_BARRIER_EXPL = 0x20,
- /// \brief Implicit barrier in code.
+ /// Implicit barrier in code.
OMP_IDENT_BARRIER_IMPL = 0x40,
- /// \brief Implicit barrier in 'for' directive.
+ /// Implicit barrier in 'for' directive.
OMP_IDENT_BARRIER_IMPL_FOR = 0x40,
- /// \brief Implicit barrier in 'sections' directive.
+ /// Implicit barrier in 'sections' directive.
OMP_IDENT_BARRIER_IMPL_SECTIONS = 0xC0,
- /// \brief Implicit barrier in 'single' directive.
+ /// Implicit barrier in 'single' directive.
OMP_IDENT_BARRIER_IMPL_SINGLE = 0x140,
/// Call of __kmp_for_static_init for static loop.
OMP_IDENT_WORK_LOOP = 0x200,
@@ -454,7 +459,7 @@ enum OpenMPLocationFlags : unsigned {
LLVM_MARK_AS_BITMASK_ENUM(/*LargestValue=*/OMP_IDENT_WORK_DISTRIBUTE)
};
-/// \brief Describes ident structure that describes a source location.
+/// Describes ident structure that describes a source location.
/// All descriptions are taken from
/// http://llvm.org/svn/llvm-project/openmp/trunk/runtime/src/kmp.h
/// Original structure:
@@ -481,24 +486,24 @@ enum OpenMPLocationFlags : unsigned {
/// */
/// } ident_t;
enum IdentFieldIndex {
- /// \brief might be used in Fortran
+ /// might be used in Fortran
IdentField_Reserved_1,
- /// \brief OMP_IDENT_xxx flags; OMP_IDENT_KMPC identifies this union member.
+ /// OMP_IDENT_xxx flags; OMP_IDENT_KMPC identifies this union member.
IdentField_Flags,
- /// \brief Not really used in Fortran any more
+ /// Not really used in Fortran any more
IdentField_Reserved_2,
- /// \brief Source[4] in Fortran, do not use for C++
+ /// Source[4] in Fortran, do not use for C++
IdentField_Reserved_3,
- /// \brief String describing the source location. The string is composed of
+ /// String describing the source location. The string is composed of
/// semi-colon separated fields which describe the source file, the function
/// and a pair of line numbers that delimit the construct.
IdentField_PSource
};
-/// \brief Schedule types for 'omp for' loops (these enumerators are taken from
+/// Schedule types for 'omp for' loops (these enumerators are taken from
/// the enum sched_type in kmp.h).
enum OpenMPSchedType {
- /// \brief Lower bound for default (unordered) versions.
+ /// Lower bound for default (unordered) versions.
OMP_sch_lower = 32,
OMP_sch_static_chunked = 33,
OMP_sch_static = 34,
@@ -508,7 +513,7 @@ enum OpenMPSchedType {
OMP_sch_auto = 38,
/// static with chunk adjustment (e.g., simd)
OMP_sch_static_balanced_chunked = 45,
- /// \brief Lower bound for 'ordered' versions.
+ /// Lower bound for 'ordered' versions.
OMP_ord_lower = 64,
OMP_ord_static_chunked = 65,
OMP_ord_static = 66,
@@ -517,7 +522,7 @@ enum OpenMPSchedType {
OMP_ord_runtime = 69,
OMP_ord_auto = 70,
OMP_sch_default = OMP_sch_static,
- /// \brief dist_schedule types
+ /// dist_schedule types
OMP_dist_sch_static_chunked = 91,
OMP_dist_sch_static = 92,
/// Support for OpenMP 4.5 monotonic and nonmonotonic schedule modifiers.
@@ -528,13 +533,13 @@ enum OpenMPSchedType {
};
enum OpenMPRTLFunction {
- /// \brief Call to void __kmpc_fork_call(ident_t *loc, kmp_int32 argc,
+ /// Call to void __kmpc_fork_call(ident_t *loc, kmp_int32 argc,
/// kmpc_micro microtask, ...);
OMPRTL__kmpc_fork_call,
- /// \brief Call to void *__kmpc_threadprivate_cached(ident_t *loc,
+ /// Call to void *__kmpc_threadprivate_cached(ident_t *loc,
/// kmp_int32 global_tid, void *data, size_t size, void ***cache);
OMPRTL__kmpc_threadprivate_cached,
- /// \brief Call to void __kmpc_threadprivate_register( ident_t *,
+ /// Call to void __kmpc_threadprivate_register( ident_t *,
/// void *data, kmpc_ctor ctor, kmpc_cctor cctor, kmpc_dtor dtor);
OMPRTL__kmpc_threadprivate_register,
// Call to __kmpc_int32 kmpc_global_thread_num(ident_t *loc);
@@ -742,11 +747,11 @@ void RegionCodeGenTy::operator()(CodeGenFunction &CGF) const {
/// UDR decl used for reduction.
static const OMPDeclareReductionDecl *
getReductionInit(const Expr *ReductionOp) {
- if (auto *CE = dyn_cast<CallExpr>(ReductionOp))
- if (auto *OVE = dyn_cast<OpaqueValueExpr>(CE->getCallee()))
- if (auto *DRE =
+ if (const auto *CE = dyn_cast<CallExpr>(ReductionOp))
+ if (const auto *OVE = dyn_cast<OpaqueValueExpr>(CE->getCallee()))
+ if (const auto *DRE =
dyn_cast<DeclRefExpr>(OVE->getSourceExpr()->IgnoreImpCasts()))
- if (auto *DRD = dyn_cast<OMPDeclareReductionDecl>(DRE->getDecl()))
+ if (const auto *DRD = dyn_cast<OMPDeclareReductionDecl>(DRE->getDecl()))
return DRD;
return nullptr;
}
@@ -759,48 +764,51 @@ static void emitInitWithReductionInitializer(CodeGenFunction &CGF,
if (DRD->getInitializer()) {
std::pair<llvm::Function *, llvm::Function *> Reduction =
CGF.CGM.getOpenMPRuntime().getUserDefinedReduction(DRD);
- auto *CE = cast<CallExpr>(InitOp);
- auto *OVE = cast<OpaqueValueExpr>(CE->getCallee());
+ const auto *CE = cast<CallExpr>(InitOp);
+ const auto *OVE = cast<OpaqueValueExpr>(CE->getCallee());
const Expr *LHS = CE->getArg(/*Arg=*/0)->IgnoreParenImpCasts();
const Expr *RHS = CE->getArg(/*Arg=*/1)->IgnoreParenImpCasts();
- auto *LHSDRE = cast<DeclRefExpr>(cast<UnaryOperator>(LHS)->getSubExpr());
- auto *RHSDRE = cast<DeclRefExpr>(cast<UnaryOperator>(RHS)->getSubExpr());
+ const auto *LHSDRE =
+ cast<DeclRefExpr>(cast<UnaryOperator>(LHS)->getSubExpr());
+ const auto *RHSDRE =
+ cast<DeclRefExpr>(cast<UnaryOperator>(RHS)->getSubExpr());
CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
PrivateScope.addPrivate(cast<VarDecl>(LHSDRE->getDecl()),
- [=]() -> Address { return Private; });
+ [=]() { return Private; });
PrivateScope.addPrivate(cast<VarDecl>(RHSDRE->getDecl()),
- [=]() -> Address { return Original; });
+ [=]() { return Original; });
(void)PrivateScope.Privatize();
RValue Func = RValue::get(Reduction.second);
CodeGenFunction::OpaqueValueMapping Map(CGF, OVE, Func);
CGF.EmitIgnoredExpr(InitOp);
} else {
llvm::Constant *Init = CGF.CGM.EmitNullConstant(Ty);
+ std::string Name = CGF.CGM.getOpenMPRuntime().getName({"init"});
auto *GV = new llvm::GlobalVariable(
CGF.CGM.getModule(), Init->getType(), /*isConstant=*/true,
- llvm::GlobalValue::PrivateLinkage, Init, ".init");
+ llvm::GlobalValue::PrivateLinkage, Init, Name);
LValue LV = CGF.MakeNaturalAlignAddrLValue(GV, Ty);
RValue InitRVal;
switch (CGF.getEvaluationKind(Ty)) {
case TEK_Scalar:
- InitRVal = CGF.EmitLoadOfLValue(LV, SourceLocation());
+ InitRVal = CGF.EmitLoadOfLValue(LV, DRD->getLocation());
break;
case TEK_Complex:
InitRVal =
- RValue::getComplex(CGF.EmitLoadOfComplex(LV, SourceLocation()));
+ RValue::getComplex(CGF.EmitLoadOfComplex(LV, DRD->getLocation()));
break;
case TEK_Aggregate:
InitRVal = RValue::getAggregate(LV.getAddress());
break;
}
- OpaqueValueExpr OVE(SourceLocation(), Ty, VK_RValue);
+ OpaqueValueExpr OVE(DRD->getLocation(), Ty, VK_RValue);
CodeGenFunction::OpaqueValueMapping OpaqueMap(CGF, &OVE, InitRVal);
CGF.EmitAnyExprToMem(&OVE, Private, Ty.getQualifiers(),
/*IsInitializer=*/false);
}
}
-/// \brief Emit initialization of arrays of complex types.
+/// Emit initialization of arrays of complex types.
/// \param DestAddr Address of the array.
/// \param Type Type of array.
/// \param Init Initial expression of array.
@@ -814,8 +822,8 @@ static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
QualType ElementTy;
// Drill down to the base element type on both arrays.
- auto ArrayTy = Type->getAsArrayTypeUnsafe();
- auto NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, DestAddr);
+ const ArrayType *ArrayTy = Type->getAsArrayTypeUnsafe();
+ llvm::Value *NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, DestAddr);
DestAddr =
CGF.Builder.CreateElementBitCast(DestAddr, DestAddr.getElementType());
if (DRD)
@@ -825,18 +833,18 @@ static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
llvm::Value *SrcBegin = nullptr;
if (DRD)
SrcBegin = SrcAddr.getPointer();
- auto DestBegin = DestAddr.getPointer();
+ llvm::Value *DestBegin = DestAddr.getPointer();
// Cast from pointer to array type to pointer to single element.
- auto DestEnd = CGF.Builder.CreateGEP(DestBegin, NumElements);
+ llvm::Value *DestEnd = CGF.Builder.CreateGEP(DestBegin, NumElements);
// The basic structure here is a while-do loop.
- auto BodyBB = CGF.createBasicBlock("omp.arrayinit.body");
- auto DoneBB = CGF.createBasicBlock("omp.arrayinit.done");
- auto IsEmpty =
+ llvm::BasicBlock *BodyBB = CGF.createBasicBlock("omp.arrayinit.body");
+ llvm::BasicBlock *DoneBB = CGF.createBasicBlock("omp.arrayinit.done");
+ llvm::Value *IsEmpty =
CGF.Builder.CreateICmpEQ(DestBegin, DestEnd, "omp.arrayinit.isempty");
CGF.Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);
// Enter the loop body, making that address the current address.
- auto EntryBB = CGF.Builder.GetInsertBlock();
+ llvm::BasicBlock *EntryBB = CGF.Builder.GetInsertBlock();
CGF.EmitBlock(BodyBB);
CharUnits ElementSize = CGF.getContext().getTypeSizeInChars(ElementTy);
@@ -871,16 +879,16 @@ static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
if (DRD) {
// Shift the address forward by one element.
- auto SrcElementNext = CGF.Builder.CreateConstGEP1_32(
+ llvm::Value *SrcElementNext = CGF.Builder.CreateConstGEP1_32(
SrcElementPHI, /*Idx0=*/1, "omp.arraycpy.dest.element");
SrcElementPHI->addIncoming(SrcElementNext, CGF.Builder.GetInsertBlock());
}
// Shift the address forward by one element.
- auto DestElementNext = CGF.Builder.CreateConstGEP1_32(
+ llvm::Value *DestElementNext = CGF.Builder.CreateConstGEP1_32(
DestElementPHI, /*Idx0=*/1, "omp.arraycpy.dest.element");
// Check whether we've reached the end.
- auto Done =
+ llvm::Value *Done =
CGF.Builder.CreateICmpEQ(DestElementNext, DestEnd, "omp.arraycpy.done");
CGF.Builder.CreateCondBr(Done, DoneBB, BodyBB);
DestElementPHI->addIncoming(DestElementNext, CGF.Builder.GetInsertBlock());
@@ -889,6 +897,25 @@ static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
}
+static llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy>
+isDeclareTargetDeclaration(const ValueDecl *VD) {
+ for (const Decl *D : VD->redecls()) {
+ if (!D->hasAttrs())
+ continue;
+ if (const auto *Attr = D->getAttr<OMPDeclareTargetDeclAttr>())
+ return Attr->getMapType();
+ }
+ if (const auto *V = dyn_cast<VarDecl>(VD)) {
+ if (const VarDecl *TD = V->getTemplateInstantiationPattern())
+ return isDeclareTargetDeclaration(TD);
+ } else if (const auto *FD = dyn_cast<FunctionDecl>(VD)) {
+ if (const auto *TD = FD->getTemplateInstantiationPattern())
+ return isDeclareTargetDeclaration(TD);
+ }
+
+ return llvm::None;
+}
+
LValue ReductionCodeGen::emitSharedLValue(CodeGenFunction &CGF, const Expr *E) {
return CGF.EmitOMPSharedLValue(E);
}
@@ -906,7 +933,7 @@ void ReductionCodeGen::emitAggregateInitialization(
// Emit VarDecl with copy init for arrays.
// Get the address of the original variable captured in current
// captured region.
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
bool EmitDeclareReductionInit =
DRD && (DRD->getInitializer() || !PrivateVD->hasInit());
@@ -926,7 +953,7 @@ ReductionCodeGen::ReductionCodeGen(ArrayRef<const Expr *> Shareds,
BaseDecls.reserve(Shareds.size());
auto IPriv = Privates.begin();
auto IRed = ReductionOps.begin();
- for (const auto *Ref : Shareds) {
+ for (const Expr *Ref : Shareds) {
ClausesData.emplace_back(Ref, *IPriv, *IRed);
std::advance(IPriv, 1);
std::advance(IRed, 1);
@@ -942,7 +969,7 @@ void ReductionCodeGen::emitSharedLValue(CodeGenFunction &CGF, unsigned N) {
}
void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N) {
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
bool AsArraySection = isa<OMPArraySectionExpr>(ClausesData[N].Ref);
@@ -955,7 +982,7 @@ void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N) {
}
llvm::Value *Size;
llvm::Value *SizeInChars;
- llvm::Type *ElemType =
+ auto *ElemType =
cast<llvm::PointerType>(SharedAddresses[N].first.getPointer()->getType())
->getElementType();
auto *ElemSizeOf = llvm::ConstantExpr::getSizeOf(ElemType);
@@ -981,7 +1008,7 @@ void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N) {
void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N,
llvm::Value *Size) {
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
if (!PrivateType->isVariablyModifiedType()) {
@@ -1002,9 +1029,10 @@ void ReductionCodeGen::emitInitialization(
CodeGenFunction &CGF, unsigned N, Address PrivateAddr, LValue SharedLVal,
llvm::function_ref<bool(CodeGenFunction &)> DefaultInit) {
assert(SharedAddresses.size() > N && "No variable was generated");
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
- auto *DRD = getReductionInit(ClausesData[N].ReductionOp);
+ const OMPDeclareReductionDecl *DRD =
+ getReductionInit(ClausesData[N].ReductionOp);
QualType PrivateType = PrivateVD->getType();
PrivateAddr = CGF.Builder.CreateElementBitCast(
PrivateAddr, CGF.ConvertTypeForMem(PrivateType));
@@ -1029,7 +1057,7 @@ void ReductionCodeGen::emitInitialization(
}
bool ReductionCodeGen::needCleanups(unsigned N) {
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
QualType::DestructionKind DTorKind = PrivateType.isDestructedType();
@@ -1038,7 +1066,7 @@ bool ReductionCodeGen::needCleanups(unsigned N) {
void ReductionCodeGen::emitCleanups(CodeGenFunction &CGF, unsigned N,
Address PrivateAddr) {
- auto *PrivateVD =
+ const auto *PrivateVD =
cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
QualType::DestructionKind DTorKind = PrivateType.isDestructedType();
@@ -1054,9 +1082,9 @@ static LValue loadToBegin(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
BaseTy = BaseTy.getNonReferenceType();
while ((BaseTy->isPointerType() || BaseTy->isReferenceType()) &&
!CGF.getContext().hasSameType(BaseTy, ElTy)) {
- if (auto *PtrTy = BaseTy->getAs<PointerType>())
+ if (const auto *PtrTy = BaseTy->getAs<PointerType>()) {
BaseLV = CGF.EmitLoadOfPointerLValue(BaseLV.getAddress(), PtrTy);
- else {
+ } else {
LValue RefLVal = CGF.MakeAddrLValue(BaseLV.getAddress(), BaseTy);
BaseLV = CGF.EmitLoadOfReferenceLValue(RefLVal);
}
@@ -1097,28 +1125,32 @@ static Address castToBase(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
return Address(Addr, BaseLVAlignment);
}
-Address ReductionCodeGen::adjustPrivateAddress(CodeGenFunction &CGF, unsigned N,
- Address PrivateAddr) {
- const DeclRefExpr *DE;
+static const VarDecl *getBaseDecl(const Expr *Ref, const DeclRefExpr *&DE) {
const VarDecl *OrigVD = nullptr;
- if (auto *OASE = dyn_cast<OMPArraySectionExpr>(ClausesData[N].Ref)) {
- auto *Base = OASE->getBase()->IgnoreParenImpCasts();
- while (auto *TempOASE = dyn_cast<OMPArraySectionExpr>(Base))
+ if (const auto *OASE = dyn_cast<OMPArraySectionExpr>(Ref)) {
+ const Expr *Base = OASE->getBase()->IgnoreParenImpCasts();
+ while (const auto *TempOASE = dyn_cast<OMPArraySectionExpr>(Base))
Base = TempOASE->getBase()->IgnoreParenImpCasts();
- while (auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
+ while (const auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
Base = TempASE->getBase()->IgnoreParenImpCasts();
DE = cast<DeclRefExpr>(Base);
OrigVD = cast<VarDecl>(DE->getDecl());
- } else if (auto *ASE = dyn_cast<ArraySubscriptExpr>(ClausesData[N].Ref)) {
- auto *Base = ASE->getBase()->IgnoreParenImpCasts();
- while (auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
+ } else if (const auto *ASE = dyn_cast<ArraySubscriptExpr>(Ref)) {
+ const Expr *Base = ASE->getBase()->IgnoreParenImpCasts();
+ while (const auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
Base = TempASE->getBase()->IgnoreParenImpCasts();
DE = cast<DeclRefExpr>(Base);
OrigVD = cast<VarDecl>(DE->getDecl());
}
- if (OrigVD) {
+ return OrigVD;
+}
+
+Address ReductionCodeGen::adjustPrivateAddress(CodeGenFunction &CGF, unsigned N,
+ Address PrivateAddr) {
+ const DeclRefExpr *DE;
+ if (const VarDecl *OrigVD = ::getBaseDecl(ClausesData[N].Ref, DE)) {
BaseDecls.emplace_back(OrigVD);
- auto OriginalBaseLValue = CGF.EmitLValue(DE);
+ LValue OriginalBaseLValue = CGF.EmitLValue(DE);
LValue BaseLValue =
loadToBegin(CGF, OrigVD->getType(), SharedAddresses[N].first.getType(),
OriginalBaseLValue);
@@ -1140,7 +1172,8 @@ Address ReductionCodeGen::adjustPrivateAddress(CodeGenFunction &CGF, unsigned N,
}
bool ReductionCodeGen::usesReductionInitializer(unsigned N) const {
- auto *DRD = getReductionInit(ClausesData[N].ReductionOp);
+ const OMPDeclareReductionDecl *DRD =
+ getReductionInit(ClausesData[N].ReductionOp);
return DRD && DRD->getInitializer();
}
@@ -1170,12 +1203,38 @@ LValue CGOpenMPTaskOutlinedRegionInfo::getThreadIDVariableLValue(
AlignmentSource::Decl);
}
-CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM)
- : CGM(CGM), OffloadEntriesInfoManager(CGM) {
- IdentTy = llvm::StructType::create(
- "ident_t", CGM.Int32Ty /* reserved_1 */, CGM.Int32Ty /* flags */,
- CGM.Int32Ty /* reserved_2 */, CGM.Int32Ty /* reserved_3 */,
- CGM.Int8PtrTy /* psource */);
+static FieldDecl *addFieldToRecordDecl(ASTContext &C, DeclContext *DC,
+ QualType FieldTy) {
+ auto *Field = FieldDecl::Create(
+ C, DC, SourceLocation(), SourceLocation(), /*Id=*/nullptr, FieldTy,
+ C.getTrivialTypeSourceInfo(FieldTy, SourceLocation()),
+ /*BW=*/nullptr, /*Mutable=*/false, /*InitStyle=*/ICIS_NoInit);
+ Field->setAccess(AS_public);
+ DC->addDecl(Field);
+ return Field;
+}
+
+CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM, StringRef FirstSeparator,
+ StringRef Separator)
+ : CGM(CGM), FirstSeparator(FirstSeparator), Separator(Separator),
+ OffloadEntriesInfoManager(CGM) {
+ ASTContext &C = CGM.getContext();
+ RecordDecl *RD = C.buildImplicitRecord("ident_t");
+ QualType KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
+ RD->startDefinition();
+ // reserved_1
+ addFieldToRecordDecl(C, RD, KmpInt32Ty);
+ // flags
+ addFieldToRecordDecl(C, RD, KmpInt32Ty);
+ // reserved_2
+ addFieldToRecordDecl(C, RD, KmpInt32Ty);
+ // reserved_3
+ addFieldToRecordDecl(C, RD, KmpInt32Ty);
+ // psource
+ addFieldToRecordDecl(C, RD, C.VoidPtrTy);
+ RD->completeDefinition();
+ IdentQTy = C.getRecordType(RD);
+ IdentTy = CGM.getTypes().ConvertRecordDeclType(RD);
KmpCriticalNameTy = llvm::ArrayType::get(CGM.Int32Ty, /*NumElements*/ 8);
loadOffloadInfoMetadata();
@@ -1185,12 +1244,23 @@ void CGOpenMPRuntime::clear() {
InternalVars.clear();
}
+std::string CGOpenMPRuntime::getName(ArrayRef<StringRef> Parts) const {
+ SmallString<128> Buffer;
+ llvm::raw_svector_ostream OS(Buffer);
+ StringRef Sep = FirstSeparator;
+ for (StringRef Part : Parts) {
+ OS << Sep << Part;
+ Sep = Separator;
+ }
+ return OS.str();
+}
+
static llvm::Function *
emitCombinerOrInitializer(CodeGenModule &CGM, QualType Ty,
const Expr *CombinerInitializer, const VarDecl *In,
const VarDecl *Out, bool IsCombiner) {
// void .omp_combiner.(Ty *in, Ty *out);
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
QualType PtrTy = C.getPointerType(Ty).withRestrict();
FunctionArgList Args;
ImplicitParamDecl OmpOutParm(C, /*DC=*/nullptr, Out->getLocation(),
@@ -1199,28 +1269,30 @@ emitCombinerOrInitializer(CodeGenModule &CGM, QualType Ty,
/*Id=*/nullptr, PtrTy, ImplicitParamDecl::Other);
Args.push_back(&OmpOutParm);
Args.push_back(&OmpInParm);
- auto &FnInfo =
+ const CGFunctionInfo &FnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
- auto *Fn = llvm::Function::Create(
- FnTy, llvm::GlobalValue::InternalLinkage,
- IsCombiner ? ".omp_combiner." : ".omp_initializer.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, FnInfo);
+ llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName(
+ {IsCombiner ? "omp_combiner" : "omp_initializer", ""});
+ auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
+ Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
Fn->removeFnAttr(llvm::Attribute::NoInline);
Fn->removeFnAttr(llvm::Attribute::OptimizeNone);
Fn->addFnAttr(llvm::Attribute::AlwaysInline);
CodeGenFunction CGF(CGM);
// Map "T omp_in;" variable to "*omp_in_parm" value in all expressions.
// Map "T omp_out;" variable to "*omp_out_parm" value in all expressions.
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, In->getLocation(),
+ Out->getLocation());
CodeGenFunction::OMPPrivateScope Scope(CGF);
Address AddrIn = CGF.GetAddrOfLocalVar(&OmpInParm);
- Scope.addPrivate(In, [&CGF, AddrIn, PtrTy]() -> Address {
+ Scope.addPrivate(In, [&CGF, AddrIn, PtrTy]() {
return CGF.EmitLoadOfPointerLValue(AddrIn, PtrTy->castAs<PointerType>())
.getAddress();
});
Address AddrOut = CGF.GetAddrOfLocalVar(&OmpOutParm);
- Scope.addPrivate(Out, [&CGF, AddrOut, PtrTy]() -> Address {
+ Scope.addPrivate(Out, [&CGF, AddrOut, PtrTy]() {
return CGF.EmitLoadOfPointerLValue(AddrOut, PtrTy->castAs<PointerType>())
.getAddress();
});
@@ -1242,7 +1314,7 @@ void CGOpenMPRuntime::emitUserDefinedReduction(
CodeGenFunction *CGF, const OMPDeclareReductionDecl *D) {
if (UDRMap.count(D) > 0)
return;
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
if (!In || !Out) {
In = &C.Idents.get("omp_in");
Out = &C.Idents.get("omp_out");
@@ -1252,7 +1324,7 @@ void CGOpenMPRuntime::emitUserDefinedReduction(
cast<VarDecl>(D->lookup(Out).front()),
/*IsCombiner=*/true);
llvm::Function *Initializer = nullptr;
- if (auto *Init = D->getInitializer()) {
+ if (const Expr *Init = D->getInitializer()) {
if (!Priv || !Orig) {
Priv = &C.Idents.get("omp_priv");
Orig = &C.Idents.get("omp_orig");
@@ -1265,7 +1337,7 @@ void CGOpenMPRuntime::emitUserDefinedReduction(
cast<VarDecl>(D->lookup(Priv).front()),
/*IsCombiner=*/false);
}
- UDRMap.insert(std::make_pair(D, std::make_pair(Combiner, Initializer)));
+ UDRMap.try_emplace(D, Combiner, Initializer);
if (CGF) {
auto &Decls = FunctionUDRMap.FindAndConstruct(CGF->CurFn);
Decls.second.push_back(D);
@@ -1281,25 +1353,6 @@ CGOpenMPRuntime::getUserDefinedReduction(const OMPDeclareReductionDecl *D) {
return UDRMap.lookup(D);
}
-// Layout information for ident_t.
-static CharUnits getIdentAlign(CodeGenModule &CGM) {
- return CGM.getPointerAlign();
-}
-static CharUnits getIdentSize(CodeGenModule &CGM) {
- assert((4 * CGM.getPointerSize()).isMultipleOf(CGM.getPointerAlign()));
- return CharUnits::fromQuantity(16) + CGM.getPointerSize();
-}
-static CharUnits getOffsetOfIdentField(IdentFieldIndex Field) {
- // All the fields except the last are i32, so this works beautifully.
- return unsigned(Field) * CharUnits::fromQuantity(4);
-}
-static Address createIdentFieldGEP(CodeGenFunction &CGF, Address Addr,
- IdentFieldIndex Field,
- const llvm::Twine &Name = "") {
- auto Offset = getOffsetOfIdentField(Field);
- return CGF.Builder.CreateStructGEP(Addr, Field, Offset, Name);
-}
-
static llvm::Value *emitParallelOrTeamsOutlinedFunction(
CodeGenModule &CGM, const OMPExecutableDirective &D, const CapturedStmt *CS,
const VarDecl *ThreadIDVar, OpenMPDirectiveKind InnermostKind,
@@ -1308,19 +1361,20 @@ static llvm::Value *emitParallelOrTeamsOutlinedFunction(
"thread id variable must be of type kmp_int32 *");
CodeGenFunction CGF(CGM, true);
bool HasCancel = false;
- if (auto *OPD = dyn_cast<OMPParallelDirective>(&D))
+ if (const auto *OPD = dyn_cast<OMPParallelDirective>(&D))
HasCancel = OPD->hasCancel();
- else if (auto *OPSD = dyn_cast<OMPParallelSectionsDirective>(&D))
+ else if (const auto *OPSD = dyn_cast<OMPParallelSectionsDirective>(&D))
HasCancel = OPSD->hasCancel();
- else if (auto *OPFD = dyn_cast<OMPParallelForDirective>(&D))
+ else if (const auto *OPFD = dyn_cast<OMPParallelForDirective>(&D))
HasCancel = OPFD->hasCancel();
- else if (auto *OPFD = dyn_cast<OMPTargetParallelForDirective>(&D))
+ else if (const auto *OPFD = dyn_cast<OMPTargetParallelForDirective>(&D))
HasCancel = OPFD->hasCancel();
- else if (auto *OPFD = dyn_cast<OMPDistributeParallelForDirective>(&D))
+ else if (const auto *OPFD = dyn_cast<OMPDistributeParallelForDirective>(&D))
HasCancel = OPFD->hasCancel();
- else if (auto *OPFD = dyn_cast<OMPTeamsDistributeParallelForDirective>(&D))
+ else if (const auto *OPFD =
+ dyn_cast<OMPTeamsDistributeParallelForDirective>(&D))
HasCancel = OPFD->hasCancel();
- else if (auto *OPFD =
+ else if (const auto *OPFD =
dyn_cast<OMPTargetTeamsDistributeParallelForDirective>(&D))
HasCancel = OPFD->hasCancel();
CGOpenMPOutlinedRegionInfo CGInfo(*CS, ThreadIDVar, CodeGen, InnermostKind,
@@ -1352,8 +1406,8 @@ llvm::Value *CGOpenMPRuntime::emitTaskOutlinedFunction(
bool Tied, unsigned &NumberOfParts) {
auto &&UntiedCodeGen = [this, &D, TaskTVar](CodeGenFunction &CGF,
PrePostActionTy &) {
- auto *ThreadID = getThreadID(CGF, D.getLocStart());
- auto *UpLoc = emitUpdateLocation(CGF, D.getLocStart());
+ llvm::Value *ThreadID = getThreadID(CGF, D.getLocStart());
+ llvm::Value *UpLoc = emitUpdateLocation(CGF, D.getLocStart());
llvm::Value *TaskArgs[] = {
UpLoc, ThreadID,
CGF.EmitLoadOfPointerLValue(CGF.GetAddrOfLocalVar(TaskTVar),
@@ -1366,21 +1420,69 @@ llvm::Value *CGOpenMPRuntime::emitTaskOutlinedFunction(
CodeGen.setAction(Action);
assert(!ThreadIDVar->getType()->isPointerType() &&
"thread id variable must be of type kmp_int32 for tasks");
- auto *CS = cast<CapturedStmt>(D.getAssociatedStmt());
- auto *TD = dyn_cast<OMPTaskDirective>(&D);
+ const OpenMPDirectiveKind Region =
+ isOpenMPTaskLoopDirective(D.getDirectiveKind()) ? OMPD_taskloop
+ : OMPD_task;
+ const CapturedStmt *CS = D.getCapturedStmt(Region);
+ const auto *TD = dyn_cast<OMPTaskDirective>(&D);
CodeGenFunction CGF(CGM, true);
CGOpenMPTaskOutlinedRegionInfo CGInfo(*CS, ThreadIDVar, CodeGen,
InnermostKind,
TD ? TD->hasCancel() : false, Action);
CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
- auto *Res = CGF.GenerateCapturedStmtFunction(*CS);
+ llvm::Value *Res = CGF.GenerateCapturedStmtFunction(*CS);
if (!Tied)
NumberOfParts = Action.getNumberOfParts();
return Res;
}
+static void buildStructValue(ConstantStructBuilder &Fields, CodeGenModule &CGM,
+ const RecordDecl *RD, const CGRecordLayout &RL,
+ ArrayRef<llvm::Constant *> Data) {
+ llvm::StructType *StructTy = RL.getLLVMType();
+ unsigned PrevIdx = 0;
+ ConstantInitBuilder CIBuilder(CGM);
+ auto DI = Data.begin();
+ for (const FieldDecl *FD : RD->fields()) {
+ unsigned Idx = RL.getLLVMFieldNo(FD);
+ // Fill the alignment.
+ for (unsigned I = PrevIdx; I < Idx; ++I)
+ Fields.add(llvm::Constant::getNullValue(StructTy->getElementType(I)));
+ PrevIdx = Idx + 1;
+ Fields.add(*DI);
+ ++DI;
+ }
+}
+
+template <class... As>
+static llvm::GlobalVariable *
+createConstantGlobalStruct(CodeGenModule &CGM, QualType Ty,
+ ArrayRef<llvm::Constant *> Data, const Twine &Name,
+ As &&... Args) {
+ const auto *RD = cast<RecordDecl>(Ty->getAsTagDecl());
+ const CGRecordLayout &RL = CGM.getTypes().getCGRecordLayout(RD);
+ ConstantInitBuilder CIBuilder(CGM);
+ ConstantStructBuilder Fields = CIBuilder.beginStruct(RL.getLLVMType());
+ buildStructValue(Fields, CGM, RD, RL, Data);
+ return Fields.finishAndCreateGlobal(
+ Name, CGM.getContext().getAlignOfGlobalVarInChars(Ty),
+ /*isConstant=*/true, std::forward<As>(Args)...);
+}
+
+template <typename T>
+static void
+createConstantGlobalStructAndAddToParent(CodeGenModule &CGM, QualType Ty,
+ ArrayRef<llvm::Constant *> Data,
+ T &Parent) {
+ const auto *RD = cast<RecordDecl>(Ty->getAsTagDecl());
+ const CGRecordLayout &RL = CGM.getTypes().getCGRecordLayout(RD);
+ ConstantStructBuilder Fields = Parent.beginStruct(RL.getLLVMType());
+ buildStructValue(Fields, CGM, RD, RL, Data);
+ Fields.finishAndAddTo(Parent);
+}
+
Address CGOpenMPRuntime::getOrCreateDefaultLocation(unsigned Flags) {
- CharUnits Align = getIdentAlign(CGM);
+ CharUnits Align = CGM.getContext().getTypeAlignInChars(IdentQTy);
llvm::Value *Entry = OpenMPDefaultLocMap.lookup(Flags);
if (!Entry) {
if (!DefaultOpenMPPSource) {
@@ -1394,17 +1496,15 @@ Address CGOpenMPRuntime::getOrCreateDefaultLocation(unsigned Flags) {
llvm::ConstantExpr::getBitCast(DefaultOpenMPPSource, CGM.Int8PtrTy);
}
- ConstantInitBuilder builder(CGM);
- auto fields = builder.beginStruct(IdentTy);
- fields.addInt(CGM.Int32Ty, 0);
- fields.addInt(CGM.Int32Ty, Flags);
- fields.addInt(CGM.Int32Ty, 0);
- fields.addInt(CGM.Int32Ty, 0);
- fields.add(DefaultOpenMPPSource);
- auto DefaultOpenMPLocation =
- fields.finishAndCreateGlobal("", Align, /*isConstant*/ true,
- llvm::GlobalValue::PrivateLinkage);
- DefaultOpenMPLocation->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
+ llvm::Constant *Data[] = {llvm::ConstantInt::getNullValue(CGM.Int32Ty),
+ llvm::ConstantInt::get(CGM.Int32Ty, Flags),
+ llvm::ConstantInt::getNullValue(CGM.Int32Ty),
+ llvm::ConstantInt::getNullValue(CGM.Int32Ty),
+ DefaultOpenMPPSource};
+ llvm::GlobalValue *DefaultOpenMPLocation = createConstantGlobalStruct(
+ CGM, IdentQTy, Data, "", llvm::GlobalValue::PrivateLinkage);
+ DefaultOpenMPLocation->setUnnamedAddr(
+ llvm::GlobalValue::UnnamedAddr::Global);
OpenMPDefaultLocMap[Flags] = Entry = DefaultOpenMPLocation;
}
@@ -1422,17 +1522,17 @@ llvm::Value *CGOpenMPRuntime::emitUpdateLocation(CodeGenFunction &CGF,
assert(CGF.CurFn && "No function in current CodeGenFunction.");
+ CharUnits Align = CGM.getContext().getTypeAlignInChars(IdentQTy);
Address LocValue = Address::invalid();
auto I = OpenMPLocThreadIDMap.find(CGF.CurFn);
if (I != OpenMPLocThreadIDMap.end())
- LocValue = Address(I->second.DebugLoc, getIdentAlign(CGF.CGM));
+ LocValue = Address(I->second.DebugLoc, Align);
// OpenMPLocThreadIDMap may have null DebugLoc and non-null ThreadID, if
// GetOpenMPThreadID was called before this routine.
if (!LocValue.isValid()) {
// Generate "ident_t .kmpc_loc.addr;"
- Address AI = CGF.CreateTempAlloca(IdentTy, getIdentAlign(CGF.CGM),
- ".kmpc_loc.addr");
+ Address AI = CGF.CreateMemTemp(IdentQTy, ".kmpc_loc.addr");
auto &Elem = OpenMPLocThreadIDMap.FindAndConstruct(CGF.CurFn);
Elem.second.DebugLoc = AI.getPointer();
LocValue = AI;
@@ -1440,29 +1540,30 @@ llvm::Value *CGOpenMPRuntime::emitUpdateLocation(CodeGenFunction &CGF,
CGBuilderTy::InsertPointGuard IPG(CGF.Builder);
CGF.Builder.SetInsertPoint(CGF.AllocaInsertPt);
CGF.Builder.CreateMemCpy(LocValue, getOrCreateDefaultLocation(Flags),
- CGM.getSize(getIdentSize(CGF.CGM)));
+ CGF.getTypeSize(IdentQTy));
}
// char **psource = &.kmpc_loc_<flags>.addr.psource;
- Address PSource = createIdentFieldGEP(CGF, LocValue, IdentField_PSource);
+ LValue Base = CGF.MakeAddrLValue(LocValue, IdentQTy);
+ auto Fields = cast<RecordDecl>(IdentQTy->getAsTagDecl())->field_begin();
+ LValue PSource =
+ CGF.EmitLValueForField(Base, *std::next(Fields, IdentField_PSource));
- auto OMPDebugLoc = OpenMPDebugLocMap.lookup(Loc.getRawEncoding());
+ llvm::Value *OMPDebugLoc = OpenMPDebugLocMap.lookup(Loc.getRawEncoding());
if (OMPDebugLoc == nullptr) {
SmallString<128> Buffer2;
llvm::raw_svector_ostream OS2(Buffer2);
// Build debug location
PresumedLoc PLoc = CGF.getContext().getSourceManager().getPresumedLoc(Loc);
OS2 << ";" << PLoc.getFilename() << ";";
- if (const FunctionDecl *FD =
- dyn_cast_or_null<FunctionDecl>(CGF.CurFuncDecl)) {
+ if (const auto *FD = dyn_cast_or_null<FunctionDecl>(CGF.CurFuncDecl))
OS2 << FD->getQualifiedNameAsString();
- }
OS2 << ";" << PLoc.getLine() << ";" << PLoc.getColumn() << ";;";
OMPDebugLoc = CGF.Builder.CreateGlobalStringPtr(OS2.str());
OpenMPDebugLocMap[Loc.getRawEncoding()] = OMPDebugLoc;
}
// *psource = ";<File>;<Function>;<Line>;<Column>;;";
- CGF.Builder.CreateStore(OMPDebugLoc, PSource);
+ CGF.EmitStoreOfScalar(OMPDebugLoc, PSource);
// Our callers always pass this to a runtime function, so for
// convenience, go ahead and return a naked pointer.
@@ -1490,8 +1591,8 @@ llvm::Value *CGOpenMPRuntime::getThreadID(CodeGenFunction &CGF,
dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo)) {
if (OMPRegionInfo->getThreadIDVariable()) {
// Check if this an outlined function with thread id passed as argument.
- auto LVal = OMPRegionInfo->getThreadIDVariableLValue(CGF);
- ThreadID = CGF.EmitLoadOfLValue(LVal, Loc).getScalarVal();
+ LValue LVal = OMPRegionInfo->getThreadIDVariableLValue(CGF);
+ ThreadID = CGF.EmitLoadOfScalar(LVal, Loc);
// If value loaded in entry block, cache it and use it everywhere in
// function.
if (CGF.Builder.GetInsertBlock() == CGF.AllocaInsertPt->getParent()) {
@@ -1509,7 +1610,7 @@ llvm::Value *CGOpenMPRuntime::getThreadID(CodeGenFunction &CGF,
// function.
CGBuilderTy::InsertPointGuard IPG(CGF.Builder);
CGF.Builder.SetInsertPoint(CGF.AllocaInsertPt);
- auto *Call = CGF.Builder.CreateCall(
+ llvm::CallInst *Call = CGF.Builder.CreateCall(
createRuntimeFunction(OMPRTL__kmpc_global_thread_num),
emitUpdateLocation(CGF, Loc));
Call->setCallingConv(CGF.getRuntimeCC());
@@ -1523,17 +1624,14 @@ void CGOpenMPRuntime::functionFinished(CodeGenFunction &CGF) {
if (OpenMPLocThreadIDMap.count(CGF.CurFn))
OpenMPLocThreadIDMap.erase(CGF.CurFn);
if (FunctionUDRMap.count(CGF.CurFn) > 0) {
- for(auto *D : FunctionUDRMap[CGF.CurFn]) {
+ for(auto *D : FunctionUDRMap[CGF.CurFn])
UDRMap.erase(D);
- }
FunctionUDRMap.erase(CGF.CurFn);
}
}
llvm::Type *CGOpenMPRuntime::getIdentTyPointerTy() {
- if (!IdentTy) {
- }
- return llvm::PointerType::getUnqual(IdentTy);
+ return IdentTy->getPointerTo();
}
llvm::Type *CGOpenMPRuntime::getKmpc_MicroPointerTy() {
@@ -1555,7 +1653,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// microtask, ...);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
getKmpc_MicroPointerTy()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ true);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_fork_call");
break;
@@ -1563,7 +1661,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_global_thread_num: {
// Build kmp_int32 __kmpc_global_thread_num(ident_t *loc);
llvm::Type *TypeParams[] = {getIdentTyPointerTy()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_global_thread_num");
break;
@@ -1574,7 +1672,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.VoidPtrTy, CGM.SizeTy,
CGM.VoidPtrTy->getPointerTo()->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidPtrTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_threadprivate_cached");
break;
@@ -1585,7 +1683,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), CGM.Int32Ty,
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_critical");
break;
@@ -1596,7 +1694,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
llvm::PointerType::getUnqual(KmpCriticalNameTy),
CGM.IntPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_critical_with_hint");
break;
@@ -1605,21 +1703,22 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void __kmpc_threadprivate_register(ident_t *, void *data,
// kmpc_ctor ctor, kmpc_cctor cctor, kmpc_dtor dtor);
// typedef void *(*kmpc_ctor)(void *);
- auto KmpcCtorTy =
+ auto *KmpcCtorTy =
llvm::FunctionType::get(CGM.VoidPtrTy, CGM.VoidPtrTy,
/*isVarArg*/ false)->getPointerTo();
// typedef void *(*kmpc_cctor)(void *, void *);
llvm::Type *KmpcCopyCtorTyArgs[] = {CGM.VoidPtrTy, CGM.VoidPtrTy};
- auto KmpcCopyCtorTy =
+ auto *KmpcCopyCtorTy =
llvm::FunctionType::get(CGM.VoidPtrTy, KmpcCopyCtorTyArgs,
- /*isVarArg*/ false)->getPointerTo();
+ /*isVarArg*/ false)
+ ->getPointerTo();
// typedef void (*kmpc_dtor)(void *);
- auto KmpcDtorTy =
+ auto *KmpcDtorTy =
llvm::FunctionType::get(CGM.VoidTy, CGM.VoidPtrTy, /*isVarArg*/ false)
->getPointerTo();
llvm::Type *FnTyArgs[] = {getIdentTyPointerTy(), CGM.VoidPtrTy, KmpcCtorTy,
KmpcCopyCtorTy, KmpcDtorTy};
- auto FnTy = llvm::FunctionType::get(CGM.VoidTy, FnTyArgs,
+ auto *FnTy = llvm::FunctionType::get(CGM.VoidTy, FnTyArgs,
/*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_threadprivate_register");
break;
@@ -1630,7 +1729,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), CGM.Int32Ty,
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_end_critical");
break;
@@ -1639,7 +1738,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build kmp_int32 __kmpc_cancel_barrier(ident_t *loc, kmp_int32
// global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name*/ "__kmpc_cancel_barrier");
break;
@@ -1647,7 +1746,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_barrier: {
// Build void __kmpc_barrier(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name*/ "__kmpc_barrier");
break;
@@ -1655,7 +1754,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_for_static_fini: {
// Build void __kmpc_for_static_fini(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_for_static_fini");
break;
@@ -1665,7 +1764,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// kmp_int32 num_threads)
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_push_num_threads");
break;
@@ -1674,7 +1773,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void __kmpc_serialized_parallel(ident_t *loc, kmp_int32
// global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_serialized_parallel");
break;
@@ -1683,7 +1782,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void __kmpc_end_serialized_parallel(ident_t *loc, kmp_int32
// global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_end_serialized_parallel");
break;
@@ -1691,7 +1790,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_flush: {
// Build void __kmpc_flush(ident_t *loc);
llvm::Type *TypeParams[] = {getIdentTyPointerTy()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_flush");
break;
@@ -1699,7 +1798,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_master: {
// Build kmp_int32 __kmpc_master(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_master");
break;
@@ -1707,7 +1806,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_end_master: {
// Build void __kmpc_end_master(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_end_master");
break;
@@ -1716,7 +1815,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build kmp_int32 __kmpc_omp_taskyield(ident_t *, kmp_int32 global_tid,
// int end_part);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.IntTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_taskyield");
break;
@@ -1724,7 +1823,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_single: {
// Build kmp_int32 __kmpc_single(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_single");
break;
@@ -1732,7 +1831,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_end_single: {
// Build void __kmpc_end_single(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_end_single");
break;
@@ -1746,7 +1845,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.Int32Ty,
CGM.SizeTy, CGM.SizeTy, KmpRoutineEntryPtrTy};
// Return void * and then cast to particular kmp_task_t type.
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidPtrTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_task_alloc");
break;
@@ -1756,7 +1855,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// *new_task);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_task");
break;
@@ -1771,7 +1870,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.SizeTy,
CGM.VoidPtrTy, CpyFnTy->getPointerTo(),
CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_copyprivate");
break;
@@ -1787,7 +1886,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
getIdentTyPointerTy(), CGM.Int32Ty, CGM.Int32Ty, CGM.SizeTy,
CGM.VoidPtrTy, ReduceFnTy->getPointerTo(),
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_reduce");
break;
@@ -1804,7 +1903,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
getIdentTyPointerTy(), CGM.Int32Ty, CGM.Int32Ty, CGM.SizeTy,
CGM.VoidPtrTy, ReduceFnTy->getPointerTo(),
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_reduce_nowait");
break;
@@ -1815,7 +1914,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), CGM.Int32Ty,
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_end_reduce");
break;
@@ -1826,7 +1925,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), CGM.Int32Ty,
llvm::PointerType::getUnqual(KmpCriticalNameTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn =
CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_end_reduce_nowait");
@@ -1837,7 +1936,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// *new_task);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn =
CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_task_begin_if0");
@@ -1848,7 +1947,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// *new_task);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy,
/*Name=*/"__kmpc_omp_task_complete_if0");
@@ -1857,7 +1956,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_ordered: {
// Build void __kmpc_ordered(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_ordered");
break;
@@ -1865,7 +1964,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_end_ordered: {
// Build void __kmpc_end_ordered(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_end_ordered");
break;
@@ -1873,7 +1972,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_omp_taskwait: {
// Build kmp_int32 __kmpc_omp_taskwait(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_omp_taskwait");
break;
@@ -1881,7 +1980,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_taskgroup: {
// Build void __kmpc_taskgroup(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_taskgroup");
break;
@@ -1889,7 +1988,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_end_taskgroup: {
// Build void __kmpc_end_taskgroup(ident_t *loc, kmp_int32 global_tid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_end_taskgroup");
break;
@@ -1898,7 +1997,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void __kmpc_push_proc_bind(ident_t *loc, kmp_int32 global_tid,
// int proc_bind)
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.IntTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_push_proc_bind");
break;
@@ -1910,7 +2009,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), CGM.Int32Ty, CGM.VoidPtrTy, CGM.Int32Ty,
CGM.VoidPtrTy, CGM.Int32Ty, CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg=*/false);
RTLFn =
CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_task_with_deps");
@@ -1923,7 +2022,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.Int32Ty, CGM.VoidPtrTy,
CGM.Int32Ty, CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_omp_wait_deps");
break;
@@ -1932,7 +2031,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build kmp_int32 __kmpc_cancellationpoint(ident_t *loc, kmp_int32
// global_tid, kmp_int32 cncl_kind)
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.IntTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_cancellationpoint");
break;
@@ -1941,7 +2040,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build kmp_int32 __kmpc_cancel(ident_t *loc, kmp_int32 global_tid,
// kmp_int32 cncl_kind)
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.IntTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_cancel");
break;
@@ -1951,7 +2050,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// kmp_int32 num_teams, kmp_int32 num_threads)
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty, CGM.Int32Ty,
CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_push_num_teams");
break;
@@ -1961,7 +2060,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// microtask, ...);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
getKmpc_MicroPointerTy()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ true);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__kmpc_fork_teams");
break;
@@ -1981,7 +2080,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.IntTy,
CGM.Int64Ty,
CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_taskloop");
break;
@@ -1993,7 +2092,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.Int32Ty,
CGM.Int32Ty,
CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_init");
break;
@@ -2001,7 +2100,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
case OMPRTL__kmpc_doacross_fini: {
// Build void __kmpc_doacross_fini(ident_t *loc, kmp_int32 gtid);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_fini");
break;
@@ -2011,7 +2110,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// *vec);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_post");
break;
@@ -2021,7 +2120,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// *vec);
llvm::Type *TypeParams[] = {getIdentTyPointerTy(), CGM.Int32Ty,
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_doacross_wait");
break;
@@ -2030,7 +2129,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void *__kmpc_task_reduction_init(int gtid, int num_data, void
// *data);
llvm::Type *TypeParams[] = {CGM.IntTy, CGM.IntTy, CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidPtrTy, TypeParams, /*isVarArg=*/false);
RTLFn =
CGM.CreateRuntimeFunction(FnTy, /*Name=*/"__kmpc_task_reduction_init");
@@ -2040,7 +2139,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
// Build void *__kmpc_task_reduction_get_th_data(int gtid, void *tg, void
// *d);
llvm::Type *TypeParams[] = {CGM.IntTy, CGM.VoidPtrTy, CGM.VoidPtrTy};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidPtrTy, TypeParams, /*isVarArg=*/false);
RTLFn = CGM.CreateRuntimeFunction(
FnTy, /*Name=*/"__kmpc_task_reduction_get_th_data");
@@ -2057,7 +2156,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.VoidPtrPtrTy,
CGM.SizeTy->getPointerTo(),
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target");
break;
@@ -2073,7 +2172,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.VoidPtrPtrTy,
CGM.SizeTy->getPointerTo(),
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_nowait");
break;
@@ -2091,7 +2190,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.Int64Ty->getPointerTo(),
CGM.Int32Ty,
CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_teams");
break;
@@ -2109,7 +2208,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.Int64Ty->getPointerTo(),
CGM.Int32Ty,
CGM.Int32Ty};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_teams_nowait");
break;
@@ -2119,7 +2218,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
QualType ParamTy =
CGM.getContext().getPointerType(getTgtBinaryDescriptorQTy());
llvm::Type *TypeParams[] = {CGM.getTypes().ConvertTypeForMem(ParamTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_register_lib");
break;
@@ -2129,7 +2228,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
QualType ParamTy =
CGM.getContext().getPointerType(getTgtBinaryDescriptorQTy());
llvm::Type *TypeParams[] = {CGM.getTypes().ConvertTypeForMem(ParamTy)};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_unregister_lib");
break;
@@ -2143,7 +2242,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.VoidPtrPtrTy,
CGM.SizeTy->getPointerTo(),
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_begin");
break;
@@ -2172,7 +2271,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.VoidPtrPtrTy,
CGM.SizeTy->getPointerTo(),
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_end");
break;
@@ -2201,7 +2300,7 @@ CGOpenMPRuntime::createRuntimeFunction(unsigned Function) {
CGM.VoidPtrPtrTy,
CGM.SizeTy->getPointerTo(),
CGM.Int64Ty->getPointerTo()};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
RTLFn = CGM.CreateRuntimeFunction(FnTy, "__tgt_target_data_update");
break;
@@ -2230,12 +2329,12 @@ llvm::Constant *CGOpenMPRuntime::createForStaticInitFunction(unsigned IVSize,
bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&
"IV size is not compatible with the omp runtime");
- auto Name = IVSize == 32 ? (IVSigned ? "__kmpc_for_static_init_4"
- : "__kmpc_for_static_init_4u")
- : (IVSigned ? "__kmpc_for_static_init_8"
- : "__kmpc_for_static_init_8u");
- auto ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
- auto PtrTy = llvm::PointerType::getUnqual(ITy);
+ StringRef Name = IVSize == 32 ? (IVSigned ? "__kmpc_for_static_init_4"
+ : "__kmpc_for_static_init_4u")
+ : (IVSigned ? "__kmpc_for_static_init_8"
+ : "__kmpc_for_static_init_8u");
+ llvm::Type *ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
+ auto *PtrTy = llvm::PointerType::getUnqual(ITy);
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), // loc
CGM.Int32Ty, // tid
@@ -2247,7 +2346,7 @@ llvm::Constant *CGOpenMPRuntime::createForStaticInitFunction(unsigned IVSize,
ITy, // incr
ITy // chunk
};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
}
@@ -2256,11 +2355,11 @@ llvm::Constant *CGOpenMPRuntime::createDispatchInitFunction(unsigned IVSize,
bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&
"IV size is not compatible with the omp runtime");
- auto Name =
+ StringRef Name =
IVSize == 32
? (IVSigned ? "__kmpc_dispatch_init_4" : "__kmpc_dispatch_init_4u")
: (IVSigned ? "__kmpc_dispatch_init_8" : "__kmpc_dispatch_init_8u");
- auto ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
+ llvm::Type *ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
llvm::Type *TypeParams[] = { getIdentTyPointerTy(), // loc
CGM.Int32Ty, // tid
CGM.Int32Ty, // schedtype
@@ -2269,7 +2368,7 @@ llvm::Constant *CGOpenMPRuntime::createDispatchInitFunction(unsigned IVSize,
ITy, // stride
ITy // chunk
};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
}
@@ -2278,7 +2377,7 @@ llvm::Constant *CGOpenMPRuntime::createDispatchFiniFunction(unsigned IVSize,
bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&
"IV size is not compatible with the omp runtime");
- auto Name =
+ StringRef Name =
IVSize == 32
? (IVSigned ? "__kmpc_dispatch_fini_4" : "__kmpc_dispatch_fini_4u")
: (IVSigned ? "__kmpc_dispatch_fini_8" : "__kmpc_dispatch_fini_8u");
@@ -2286,7 +2385,7 @@ llvm::Constant *CGOpenMPRuntime::createDispatchFiniFunction(unsigned IVSize,
getIdentTyPointerTy(), // loc
CGM.Int32Ty, // tid
};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
return CGM.CreateRuntimeFunction(FnTy, Name);
}
@@ -2295,12 +2394,12 @@ llvm::Constant *CGOpenMPRuntime::createDispatchNextFunction(unsigned IVSize,
bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&
"IV size is not compatible with the omp runtime");
- auto Name =
+ StringRef Name =
IVSize == 32
? (IVSigned ? "__kmpc_dispatch_next_4" : "__kmpc_dispatch_next_4u")
: (IVSigned ? "__kmpc_dispatch_next_8" : "__kmpc_dispatch_next_8u");
- auto ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
- auto PtrTy = llvm::PointerType::getUnqual(ITy);
+ llvm::Type *ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
+ auto *PtrTy = llvm::PointerType::getUnqual(ITy);
llvm::Type *TypeParams[] = {
getIdentTyPointerTy(), // loc
CGM.Int32Ty, // tid
@@ -2309,18 +2408,48 @@ llvm::Constant *CGOpenMPRuntime::createDispatchNextFunction(unsigned IVSize,
PtrTy, // p_upper
PtrTy // p_stride
};
- llvm::FunctionType *FnTy =
+ auto *FnTy =
llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
}
+Address CGOpenMPRuntime::getAddrOfDeclareTargetLink(const VarDecl *VD) {
+ if (CGM.getLangOpts().OpenMPSimd)
+ return Address::invalid();
+ llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(VD);
+ if (Res && *Res == OMPDeclareTargetDeclAttr::MT_Link) {
+ SmallString<64> PtrName;
+ {
+ llvm::raw_svector_ostream OS(PtrName);
+ OS << CGM.getMangledName(GlobalDecl(VD)) << "_decl_tgt_link_ptr";
+ }
+ llvm::Value *Ptr = CGM.getModule().getNamedValue(PtrName);
+ if (!Ptr) {
+ QualType PtrTy = CGM.getContext().getPointerType(VD->getType());
+ Ptr = getOrCreateInternalVariable(CGM.getTypes().ConvertTypeForMem(PtrTy),
+ PtrName);
+ if (!CGM.getLangOpts().OpenMPIsDevice) {
+ auto *GV = cast<llvm::GlobalVariable>(Ptr);
+ GV->setLinkage(llvm::GlobalValue::ExternalLinkage);
+ GV->setInitializer(CGM.GetAddrOfGlobal(VD));
+ }
+ CGM.addUsedGlobal(cast<llvm::GlobalValue>(Ptr));
+ registerTargetGlobalVariable(VD, cast<llvm::Constant>(Ptr));
+ }
+ return Address(Ptr, CGM.getContext().getDeclAlign(VD));
+ }
+ return Address::invalid();
+}
+
llvm::Constant *
CGOpenMPRuntime::getOrCreateThreadPrivateCache(const VarDecl *VD) {
assert(!CGM.getLangOpts().OpenMPUseTLS ||
!CGM.getContext().getTargetInfo().isTLSSupported());
// Lookup the entry, lazily creating it if necessary.
- return getOrCreateInternalVariable(CGM.Int8PtrPtrTy,
- Twine(CGM.getMangledName(VD)) + ".cache.");
+ std::string Suffix = getName({"cache", ""});
+ return getOrCreateInternalVariable(
+ CGM.Int8PtrPtrTy, Twine(CGM.getMangledName(VD)).concat(Suffix));
}
Address CGOpenMPRuntime::getAddrOfThreadPrivate(CodeGenFunction &CGF,
@@ -2331,7 +2460,7 @@ Address CGOpenMPRuntime::getAddrOfThreadPrivate(CodeGenFunction &CGF,
CGM.getContext().getTargetInfo().isTLSSupported())
return VDAddr;
- auto VarTy = VDAddr.getElementType();
+ llvm::Type *VarTy = VDAddr.getElementType();
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
CGF.Builder.CreatePointerCast(VDAddr.getPointer(),
CGM.Int8PtrTy),
@@ -2347,15 +2476,14 @@ void CGOpenMPRuntime::emitThreadPrivateVarInit(
llvm::Value *CopyCtor, llvm::Value *Dtor, SourceLocation Loc) {
// Call kmp_int32 __kmpc_global_thread_num(&loc) to init OpenMP runtime
// library.
- auto OMPLoc = emitUpdateLocation(CGF, Loc);
+ llvm::Value *OMPLoc = emitUpdateLocation(CGF, Loc);
CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_global_thread_num),
OMPLoc);
// Call __kmpc_threadprivate_register(&loc, &var, ctor, cctor/*NULL*/, dtor)
// to register constructor/destructor for variable.
- llvm::Value *Args[] = {OMPLoc,
- CGF.Builder.CreatePointerCast(VDAddr.getPointer(),
- CGM.VoidPtrTy),
- Ctor, CopyCtor, Dtor};
+ llvm::Value *Args[] = {
+ OMPLoc, CGF.Builder.CreatePointerCast(VDAddr.getPointer(), CGM.VoidPtrTy),
+ Ctor, CopyCtor, Dtor};
CGF.EmitRuntimeCall(
createRuntimeFunction(OMPRTL__kmpc_threadprivate_register), Args);
}
@@ -2373,29 +2501,31 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
QualType ASTTy = VD->getType();
llvm::Value *Ctor = nullptr, *CopyCtor = nullptr, *Dtor = nullptr;
- auto Init = VD->getAnyInitializer();
+ const Expr *Init = VD->getAnyInitializer();
if (CGM.getLangOpts().CPlusPlus && PerformInit) {
// Generate function that re-emits the declaration's initializer into the
// threadprivate copy of the variable VD
CodeGenFunction CtorCGF(CGM);
FunctionArgList Args;
- ImplicitParamDecl Dst(CGM.getContext(), CGM.getContext().VoidPtrTy,
+ ImplicitParamDecl Dst(CGM.getContext(), /*DC=*/nullptr, Loc,
+ /*Id=*/nullptr, CGM.getContext().VoidPtrTy,
ImplicitParamDecl::Other);
Args.push_back(&Dst);
- auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
+ const auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
CGM.getContext().VoidPtrTy, Args);
- auto FTy = CGM.getTypes().GetFunctionType(FI);
- auto Fn = CGM.CreateGlobalInitOrDestructFunction(
- FTy, ".__kmpc_global_ctor_.", FI, Loc);
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ std::string Name = getName({"__kmpc_global_ctor_", ""});
+ llvm::Function *Fn =
+ CGM.CreateGlobalInitOrDestructFunction(FTy, Name, FI, Loc);
CtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidPtrTy, Fn, FI,
- Args, SourceLocation());
- auto ArgVal = CtorCGF.EmitLoadOfScalar(
+ Args, Loc, Loc);
+ llvm::Value *ArgVal = CtorCGF.EmitLoadOfScalar(
CtorCGF.GetAddrOfLocalVar(&Dst), /*Volatile=*/false,
CGM.getContext().VoidPtrTy, Dst.getLocation());
Address Arg = Address(ArgVal, VDAddr.getAlignment());
- Arg = CtorCGF.Builder.CreateElementBitCast(Arg,
- CtorCGF.ConvertTypeForMem(ASTTy));
+ Arg = CtorCGF.Builder.CreateElementBitCast(
+ Arg, CtorCGF.ConvertTypeForMem(ASTTy));
CtorCGF.EmitAnyExprToMem(Init, Arg, Init->getType().getQualifiers(),
/*IsInitializer=*/true);
ArgVal = CtorCGF.EmitLoadOfScalar(
@@ -2410,21 +2540,23 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
// of the variable VD
CodeGenFunction DtorCGF(CGM);
FunctionArgList Args;
- ImplicitParamDecl Dst(CGM.getContext(), CGM.getContext().VoidPtrTy,
+ ImplicitParamDecl Dst(CGM.getContext(), /*DC=*/nullptr, Loc,
+ /*Id=*/nullptr, CGM.getContext().VoidPtrTy,
ImplicitParamDecl::Other);
Args.push_back(&Dst);
- auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
+ const auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
CGM.getContext().VoidTy, Args);
- auto FTy = CGM.getTypes().GetFunctionType(FI);
- auto Fn = CGM.CreateGlobalInitOrDestructFunction(
- FTy, ".__kmpc_global_dtor_.", FI, Loc);
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ std::string Name = getName({"__kmpc_global_dtor_", ""});
+ llvm::Function *Fn =
+ CGM.CreateGlobalInitOrDestructFunction(FTy, Name, FI, Loc);
auto NL = ApplyDebugLocation::CreateEmpty(DtorCGF);
DtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, Fn, FI, Args,
- SourceLocation());
+ Loc, Loc);
// Create a scope with an artificial location for the body of this function.
auto AL = ApplyDebugLocation::CreateArtificial(DtorCGF);
- auto ArgVal = DtorCGF.EmitLoadOfScalar(
+ llvm::Value *ArgVal = DtorCGF.EmitLoadOfScalar(
DtorCGF.GetAddrOfLocalVar(&Dst),
/*Volatile=*/false, CGM.getContext().VoidPtrTy, Dst.getLocation());
DtorCGF.emitDestroy(Address(ArgVal, VDAddr.getAlignment()), ASTTy,
@@ -2438,34 +2570,36 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
return nullptr;
llvm::Type *CopyCtorTyArgs[] = {CGM.VoidPtrTy, CGM.VoidPtrTy};
- auto CopyCtorTy =
- llvm::FunctionType::get(CGM.VoidPtrTy, CopyCtorTyArgs,
- /*isVarArg=*/false)->getPointerTo();
+ auto *CopyCtorTy = llvm::FunctionType::get(CGM.VoidPtrTy, CopyCtorTyArgs,
+ /*isVarArg=*/false)
+ ->getPointerTo();
// Copying constructor for the threadprivate variable.
// Must be NULL - reserved by runtime, but currently it requires that this
// parameter is always NULL. Otherwise it fires assertion.
CopyCtor = llvm::Constant::getNullValue(CopyCtorTy);
if (Ctor == nullptr) {
- auto CtorTy = llvm::FunctionType::get(CGM.VoidPtrTy, CGM.VoidPtrTy,
- /*isVarArg=*/false)->getPointerTo();
+ auto *CtorTy = llvm::FunctionType::get(CGM.VoidPtrTy, CGM.VoidPtrTy,
+ /*isVarArg=*/false)
+ ->getPointerTo();
Ctor = llvm::Constant::getNullValue(CtorTy);
}
if (Dtor == nullptr) {
- auto DtorTy = llvm::FunctionType::get(CGM.VoidTy, CGM.VoidPtrTy,
- /*isVarArg=*/false)->getPointerTo();
+ auto *DtorTy = llvm::FunctionType::get(CGM.VoidTy, CGM.VoidPtrTy,
+ /*isVarArg=*/false)
+ ->getPointerTo();
Dtor = llvm::Constant::getNullValue(DtorTy);
}
if (!CGF) {
- auto InitFunctionTy =
+ auto *InitFunctionTy =
llvm::FunctionType::get(CGM.VoidTy, /*isVarArg*/ false);
- auto InitFunction = CGM.CreateGlobalInitOrDestructFunction(
- InitFunctionTy, ".__omp_threadprivate_init_.",
- CGM.getTypes().arrangeNullaryFunction());
+ std::string Name = getName({"__omp_threadprivate_init_", ""});
+ llvm::Function *InitFunction = CGM.CreateGlobalInitOrDestructFunction(
+ InitFunctionTy, Name, CGM.getTypes().arrangeNullaryFunction());
CodeGenFunction InitCGF(CGM);
FunctionArgList ArgList;
InitCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, InitFunction,
CGM.getTypes().arrangeNullaryFunction(), ArgList,
- Loc);
+ Loc, Loc);
emitThreadPrivateVarInit(InitCGF, VDAddr, Ctor, CopyCtor, Dtor, Loc);
InitCGF.FinishFunction();
return InitFunction;
@@ -2475,19 +2609,156 @@ llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
return nullptr;
}
+/// Obtain information that uniquely identifies a target entry. This
+/// consists of the file and device IDs as well as line number associated with
+/// the relevant entry source location.
+static void getTargetEntryUniqueInfo(ASTContext &C, SourceLocation Loc,
+ unsigned &DeviceID, unsigned &FileID,
+ unsigned &LineNum) {
+ SourceManager &SM = C.getSourceManager();
+
+ // The loc should be always valid and have a file ID (the user cannot use
+ // #pragma directives in macros)
+
+ assert(Loc.isValid() && "Source location is expected to be always valid.");
+
+ PresumedLoc PLoc = SM.getPresumedLoc(Loc);
+ assert(PLoc.isValid() && "Source location is expected to be always valid.");
+
+ llvm::sys::fs::UniqueID ID;
+ if (auto EC = llvm::sys::fs::getUniqueID(PLoc.getFilename(), ID))
+ SM.getDiagnostics().Report(diag::err_cannot_open_file)
+ << PLoc.getFilename() << EC.message();
+
+ DeviceID = ID.getDevice();
+ FileID = ID.getFile();
+ LineNum = PLoc.getLine();
+}
+
+bool CGOpenMPRuntime::emitDeclareTargetVarDefinition(const VarDecl *VD,
+ llvm::GlobalVariable *Addr,
+ bool PerformInit) {
+ Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(VD);
+ if (!Res || *Res == OMPDeclareTargetDeclAttr::MT_Link)
+ return false;
+ VD = VD->getDefinition(CGM.getContext());
+ if (VD && !DeclareTargetWithDefinition.insert(VD).second)
+ return CGM.getLangOpts().OpenMPIsDevice;
+
+ QualType ASTTy = VD->getType();
+
+ SourceLocation Loc = VD->getCanonicalDecl()->getLocStart();
+ // Produce the unique prefix to identify the new target regions. We use
+ // the source location of the variable declaration which we know to not
+ // conflict with any target region.
+ unsigned DeviceID;
+ unsigned FileID;
+ unsigned Line;
+ getTargetEntryUniqueInfo(CGM.getContext(), Loc, DeviceID, FileID, Line);
+ SmallString<128> Buffer, Out;
+ {
+ llvm::raw_svector_ostream OS(Buffer);
+ OS << "__omp_offloading_" << llvm::format("_%x", DeviceID)
+ << llvm::format("_%x_", FileID) << VD->getName() << "_l" << Line;
+ }
+
+ const Expr *Init = VD->getAnyInitializer();
+ if (CGM.getLangOpts().CPlusPlus && PerformInit) {
+ llvm::Constant *Ctor;
+ llvm::Constant *ID;
+ if (CGM.getLangOpts().OpenMPIsDevice) {
+ // Generate function that re-emits the declaration's initializer into
+ // the threadprivate copy of the variable VD
+ CodeGenFunction CtorCGF(CGM);
+
+ const CGFunctionInfo &FI = CGM.getTypes().arrangeNullaryFunction();
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ llvm::Function *Fn = CGM.CreateGlobalInitOrDestructFunction(
+ FTy, Twine(Buffer, "_ctor"), FI, Loc);
+ auto NL = ApplyDebugLocation::CreateEmpty(CtorCGF);
+ CtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, Fn, FI,
+ FunctionArgList(), Loc, Loc);
+ auto AL = ApplyDebugLocation::CreateArtificial(CtorCGF);
+ CtorCGF.EmitAnyExprToMem(Init,
+ Address(Addr, CGM.getContext().getDeclAlign(VD)),
+ Init->getType().getQualifiers(),
+ /*IsInitializer=*/true);
+ CtorCGF.FinishFunction();
+ Ctor = Fn;
+ ID = llvm::ConstantExpr::getBitCast(Fn, CGM.Int8PtrTy);
+ CGM.addUsedGlobal(cast<llvm::GlobalValue>(Ctor));
+ } else {
+ Ctor = new llvm::GlobalVariable(
+ CGM.getModule(), CGM.Int8Ty, /*isConstant=*/true,
+ llvm::GlobalValue::PrivateLinkage,
+ llvm::Constant::getNullValue(CGM.Int8Ty), Twine(Buffer, "_ctor"));
+ ID = Ctor;
+ }
+
+ // Register the information for the entry associated with the constructor.
+ Out.clear();
+ OffloadEntriesInfoManager.registerTargetRegionEntryInfo(
+ DeviceID, FileID, Twine(Buffer, "_ctor").toStringRef(Out), Line, Ctor,
+ ID, OffloadEntriesInfoManagerTy::OMPTargetRegionEntryCtor);
+ }
+ if (VD->getType().isDestructedType() != QualType::DK_none) {
+ llvm::Constant *Dtor;
+ llvm::Constant *ID;
+ if (CGM.getLangOpts().OpenMPIsDevice) {
+ // Generate function that emits destructor call for the threadprivate
+ // copy of the variable VD
+ CodeGenFunction DtorCGF(CGM);
+
+ const CGFunctionInfo &FI = CGM.getTypes().arrangeNullaryFunction();
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ llvm::Function *Fn = CGM.CreateGlobalInitOrDestructFunction(
+ FTy, Twine(Buffer, "_dtor"), FI, Loc);
+ auto NL = ApplyDebugLocation::CreateEmpty(DtorCGF);
+ DtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, Fn, FI,
+ FunctionArgList(), Loc, Loc);
+ // Create a scope with an artificial location for the body of this
+ // function.
+ auto AL = ApplyDebugLocation::CreateArtificial(DtorCGF);
+ DtorCGF.emitDestroy(Address(Addr, CGM.getContext().getDeclAlign(VD)),
+ ASTTy, DtorCGF.getDestroyer(ASTTy.isDestructedType()),
+ DtorCGF.needsEHCleanup(ASTTy.isDestructedType()));
+ DtorCGF.FinishFunction();
+ Dtor = Fn;
+ ID = llvm::ConstantExpr::getBitCast(Fn, CGM.Int8PtrTy);
+ CGM.addUsedGlobal(cast<llvm::GlobalValue>(Dtor));
+ } else {
+ Dtor = new llvm::GlobalVariable(
+ CGM.getModule(), CGM.Int8Ty, /*isConstant=*/true,
+ llvm::GlobalValue::PrivateLinkage,
+ llvm::Constant::getNullValue(CGM.Int8Ty), Twine(Buffer, "_dtor"));
+ ID = Dtor;
+ }
+ // Register the information for the entry associated with the destructor.
+ Out.clear();
+ OffloadEntriesInfoManager.registerTargetRegionEntryInfo(
+ DeviceID, FileID, Twine(Buffer, "_dtor").toStringRef(Out), Line, Dtor,
+ ID, OffloadEntriesInfoManagerTy::OMPTargetRegionEntryDtor);
+ }
+ return CGM.getLangOpts().OpenMPIsDevice;
+}
+
Address CGOpenMPRuntime::getAddrOfArtificialThreadPrivate(CodeGenFunction &CGF,
QualType VarType,
StringRef Name) {
- llvm::Twine VarName(Name, ".artificial.");
+ std::string Suffix = getName({"artificial", ""});
+ std::string CacheSuffix = getName({"cache", ""});
llvm::Type *VarLVType = CGF.ConvertTypeForMem(VarType);
- llvm::Value *GAddr = getOrCreateInternalVariable(VarLVType, VarName);
+ llvm::Value *GAddr =
+ getOrCreateInternalVariable(VarLVType, Twine(Name).concat(Suffix));
llvm::Value *Args[] = {
emitUpdateLocation(CGF, SourceLocation()),
getThreadID(CGF, SourceLocation()),
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(GAddr, CGM.VoidPtrTy),
CGF.Builder.CreateIntCast(CGF.getTypeSize(VarType), CGM.SizeTy,
/*IsSigned=*/false),
- getOrCreateInternalVariable(CGM.VoidPtrPtrTy, VarName + ".cache.")};
+ getOrCreateInternalVariable(
+ CGM.VoidPtrPtrTy, Twine(Name).concat(Suffix).concat(CacheSuffix))};
return Address(
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
CGF.EmitRuntimeCall(
@@ -2496,13 +2767,6 @@ Address CGOpenMPRuntime::getAddrOfArtificialThreadPrivate(CodeGenFunction &CGF,
CGM.getPointerAlign());
}
-/// \brief Emits code for OpenMP 'if' clause using specified \a CodeGen
-/// function. Here is the logic:
-/// if (Cond) {
-/// ThenGen();
-/// } else {
-/// ElseGen();
-/// }
void CGOpenMPRuntime::emitOMPIfClause(CodeGenFunction &CGF, const Expr *Cond,
const RegionCodeGenTy &ThenGen,
const RegionCodeGenTy &ElseGen) {
@@ -2521,9 +2785,9 @@ void CGOpenMPRuntime::emitOMPIfClause(CodeGenFunction &CGF, const Expr *Cond,
// Otherwise, the condition did not fold, or we couldn't elide it. Just
// emit the conditional branch.
- auto ThenBlock = CGF.createBasicBlock("omp_if.then");
- auto ElseBlock = CGF.createBasicBlock("omp_if.else");
- auto ContBlock = CGF.createBasicBlock("omp_if.end");
+ llvm::BasicBlock *ThenBlock = CGF.createBasicBlock("omp_if.then");
+ llvm::BasicBlock *ElseBlock = CGF.createBasicBlock("omp_if.else");
+ llvm::BasicBlock *ContBlock = CGF.createBasicBlock("omp_if.end");
CGF.EmitBranchOnBoolExpr(Cond, ThenBlock, ElseBlock, /*TrueCount=*/0);
// Emit the 'then' code.
@@ -2548,11 +2812,11 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
const Expr *IfCond) {
if (!CGF.HaveInsertPoint())
return;
- auto *RTLoc = emitUpdateLocation(CGF, Loc);
+ llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);
auto &&ThenGen = [OutlinedFn, CapturedVars, RTLoc](CodeGenFunction &CGF,
PrePostActionTy &) {
// Build call __kmpc_fork_call(loc, n, microtask, var1, .., varn);
- auto &RT = CGF.CGM.getOpenMPRuntime();
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
llvm::Value *Args[] = {
RTLoc,
CGF.Builder.getInt32(CapturedVars.size()), // Number of captured vars
@@ -2561,13 +2825,13 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
RealArgs.append(std::begin(Args), std::end(Args));
RealArgs.append(CapturedVars.begin(), CapturedVars.end());
- auto RTLFn = RT.createRuntimeFunction(OMPRTL__kmpc_fork_call);
+ llvm::Value *RTLFn = RT.createRuntimeFunction(OMPRTL__kmpc_fork_call);
CGF.EmitRuntimeCall(RTLFn, RealArgs);
};
auto &&ElseGen = [OutlinedFn, CapturedVars, RTLoc, Loc](CodeGenFunction &CGF,
PrePostActionTy &) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
- auto ThreadID = RT.getThreadID(CGF, Loc);
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
+ llvm::Value *ThreadID = RT.getThreadID(CGF, Loc);
// Build calls:
// __kmpc_serialized_parallel(&Loc, GTid);
llvm::Value *Args[] = {RTLoc, ThreadID};
@@ -2575,13 +2839,12 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
RT.createRuntimeFunction(OMPRTL__kmpc_serialized_parallel), Args);
// OutlinedFn(&GTid, &zero, CapturedStruct);
- auto ThreadIDAddr = RT.emitThreadIDAddress(CGF, Loc);
- Address ZeroAddr =
- CGF.CreateTempAlloca(CGF.Int32Ty, CharUnits::fromQuantity(4),
- /*Name*/ ".zero.addr");
+ Address ZeroAddr = CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
+ /*Name*/ ".zero.addr");
CGF.InitTempAlloca(ZeroAddr, CGF.Builder.getInt32(/*C*/ 0));
llvm::SmallVector<llvm::Value *, 16> OutlinedFnArgs;
- OutlinedFnArgs.push_back(ThreadIDAddr.getPointer());
+ // ThreadId for serialized parallels is 0.
+ OutlinedFnArgs.push_back(ZeroAddr.getPointer());
OutlinedFnArgs.push_back(ZeroAddr.getPointer());
OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());
RT.emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);
@@ -2592,9 +2855,9 @@ void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
RT.createRuntimeFunction(OMPRTL__kmpc_end_serialized_parallel),
EndArgs);
};
- if (IfCond)
+ if (IfCond) {
emitOMPIfClause(CGF, IfCond, ThenGen, ElseGen);
- else {
+ } else {
RegionCodeGenTy ThenRCG(ThenGen);
ThenRCG(CGF);
}
@@ -2613,10 +2876,10 @@ Address CGOpenMPRuntime::emitThreadIDAddress(CodeGenFunction &CGF,
if (OMPRegionInfo->getThreadIDVariable())
return OMPRegionInfo->getThreadIDVariableLValue(CGF).getAddress();
- auto ThreadID = getThreadID(CGF, Loc);
- auto Int32Ty =
+ llvm::Value *ThreadID = getThreadID(CGF, Loc);
+ QualType Int32Ty =
CGF.getContext().getIntTypeForBitwidth(/*DestWidth*/ 32, /*Signed*/ true);
- auto ThreadIDTemp = CGF.CreateMemTemp(Int32Ty, /*Name*/ ".threadid_temp.");
+ Address ThreadIDTemp = CGF.CreateMemTemp(Int32Ty, /*Name*/ ".threadid_temp.");
CGF.EmitStoreOfScalar(ThreadID,
CGF.MakeAddrLValue(ThreadIDTemp, Int32Ty));
@@ -2629,8 +2892,8 @@ CGOpenMPRuntime::getOrCreateInternalVariable(llvm::Type *Ty,
SmallString<256> Buffer;
llvm::raw_svector_ostream Out(Buffer);
Out << Name;
- auto RuntimeName = Out.str();
- auto &Elem = *InternalVars.insert(std::make_pair(RuntimeName, nullptr)).first;
+ StringRef RuntimeName = Out.str();
+ auto &Elem = *InternalVars.try_emplace(RuntimeName, nullptr).first;
if (Elem.second) {
assert(Elem.second->getType()->getPointerElementType() == Ty &&
"OMP internal variable has different type than requested");
@@ -2644,8 +2907,9 @@ CGOpenMPRuntime::getOrCreateInternalVariable(llvm::Type *Ty,
}
llvm::Value *CGOpenMPRuntime::getCriticalRegionLock(StringRef CriticalName) {
- llvm::Twine Name(".gomp_critical_user_", CriticalName);
- return getOrCreateInternalVariable(KmpCriticalNameTy, Name.concat(".var"));
+ std::string Prefix = Twine("gomp_critical_user_", CriticalName).str();
+ std::string Name = getName({Prefix, "var"});
+ return getOrCreateInternalVariable(KmpCriticalNameTy, Name);
}
namespace {
@@ -2779,21 +3043,28 @@ static Address emitAddrOfVarFromArray(CodeGenFunction &CGF, Address Array,
static llvm::Value *emitCopyprivateCopyFunction(
CodeGenModule &CGM, llvm::Type *ArgsType,
ArrayRef<const Expr *> CopyprivateVars, ArrayRef<const Expr *> DestExprs,
- ArrayRef<const Expr *> SrcExprs, ArrayRef<const Expr *> AssignmentOps) {
- auto &C = CGM.getContext();
+ ArrayRef<const Expr *> SrcExprs, ArrayRef<const Expr *> AssignmentOps,
+ SourceLocation Loc) {
+ ASTContext &C = CGM.getContext();
// void copy_func(void *LHSArg, void *RHSArg);
FunctionArgList Args;
- ImplicitParamDecl LHSArg(C, C.VoidPtrTy, ImplicitParamDecl::Other);
- ImplicitParamDecl RHSArg(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl LHSArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
+ ImplicitParamDecl RHSArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
Args.push_back(&LHSArg);
Args.push_back(&RHSArg);
- auto &CGFI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *Fn = llvm::Function::Create(
- CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
- ".omp.copyprivate.copy_func", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, CGFI);
+ const auto &CGFI =
+ CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
+ std::string Name =
+ CGM.getOpenMPRuntime().getName({"omp", "copyprivate", "copy_func"});
+ auto *Fn = llvm::Function::Create(CGM.getTypes().GetFunctionType(CGFI),
+ llvm::GlobalValue::InternalLinkage, Name,
+ &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+ Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
// Dest = (void*[n])(LHSArg);
// Src = (void*[n])(RHSArg);
Address LHS(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
@@ -2807,13 +3078,15 @@ static llvm::Value *emitCopyprivateCopyFunction(
// ...
// *(Typen*)Dst[n] = *(Typen*)Src[n];
for (unsigned I = 0, E = AssignmentOps.size(); I < E; ++I) {
- auto DestVar = cast<VarDecl>(cast<DeclRefExpr>(DestExprs[I])->getDecl());
+ const auto *DestVar =
+ cast<VarDecl>(cast<DeclRefExpr>(DestExprs[I])->getDecl());
Address DestAddr = emitAddrOfVarFromArray(CGF, LHS, I, DestVar);
- auto SrcVar = cast<VarDecl>(cast<DeclRefExpr>(SrcExprs[I])->getDecl());
+ const auto *SrcVar =
+ cast<VarDecl>(cast<DeclRefExpr>(SrcExprs[I])->getDecl());
Address SrcAddr = emitAddrOfVarFromArray(CGF, RHS, I, SrcVar);
- auto *VD = cast<DeclRefExpr>(CopyprivateVars[I])->getDecl();
+ const auto *VD = cast<DeclRefExpr>(CopyprivateVars[I])->getDecl();
QualType Type = VD->getType();
CGF.EmitOMPCopy(Type, DestAddr, SrcAddr, DestVar, SrcVar, AssignmentOps[I]);
}
@@ -2833,7 +3106,7 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
assert(CopyprivateVars.size() == SrcExprs.size() &&
CopyprivateVars.size() == DstExprs.size() &&
CopyprivateVars.size() == AssignmentOps.size());
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
// int32 did_it = 0;
// if(__kmpc_single(ident_t *, gtid)) {
// SingleOpGen();
@@ -2846,7 +3119,8 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
Address DidIt = Address::invalid();
if (!CopyprivateVars.empty()) {
// int32 did_it = 0;
- auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
+ QualType KmpInt32Ty =
+ C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
DidIt = CGF.CreateMemTemp(KmpInt32Ty, ".omp.copyprivate.did_it");
CGF.Builder.CreateStore(CGF.Builder.getInt32(0), DidIt);
}
@@ -2866,7 +3140,7 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
// <copy_func>, did_it);
if (DidIt.isValid()) {
llvm::APInt ArraySize(/*unsigned int numBits=*/32, CopyprivateVars.size());
- auto CopyprivateArrayTy =
+ QualType CopyprivateArrayTy =
C.getConstantArrayType(C.VoidPtrTy, ArraySize, ArrayType::Normal,
/*IndexTypeQuals=*/0);
// Create a list of all private variables for copyprivate.
@@ -2882,14 +3156,14 @@ void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
}
// Build function that copies private values from single region to all other
// threads in the corresponding parallel region.
- auto *CpyFn = emitCopyprivateCopyFunction(
+ llvm::Value *CpyFn = emitCopyprivateCopyFunction(
CGM, CGF.ConvertTypeForMem(CopyprivateArrayTy)->getPointerTo(),
- CopyprivateVars, SrcExprs, DstExprs, AssignmentOps);
- auto *BufSize = CGF.getTypeSize(CopyprivateArrayTy);
+ CopyprivateVars, SrcExprs, DstExprs, AssignmentOps, Loc);
+ llvm::Value *BufSize = CGF.getTypeSize(CopyprivateArrayTy);
Address CL =
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(CopyprivateList,
CGF.VoidPtrTy);
- auto *DidItVal = CGF.Builder.CreateLoad(DidIt);
+ llvm::Value *DidItVal = CGF.Builder.CreateLoad(DidIt);
llvm::Value *Args[] = {
emitUpdateLocation(CGF, Loc), // ident_t *<loc>
getThreadID(CGF, Loc), // i32 <gtid>
@@ -2948,19 +3222,19 @@ void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
if (auto *OMPRegionInfo =
dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo)) {
if (!ForceSimpleCall && OMPRegionInfo->hasCancel()) {
- auto *Result = CGF.EmitRuntimeCall(
+ llvm::Value *Result = CGF.EmitRuntimeCall(
createRuntimeFunction(OMPRTL__kmpc_cancel_barrier), Args);
if (EmitChecks) {
// if (__kmpc_cancel_barrier()) {
// exit from construct;
// }
- auto *ExitBB = CGF.createBasicBlock(".cancel.exit");
- auto *ContBB = CGF.createBasicBlock(".cancel.continue");
- auto *Cmp = CGF.Builder.CreateIsNotNull(Result);
+ llvm::BasicBlock *ExitBB = CGF.createBasicBlock(".cancel.exit");
+ llvm::BasicBlock *ContBB = CGF.createBasicBlock(".cancel.continue");
+ llvm::Value *Cmp = CGF.Builder.CreateIsNotNull(Result);
CGF.Builder.CreateCondBr(Cmp, ExitBB, ContBB);
CGF.EmitBlock(ExitBB);
// exit from construct;
- auto CancelDestination =
+ CodeGenFunction::JumpDest CancelDestination =
CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
CGF.EmitBranchThroughCleanup(CancelDestination);
CGF.EmitBlock(ContBB, /*IsFinished=*/true);
@@ -2971,7 +3245,7 @@ void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_barrier), Args);
}
-/// \brief Map the OpenMP loop schedule to the runtime enumeration.
+/// Map the OpenMP loop schedule to the runtime enumeration.
static OpenMPSchedType getRuntimeSchedule(OpenMPScheduleClauseKind ScheduleKind,
bool Chunked, bool Ordered) {
switch (ScheduleKind) {
@@ -2993,7 +3267,7 @@ static OpenMPSchedType getRuntimeSchedule(OpenMPScheduleClauseKind ScheduleKind,
llvm_unreachable("Unexpected runtime schedule");
}
-/// \brief Map the OpenMP distribute schedule to the runtime enumeration.
+/// Map the OpenMP distribute schedule to the runtime enumeration.
static OpenMPSchedType
getRuntimeSchedule(OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) {
// only static is allowed for dist_schedule
@@ -3002,19 +3276,20 @@ getRuntimeSchedule(OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) {
bool CGOpenMPRuntime::isStaticNonchunked(OpenMPScheduleClauseKind ScheduleKind,
bool Chunked) const {
- auto Schedule = getRuntimeSchedule(ScheduleKind, Chunked, /*Ordered=*/false);
+ OpenMPSchedType Schedule =
+ getRuntimeSchedule(ScheduleKind, Chunked, /*Ordered=*/false);
return Schedule == OMP_sch_static;
}
bool CGOpenMPRuntime::isStaticNonchunked(
OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) const {
- auto Schedule = getRuntimeSchedule(ScheduleKind, Chunked);
+ OpenMPSchedType Schedule = getRuntimeSchedule(ScheduleKind, Chunked);
return Schedule == OMP_dist_sch_static;
}
bool CGOpenMPRuntime::isDynamic(OpenMPScheduleClauseKind ScheduleKind) const {
- auto Schedule =
+ OpenMPSchedType Schedule =
getRuntimeSchedule(ScheduleKind, /*Chunked=*/false, /*Ordered=*/false);
assert(Schedule != OMP_sch_static_chunked && "cannot be chunked here");
return Schedule != OMP_sch_static;
@@ -3147,12 +3422,12 @@ void CGOpenMPRuntime::emitForStaticInit(CodeGenFunction &CGF,
ScheduleKind.Schedule, Values.Chunk != nullptr, Values.Ordered);
assert(isOpenMPWorksharingDirective(DKind) &&
"Expected loop-based or sections-based directive.");
- auto *UpdatedLocation = emitUpdateLocation(CGF, Loc,
+ llvm::Value *UpdatedLocation = emitUpdateLocation(CGF, Loc,
isOpenMPLoopDirective(DKind)
? OMP_IDENT_WORK_LOOP
: OMP_IDENT_WORK_SECTIONS);
- auto *ThreadId = getThreadID(CGF, Loc);
- auto *StaticInitFunction =
+ llvm::Value *ThreadId = getThreadID(CGF, Loc);
+ llvm::Constant *StaticInitFunction =
createForStaticInitFunction(Values.IVSize, Values.IVSigned);
emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
ScheduleNum, ScheduleKind.M1, ScheduleKind.M2, Values);
@@ -3164,10 +3439,10 @@ void CGOpenMPRuntime::emitDistributeStaticInit(
const CGOpenMPRuntime::StaticRTInput &Values) {
OpenMPSchedType ScheduleNum =
getRuntimeSchedule(SchedKind, Values.Chunk != nullptr);
- auto *UpdatedLocation =
+ llvm::Value *UpdatedLocation =
emitUpdateLocation(CGF, Loc, OMP_IDENT_WORK_DISTRIBUTE);
- auto *ThreadId = getThreadID(CGF, Loc);
- auto *StaticInitFunction =
+ llvm::Value *ThreadId = getThreadID(CGF, Loc);
+ llvm::Constant *StaticInitFunction =
createForStaticInitFunction(Values.IVSize, Values.IVSigned);
emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
ScheduleNum, OMPC_SCHEDULE_MODIFIER_unknown,
@@ -3223,7 +3498,7 @@ llvm::Value *CGOpenMPRuntime::emitForNext(CodeGenFunction &CGF,
llvm::Value *Call =
CGF.EmitRuntimeCall(createDispatchNextFunction(IVSize, IVSigned), Args);
return CGF.EmitScalarConversion(
- Call, CGF.getContext().getIntTypeForBitwidth(32, /* Signed */ true),
+ Call, CGF.getContext().getIntTypeForBitwidth(32, /*Signed=*/1),
CGF.getContext().BoolTy, Loc);
}
@@ -3285,13 +3560,13 @@ void CGOpenMPRuntime::emitFlush(CodeGenFunction &CGF, ArrayRef<const Expr *>,
}
namespace {
-/// \brief Indexes of fields for type kmp_task_t.
+/// Indexes of fields for type kmp_task_t.
enum KmpTaskTFields {
- /// \brief List of shared variables.
+ /// List of shared variables.
KmpTaskTShareds,
- /// \brief Task routine.
+ /// Task routine.
KmpTaskTRoutine,
- /// \brief Partition id for the untied tasks.
+ /// Partition id for the untied tasks.
KmpTaskTPartId,
/// Function with call of destructors for private variables.
Data1,
@@ -3311,11 +3586,11 @@ enum KmpTaskTFields {
} // anonymous namespace
bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::empty() const {
- // FIXME: Add other entries type when they become supported.
- return OffloadEntriesTargetRegion.empty();
+ return OffloadEntriesTargetRegion.empty() &&
+ OffloadEntriesDeviceGlobalVar.empty();
}
-/// \brief Initialize target region entry.
+/// Initialize target region entry.
void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
initializeTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
StringRef ParentName, unsigned LineNum,
@@ -3325,7 +3600,7 @@ void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
"code generation.");
OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] =
OffloadEntryInfoTargetRegion(Order, /*Addr=*/nullptr, /*ID=*/nullptr,
- /*Flags=*/0);
+ OMPTargetRegionEntryTargetRegion);
++OffloadingEntriesNum;
}
@@ -3333,22 +3608,27 @@ void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
registerTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
StringRef ParentName, unsigned LineNum,
llvm::Constant *Addr, llvm::Constant *ID,
- int32_t Flags) {
+ OMPTargetRegionEntryKind Flags) {
// If we are emitting code for a target, the entry is already initialized,
// only has to be registered.
if (CGM.getLangOpts().OpenMPIsDevice) {
- assert(hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum) &&
- "Entry must exist.");
+ if (!hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum)) {
+ unsigned DiagID = CGM.getDiags().getCustomDiagID(
+ DiagnosticsEngine::Error,
+ "Unable to find target region on line '%0' in the device code.");
+ CGM.getDiags().Report(DiagID) << LineNum;
+ return;
+ }
auto &Entry =
OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum];
assert(Entry.isValid() && "Entry not initialized!");
Entry.setAddress(Addr);
Entry.setID(ID);
Entry.setFlags(Flags);
- return;
} else {
- OffloadEntryInfoTargetRegion Entry(OffloadingEntriesNum++, Addr, ID, Flags);
+ OffloadEntryInfoTargetRegion Entry(OffloadingEntriesNum, Addr, ID, Flags);
OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] = Entry;
+ ++OffloadingEntriesNum;
}
}
@@ -3376,48 +3656,69 @@ bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::hasTargetRegionEntryInfo(
void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::actOnTargetRegionEntriesInfo(
const OffloadTargetRegionEntryInfoActTy &Action) {
// Scan all target region entries and perform the provided action.
- for (auto &D : OffloadEntriesTargetRegion)
- for (auto &F : D.second)
- for (auto &P : F.second)
- for (auto &L : P.second)
+ for (const auto &D : OffloadEntriesTargetRegion)
+ for (const auto &F : D.second)
+ for (const auto &P : F.second)
+ for (const auto &L : P.second)
Action(D.first, F.first, P.first(), L.first, L.second);
}
-/// \brief Create a Ctor/Dtor-like function whose body is emitted through
-/// \a Codegen. This is used to emit the two functions that register and
-/// unregister the descriptor of the current compilation unit.
-static llvm::Function *
-createOffloadingBinaryDescriptorFunction(CodeGenModule &CGM, StringRef Name,
- const RegionCodeGenTy &Codegen) {
- auto &C = CGM.getContext();
- FunctionArgList Args;
- ImplicitParamDecl DummyPtr(C, C.VoidPtrTy, ImplicitParamDecl::Other);
- Args.push_back(&DummyPtr);
+void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
+ initializeDeviceGlobalVarEntryInfo(StringRef Name,
+ OMPTargetGlobalVarEntryKind Flags,
+ unsigned Order) {
+ assert(CGM.getLangOpts().OpenMPIsDevice && "Initialization of entries is "
+ "only required for the device "
+ "code generation.");
+ OffloadEntriesDeviceGlobalVar.try_emplace(Name, Order, Flags);
+ ++OffloadingEntriesNum;
+}
- CodeGenFunction CGF(CGM);
- auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto FTy = CGM.getTypes().GetFunctionType(FI);
- auto *Fn =
- CGM.CreateGlobalInitOrDestructFunction(FTy, Name, FI, SourceLocation());
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FI, Args, SourceLocation());
- Codegen(CGF);
- CGF.FinishFunction();
- return Fn;
+void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
+ registerDeviceGlobalVarEntryInfo(StringRef VarName, llvm::Constant *Addr,
+ CharUnits VarSize,
+ OMPTargetGlobalVarEntryKind Flags,
+ llvm::GlobalValue::LinkageTypes Linkage) {
+ if (CGM.getLangOpts().OpenMPIsDevice) {
+ auto &Entry = OffloadEntriesDeviceGlobalVar[VarName];
+ assert(Entry.isValid() && Entry.getFlags() == Flags &&
+ "Entry not initialized!");
+ assert((!Entry.getAddress() || Entry.getAddress() == Addr) &&
+ "Resetting with the new address.");
+ if (Entry.getAddress() && hasDeviceGlobalVarEntryInfo(VarName))
+ return;
+ Entry.setAddress(Addr);
+ Entry.setVarSize(VarSize);
+ Entry.setLinkage(Linkage);
+ } else {
+ if (hasDeviceGlobalVarEntryInfo(VarName))
+ return;
+ OffloadEntriesDeviceGlobalVar.try_emplace(
+ VarName, OffloadingEntriesNum, Addr, VarSize, Flags, Linkage);
+ ++OffloadingEntriesNum;
+ }
+}
+
+void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
+ actOnDeviceGlobalVarEntriesInfo(
+ const OffloadDeviceGlobalVarEntryInfoActTy &Action) {
+ // Scan all target region entries and perform the provided action.
+ for (const auto &E : OffloadEntriesDeviceGlobalVar)
+ Action(E.getKey(), E.getValue());
}
llvm::Function *
CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
-
// If we don't have entries or if we are emitting code for the device, we
// don't need to do anything.
if (CGM.getLangOpts().OpenMPIsDevice || OffloadEntriesInfoManager.empty())
return nullptr;
- auto &M = CGM.getModule();
- auto &C = CGM.getContext();
+ llvm::Module &M = CGM.getModule();
+ ASTContext &C = CGM.getContext();
// Get list of devices we care about
- auto &Devices = CGM.getLangOpts().OMPTargetTriples;
+ const std::vector<llvm::Triple> &Devices = CGM.getLangOpts().OMPTargetTriples;
// We should be creating an offloading descriptor only if there are devices
// specified.
@@ -3425,46 +3726,49 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
// Create the external variables that will point to the begin and end of the
// host entries section. These will be defined by the linker.
- auto *OffloadEntryTy =
+ llvm::Type *OffloadEntryTy =
CGM.getTypes().ConvertTypeForMem(getTgtOffloadEntryQTy());
- llvm::GlobalVariable *HostEntriesBegin = new llvm::GlobalVariable(
- M, OffloadEntryTy, /*isConstant=*/true,
- llvm::GlobalValue::ExternalLinkage, /*Initializer=*/nullptr,
- ".omp_offloading.entries_begin");
- llvm::GlobalVariable *HostEntriesEnd = new llvm::GlobalVariable(
+ std::string EntriesBeginName = getName({"omp_offloading", "entries_begin"});
+ auto *HostEntriesBegin = new llvm::GlobalVariable(
M, OffloadEntryTy, /*isConstant=*/true,
llvm::GlobalValue::ExternalLinkage, /*Initializer=*/nullptr,
- ".omp_offloading.entries_end");
+ EntriesBeginName);
+ std::string EntriesEndName = getName({"omp_offloading", "entries_end"});
+ auto *HostEntriesEnd =
+ new llvm::GlobalVariable(M, OffloadEntryTy, /*isConstant=*/true,
+ llvm::GlobalValue::ExternalLinkage,
+ /*Initializer=*/nullptr, EntriesEndName);
// Create all device images
auto *DeviceImageTy = cast<llvm::StructType>(
CGM.getTypes().ConvertTypeForMem(getTgtDeviceImageQTy()));
ConstantInitBuilder DeviceImagesBuilder(CGM);
- auto DeviceImagesEntries = DeviceImagesBuilder.beginArray(DeviceImageTy);
+ ConstantArrayBuilder DeviceImagesEntries =
+ DeviceImagesBuilder.beginArray(DeviceImageTy);
- for (unsigned i = 0; i < Devices.size(); ++i) {
- StringRef T = Devices[i].getTriple();
+ for (const llvm::Triple &Device : Devices) {
+ StringRef T = Device.getTriple();
+ std::string BeginName = getName({"omp_offloading", "img_start", ""});
auto *ImgBegin = new llvm::GlobalVariable(
M, CGM.Int8Ty, /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage,
- /*Initializer=*/nullptr,
- Twine(".omp_offloading.img_start.") + Twine(T));
+ /*Initializer=*/nullptr, Twine(BeginName).concat(T));
+ std::string EndName = getName({"omp_offloading", "img_end", ""});
auto *ImgEnd = new llvm::GlobalVariable(
M, CGM.Int8Ty, /*isConstant=*/true, llvm::GlobalValue::ExternalLinkage,
- /*Initializer=*/nullptr, Twine(".omp_offloading.img_end.") + Twine(T));
+ /*Initializer=*/nullptr, Twine(EndName).concat(T));
- auto Dev = DeviceImagesEntries.beginStruct(DeviceImageTy);
- Dev.add(ImgBegin);
- Dev.add(ImgEnd);
- Dev.add(HostEntriesBegin);
- Dev.add(HostEntriesEnd);
- Dev.finishAndAddTo(DeviceImagesEntries);
+ llvm::Constant *Data[] = {ImgBegin, ImgEnd, HostEntriesBegin,
+ HostEntriesEnd};
+ createConstantGlobalStructAndAddToParent(CGM, getTgtDeviceImageQTy(), Data,
+ DeviceImagesEntries);
}
// Create device images global array.
+ std::string ImagesName = getName({"omp_offloading", "device_images"});
llvm::GlobalVariable *DeviceImages =
- DeviceImagesEntries.finishAndCreateGlobal(".omp_offloading.device_images",
- CGM.getPointerAlign(),
- /*isConstant=*/true);
+ DeviceImagesEntries.finishAndCreateGlobal(ImagesName,
+ CGM.getPointerAlign(),
+ /*isConstant=*/true);
DeviceImages->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
// This is a Zero array to be used in the creation of the constant expressions
@@ -3472,49 +3776,64 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
llvm::Constant::getNullValue(CGM.Int32Ty)};
// Create the target region descriptor.
- auto *BinaryDescriptorTy = cast<llvm::StructType>(
- CGM.getTypes().ConvertTypeForMem(getTgtBinaryDescriptorQTy()));
- ConstantInitBuilder DescBuilder(CGM);
- auto DescInit = DescBuilder.beginStruct(BinaryDescriptorTy);
- DescInit.addInt(CGM.Int32Ty, Devices.size());
- DescInit.add(llvm::ConstantExpr::getGetElementPtr(DeviceImages->getValueType(),
- DeviceImages,
- Index));
- DescInit.add(HostEntriesBegin);
- DescInit.add(HostEntriesEnd);
-
- auto *Desc = DescInit.finishAndCreateGlobal(".omp_offloading.descriptor",
- CGM.getPointerAlign(),
- /*isConstant=*/true);
+ llvm::Constant *Data[] = {
+ llvm::ConstantInt::get(CGM.Int32Ty, Devices.size()),
+ llvm::ConstantExpr::getGetElementPtr(DeviceImages->getValueType(),
+ DeviceImages, Index),
+ HostEntriesBegin, HostEntriesEnd};
+ std::string Descriptor = getName({"omp_offloading", "descriptor"});
+ llvm::GlobalVariable *Desc = createConstantGlobalStruct(
+ CGM, getTgtBinaryDescriptorQTy(), Data, Descriptor);
// Emit code to register or unregister the descriptor at execution
// startup or closing, respectively.
- // Create a variable to drive the registration and unregistration of the
- // descriptor, so we can reuse the logic that emits Ctors and Dtors.
- auto *IdentInfo = &C.Idents.get(".omp_offloading.reg_unreg_var");
- ImplicitParamDecl RegUnregVar(C, C.getTranslationUnitDecl(), SourceLocation(),
- IdentInfo, C.CharTy, ImplicitParamDecl::Other);
-
- auto *UnRegFn = createOffloadingBinaryDescriptorFunction(
- CGM, ".omp_offloading.descriptor_unreg",
- [&](CodeGenFunction &CGF, PrePostActionTy &) {
- CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__tgt_unregister_lib),
- Desc);
- });
- auto *RegFn = createOffloadingBinaryDescriptorFunction(
- CGM, ".omp_offloading.descriptor_reg",
- [&](CodeGenFunction &CGF, PrePostActionTy &) {
- CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__tgt_register_lib),
- Desc);
- CGM.getCXXABI().registerGlobalDtor(CGF, RegUnregVar, UnRegFn, Desc);
- });
+ llvm::Function *UnRegFn;
+ {
+ FunctionArgList Args;
+ ImplicitParamDecl DummyPtr(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ Args.push_back(&DummyPtr);
+
+ CodeGenFunction CGF(CGM);
+ // Disable debug info for global (de-)initializer because they are not part
+ // of some particular construct.
+ CGF.disableDebugInfo();
+ const auto &FI =
+ CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ std::string UnregName = getName({"omp_offloading", "descriptor_unreg"});
+ UnRegFn = CGM.CreateGlobalInitOrDestructFunction(FTy, UnregName, FI);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, UnRegFn, FI, Args);
+ CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__tgt_unregister_lib),
+ Desc);
+ CGF.FinishFunction();
+ }
+ llvm::Function *RegFn;
+ {
+ CodeGenFunction CGF(CGM);
+ // Disable debug info for global (de-)initializer because they are not part
+ // of some particular construct.
+ CGF.disableDebugInfo();
+ const auto &FI = CGM.getTypes().arrangeNullaryFunction();
+ llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
+ std::string Descriptor = getName({"omp_offloading", "descriptor_reg"});
+ RegFn = CGM.CreateGlobalInitOrDestructFunction(FTy, Descriptor, FI);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, RegFn, FI, FunctionArgList());
+ CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__tgt_register_lib), Desc);
+ // Create a variable to drive the registration and unregistration of the
+ // descriptor, so we can reuse the logic that emits Ctors and Dtors.
+ ImplicitParamDecl RegUnregVar(C, C.getTranslationUnitDecl(),
+ SourceLocation(), nullptr, C.CharTy,
+ ImplicitParamDecl::Other);
+ CGM.getCXXABI().registerGlobalDtor(CGF, RegUnregVar, UnRegFn, Desc);
+ CGF.FinishFunction();
+ }
if (CGM.supportsCOMDAT()) {
// It is sufficient to call registration function only once, so create a
// COMDAT group for registration/unregistration functions and associated
// data. That would reduce startup time and code size. Registration
// function serves as a COMDAT group key.
- auto ComdatKey = M.getOrInsertComdat(RegFn->getName());
+ llvm::Comdat *ComdatKey = M.getOrInsertComdat(RegFn->getName());
RegFn->setLinkage(llvm::GlobalValue::LinkOnceAnyLinkage);
RegFn->setVisibility(llvm::GlobalValue::HiddenVisibility);
RegFn->setComdat(ComdatKey);
@@ -3525,48 +3844,35 @@ CGOpenMPRuntime::createOffloadingBinaryDescriptorRegistration() {
return RegFn;
}
-void CGOpenMPRuntime::createOffloadEntry(llvm::Constant *ID,
- llvm::Constant *Addr, uint64_t Size,
- int32_t Flags) {
+void CGOpenMPRuntime::createOffloadEntry(
+ llvm::Constant *ID, llvm::Constant *Addr, uint64_t Size, int32_t Flags,
+ llvm::GlobalValue::LinkageTypes Linkage) {
StringRef Name = Addr->getName();
- auto *TgtOffloadEntryType = cast<llvm::StructType>(
- CGM.getTypes().ConvertTypeForMem(getTgtOffloadEntryQTy()));
- llvm::LLVMContext &C = CGM.getModule().getContext();
llvm::Module &M = CGM.getModule();
-
- // Make sure the address has the right type.
- llvm::Constant *AddrPtr = llvm::ConstantExpr::getBitCast(ID, CGM.VoidPtrTy);
+ llvm::LLVMContext &C = M.getContext();
// Create constant string with the name.
llvm::Constant *StrPtrInit = llvm::ConstantDataArray::getString(C, Name);
- llvm::GlobalVariable *Str =
- new llvm::GlobalVariable(M, StrPtrInit->getType(), /*isConstant=*/true,
- llvm::GlobalValue::InternalLinkage, StrPtrInit,
- ".omp_offloading.entry_name");
+ std::string StringName = getName({"omp_offloading", "entry_name"});
+ auto *Str = new llvm::GlobalVariable(
+ M, StrPtrInit->getType(), /*isConstant=*/true,
+ llvm::GlobalValue::InternalLinkage, StrPtrInit, StringName);
Str->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
- llvm::Constant *StrPtr = llvm::ConstantExpr::getBitCast(Str, CGM.Int8PtrTy);
-
- // We can't have any padding between symbols, so we need to have 1-byte
- // alignment.
- auto Align = CharUnits::fromQuantity(1);
-
- // Create the entry struct.
- ConstantInitBuilder EntryBuilder(CGM);
- auto EntryInit = EntryBuilder.beginStruct(TgtOffloadEntryType);
- EntryInit.add(AddrPtr);
- EntryInit.add(StrPtr);
- EntryInit.addInt(CGM.SizeTy, Size);
- EntryInit.addInt(CGM.Int32Ty, Flags);
- EntryInit.addInt(CGM.Int32Ty, 0);
- llvm::GlobalVariable *Entry =
- EntryInit.finishAndCreateGlobal(".omp_offloading.entry",
- Align,
- /*constant*/ true,
- llvm::GlobalValue::ExternalLinkage);
+
+ llvm::Constant *Data[] = {llvm::ConstantExpr::getBitCast(ID, CGM.VoidPtrTy),
+ llvm::ConstantExpr::getBitCast(Str, CGM.Int8PtrTy),
+ llvm::ConstantInt::get(CGM.SizeTy, Size),
+ llvm::ConstantInt::get(CGM.Int32Ty, Flags),
+ llvm::ConstantInt::get(CGM.Int32Ty, 0)};
+ std::string EntryName = getName({"omp_offloading", "entry", ""});
+ llvm::GlobalVariable *Entry = createConstantGlobalStruct(
+ CGM, getTgtOffloadEntryQTy(), Data, Twine(EntryName).concat(Name),
+ llvm::GlobalValue::WeakAnyLinkage);
// The entry has to be created in the section the linker expects it to be.
- Entry->setSection(".omp_offloading.entries");
+ std::string Section = getName({"omp_offloading", "entries"});
+ Entry->setSection(Section);
}
void CGOpenMPRuntime::createOffloadEntriesAndInfoMetadata() {
@@ -3579,71 +3885,142 @@ void CGOpenMPRuntime::createOffloadEntriesAndInfoMetadata() {
// Right now we only generate metadata for function that contain target
// regions.
- // If we do not have entries, we dont need to do anything.
+ // If we do not have entries, we don't need to do anything.
if (OffloadEntriesInfoManager.empty())
return;
llvm::Module &M = CGM.getModule();
llvm::LLVMContext &C = M.getContext();
- SmallVector<OffloadEntriesInfoManagerTy::OffloadEntryInfo *, 16>
+ SmallVector<const OffloadEntriesInfoManagerTy::OffloadEntryInfo *, 16>
OrderedEntries(OffloadEntriesInfoManager.size());
- // Create the offloading info metadata node.
- llvm::NamedMDNode *MD = M.getOrInsertNamedMetadata("omp_offload.info");
-
// Auxiliary methods to create metadata values and strings.
- auto getMDInt = [&](unsigned v) {
+ auto &&GetMDInt = [this](unsigned V) {
return llvm::ConstantAsMetadata::get(
- llvm::ConstantInt::get(llvm::Type::getInt32Ty(C), v));
+ llvm::ConstantInt::get(CGM.Int32Ty, V));
};
- auto getMDString = [&](StringRef v) { return llvm::MDString::get(C, v); };
+ auto &&GetMDString = [&C](StringRef V) { return llvm::MDString::get(C, V); };
+
+ // Create the offloading info metadata node.
+ llvm::NamedMDNode *MD = M.getOrInsertNamedMetadata("omp_offload.info");
// Create function that emits metadata for each target region entry;
- auto &&TargetRegionMetadataEmitter = [&](
- unsigned DeviceID, unsigned FileID, StringRef ParentName, unsigned Line,
- OffloadEntriesInfoManagerTy::OffloadEntryInfoTargetRegion &E) {
- llvm::SmallVector<llvm::Metadata *, 32> Ops;
- // Generate metadata for target regions. Each entry of this metadata
- // contains:
- // - Entry 0 -> Kind of this type of metadata (0).
- // - Entry 1 -> Device ID of the file where the entry was identified.
- // - Entry 2 -> File ID of the file where the entry was identified.
- // - Entry 3 -> Mangled name of the function where the entry was identified.
- // - Entry 4 -> Line in the file where the entry was identified.
- // - Entry 5 -> Order the entry was created.
- // The first element of the metadata node is the kind.
- Ops.push_back(getMDInt(E.getKind()));
- Ops.push_back(getMDInt(DeviceID));
- Ops.push_back(getMDInt(FileID));
- Ops.push_back(getMDString(ParentName));
- Ops.push_back(getMDInt(Line));
- Ops.push_back(getMDInt(E.getOrder()));
-
- // Save this entry in the right position of the ordered entries array.
- OrderedEntries[E.getOrder()] = &E;
-
- // Add metadata to the named metadata node.
- MD->addOperand(llvm::MDNode::get(C, Ops));
- };
+ auto &&TargetRegionMetadataEmitter =
+ [&C, MD, &OrderedEntries, &GetMDInt, &GetMDString](
+ unsigned DeviceID, unsigned FileID, StringRef ParentName,
+ unsigned Line,
+ const OffloadEntriesInfoManagerTy::OffloadEntryInfoTargetRegion &E) {
+ // Generate metadata for target regions. Each entry of this metadata
+ // contains:
+ // - Entry 0 -> Kind of this type of metadata (0).
+ // - Entry 1 -> Device ID of the file where the entry was identified.
+ // - Entry 2 -> File ID of the file where the entry was identified.
+ // - Entry 3 -> Mangled name of the function where the entry was
+ // identified.
+ // - Entry 4 -> Line in the file where the entry was identified.
+ // - Entry 5 -> Order the entry was created.
+ // The first element of the metadata node is the kind.
+ llvm::Metadata *Ops[] = {GetMDInt(E.getKind()), GetMDInt(DeviceID),
+ GetMDInt(FileID), GetMDString(ParentName),
+ GetMDInt(Line), GetMDInt(E.getOrder())};
+
+ // Save this entry in the right position of the ordered entries array.
+ OrderedEntries[E.getOrder()] = &E;
+
+ // Add metadata to the named metadata node.
+ MD->addOperand(llvm::MDNode::get(C, Ops));
+ };
OffloadEntriesInfoManager.actOnTargetRegionEntriesInfo(
TargetRegionMetadataEmitter);
- for (auto *E : OrderedEntries) {
+ // Create function that emits metadata for each device global variable entry;
+ auto &&DeviceGlobalVarMetadataEmitter =
+ [&C, &OrderedEntries, &GetMDInt, &GetMDString,
+ MD](StringRef MangledName,
+ const OffloadEntriesInfoManagerTy::OffloadEntryInfoDeviceGlobalVar
+ &E) {
+ // Generate metadata for global variables. Each entry of this metadata
+ // contains:
+ // - Entry 0 -> Kind of this type of metadata (1).
+ // - Entry 1 -> Mangled name of the variable.
+ // - Entry 2 -> Declare target kind.
+ // - Entry 3 -> Order the entry was created.
+ // The first element of the metadata node is the kind.
+ llvm::Metadata *Ops[] = {
+ GetMDInt(E.getKind()), GetMDString(MangledName),
+ GetMDInt(E.getFlags()), GetMDInt(E.getOrder())};
+
+ // Save this entry in the right position of the ordered entries array.
+ OrderedEntries[E.getOrder()] = &E;
+
+ // Add metadata to the named metadata node.
+ MD->addOperand(llvm::MDNode::get(C, Ops));
+ };
+
+ OffloadEntriesInfoManager.actOnDeviceGlobalVarEntriesInfo(
+ DeviceGlobalVarMetadataEmitter);
+
+ for (const auto *E : OrderedEntries) {
assert(E && "All ordered entries must exist!");
- if (auto *CE =
+ if (const auto *CE =
dyn_cast<OffloadEntriesInfoManagerTy::OffloadEntryInfoTargetRegion>(
E)) {
- assert(CE->getID() && CE->getAddress() &&
- "Entry ID and Addr are invalid!");
- createOffloadEntry(CE->getID(), CE->getAddress(), /*Size=*/0);
- } else
+ if (!CE->getID() || !CE->getAddress()) {
+ unsigned DiagID = CGM.getDiags().getCustomDiagID(
+ DiagnosticsEngine::Error,
+ "Offloading entry for target region is incorrect: either the "
+ "address or the ID is invalid.");
+ CGM.getDiags().Report(DiagID);
+ continue;
+ }
+ createOffloadEntry(CE->getID(), CE->getAddress(), /*Size=*/0,
+ CE->getFlags(), llvm::GlobalValue::WeakAnyLinkage);
+ } else if (const auto *CE =
+ dyn_cast<OffloadEntriesInfoManagerTy::
+ OffloadEntryInfoDeviceGlobalVar>(E)) {
+ OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind Flags =
+ static_cast<OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind>(
+ CE->getFlags());
+ switch (Flags) {
+ case OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryTo: {
+ if (!CE->getAddress()) {
+ unsigned DiagID = CGM.getDiags().getCustomDiagID(
+ DiagnosticsEngine::Error,
+ "Offloading entry for declare target variable is incorrect: the "
+ "address is invalid.");
+ CGM.getDiags().Report(DiagID);
+ continue;
+ }
+ break;
+ }
+ case OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryLink:
+ assert(((CGM.getLangOpts().OpenMPIsDevice && !CE->getAddress()) ||
+ (!CGM.getLangOpts().OpenMPIsDevice && CE->getAddress())) &&
+ "Declaret target link address is set.");
+ if (CGM.getLangOpts().OpenMPIsDevice)
+ continue;
+ if (!CE->getAddress()) {
+ unsigned DiagID = CGM.getDiags().getCustomDiagID(
+ DiagnosticsEngine::Error,
+ "Offloading entry for declare target variable is incorrect: the "
+ "address is invalid.");
+ CGM.getDiags().Report(DiagID);
+ continue;
+ }
+ break;
+ }
+ createOffloadEntry(CE->getAddress(), CE->getAddress(),
+ CE->getVarSize().getQuantity(), Flags,
+ CE->getLinkage());
+ } else {
llvm_unreachable("Unsupported entry kind.");
+ }
}
}
-/// \brief Loads all the offload entries information from the host IR
+/// Loads all the offload entries information from the host IR
/// metadata.
void CGOpenMPRuntime::loadOffloadInfoMetadata() {
// If we are in target mode, load the metadata from the host IR. This code has
@@ -3656,44 +4033,57 @@ void CGOpenMPRuntime::loadOffloadInfoMetadata() {
return;
auto Buf = llvm::MemoryBuffer::getFile(CGM.getLangOpts().OMPHostIRFile);
- if (Buf.getError())
+ if (auto EC = Buf.getError()) {
+ CGM.getDiags().Report(diag::err_cannot_open_file)
+ << CGM.getLangOpts().OMPHostIRFile << EC.message();
return;
+ }
llvm::LLVMContext C;
auto ME = expectedToErrorOrAndEmitErrors(
C, llvm::parseBitcodeFile(Buf.get()->getMemBufferRef(), C));
- if (ME.getError())
+ if (auto EC = ME.getError()) {
+ unsigned DiagID = CGM.getDiags().getCustomDiagID(
+ DiagnosticsEngine::Error, "Unable to parse host IR file '%0':'%1'");
+ CGM.getDiags().Report(DiagID)
+ << CGM.getLangOpts().OMPHostIRFile << EC.message();
return;
+ }
llvm::NamedMDNode *MD = ME.get()->getNamedMetadata("omp_offload.info");
if (!MD)
return;
- for (auto I : MD->operands()) {
- llvm::MDNode *MN = cast<llvm::MDNode>(I);
-
- auto getMDInt = [&](unsigned Idx) {
- llvm::ConstantAsMetadata *V =
- cast<llvm::ConstantAsMetadata>(MN->getOperand(Idx));
+ for (llvm::MDNode *MN : MD->operands()) {
+ auto &&GetMDInt = [MN](unsigned Idx) {
+ auto *V = cast<llvm::ConstantAsMetadata>(MN->getOperand(Idx));
return cast<llvm::ConstantInt>(V->getValue())->getZExtValue();
};
- auto getMDString = [&](unsigned Idx) {
- llvm::MDString *V = cast<llvm::MDString>(MN->getOperand(Idx));
+ auto &&GetMDString = [MN](unsigned Idx) {
+ auto *V = cast<llvm::MDString>(MN->getOperand(Idx));
return V->getString();
};
- switch (getMDInt(0)) {
+ switch (GetMDInt(0)) {
default:
llvm_unreachable("Unexpected metadata!");
break;
case OffloadEntriesInfoManagerTy::OffloadEntryInfo::
- OFFLOAD_ENTRY_INFO_TARGET_REGION:
+ OffloadingEntryInfoTargetRegion:
OffloadEntriesInfoManager.initializeTargetRegionEntryInfo(
- /*DeviceID=*/getMDInt(1), /*FileID=*/getMDInt(2),
- /*ParentName=*/getMDString(3), /*Line=*/getMDInt(4),
- /*Order=*/getMDInt(5));
+ /*DeviceID=*/GetMDInt(1), /*FileID=*/GetMDInt(2),
+ /*ParentName=*/GetMDString(3), /*Line=*/GetMDInt(4),
+ /*Order=*/GetMDInt(5));
+ break;
+ case OffloadEntriesInfoManagerTy::OffloadEntryInfo::
+ OffloadingEntryInfoDeviceGlobalVar:
+ OffloadEntriesInfoManager.initializeDeviceGlobalVarEntryInfo(
+ /*MangledName=*/GetMDString(1),
+ static_cast<OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind>(
+ /*Flags=*/GetMDInt(2)),
+ /*Order=*/GetMDInt(3));
break;
}
}
@@ -3702,7 +4092,7 @@ void CGOpenMPRuntime::loadOffloadInfoMetadata() {
void CGOpenMPRuntime::emitKmpRoutineEntryT(QualType KmpInt32Ty) {
if (!KmpRoutineEntryPtrTy) {
// Build typedef kmp_int32 (* kmp_routine_entry_t)(kmp_int32, void *); type.
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
QualType KmpRoutineEntryTyArgs[] = {KmpInt32Ty, C.VoidPtrTy};
FunctionProtoType::ExtProtoInfo EPI;
KmpRoutineEntryPtrQTy = C.getPointerType(
@@ -3711,19 +4101,7 @@ void CGOpenMPRuntime::emitKmpRoutineEntryT(QualType KmpInt32Ty) {
}
}
-static FieldDecl *addFieldToRecordDecl(ASTContext &C, DeclContext *DC,
- QualType FieldTy) {
- auto *Field = FieldDecl::Create(
- C, DC, SourceLocation(), SourceLocation(), /*Id=*/nullptr, FieldTy,
- C.getTrivialTypeSourceInfo(FieldTy, SourceLocation()),
- /*BW=*/nullptr, /*Mutable=*/false, /*InitStyle=*/ICIS_NoInit);
- Field->setAccess(AS_public);
- DC->addDecl(Field);
- return Field;
-}
-
QualType CGOpenMPRuntime::getTgtOffloadEntryQTy() {
-
// Make sure the type of the entry is already created. This is the type we
// have to create:
// struct __tgt_offload_entry{
@@ -3736,7 +4114,7 @@ QualType CGOpenMPRuntime::getTgtOffloadEntryQTy() {
// };
if (TgtOffloadEntryQTy.isNull()) {
ASTContext &C = CGM.getContext();
- auto *RD = C.buildImplicitRecord("__tgt_offload_entry");
+ RecordDecl *RD = C.buildImplicitRecord("__tgt_offload_entry");
RD->startDefinition();
addFieldToRecordDecl(C, RD, C.VoidPtrTy);
addFieldToRecordDecl(C, RD, C.getPointerType(C.CharTy));
@@ -3746,6 +4124,7 @@ QualType CGOpenMPRuntime::getTgtOffloadEntryQTy() {
addFieldToRecordDecl(
C, RD, C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true));
RD->completeDefinition();
+ RD->addAttr(PackedAttr::CreateImplicit(C));
TgtOffloadEntryQTy = C.getRecordType(RD);
}
return TgtOffloadEntryQTy;
@@ -3765,7 +4144,7 @@ QualType CGOpenMPRuntime::getTgtDeviceImageQTy() {
// };
if (TgtDeviceImageQTy.isNull()) {
ASTContext &C = CGM.getContext();
- auto *RD = C.buildImplicitRecord("__tgt_device_image");
+ RecordDecl *RD = C.buildImplicitRecord("__tgt_device_image");
RD->startDefinition();
addFieldToRecordDecl(C, RD, C.VoidPtrTy);
addFieldToRecordDecl(C, RD, C.VoidPtrTy);
@@ -3789,7 +4168,7 @@ QualType CGOpenMPRuntime::getTgtBinaryDescriptorQTy() {
// };
if (TgtBinaryDescriptorQTy.isNull()) {
ASTContext &C = CGM.getContext();
- auto *RD = C.buildImplicitRecord("__tgt_bin_desc");
+ RecordDecl *RD = C.buildImplicitRecord("__tgt_bin_desc");
RD->startDefinition();
addFieldToRecordDecl(
C, RD, C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true));
@@ -3818,17 +4197,16 @@ typedef std::pair<CharUnits /*Align*/, PrivateHelpersTy> PrivateDataTy;
static RecordDecl *
createPrivatesRecordDecl(CodeGenModule &CGM, ArrayRef<PrivateDataTy> Privates) {
if (!Privates.empty()) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
// Build struct .kmp_privates_t. {
// /* private vars */
// };
- auto *RD = C.buildImplicitRecord(".kmp_privates.t");
+ RecordDecl *RD = C.buildImplicitRecord(".kmp_privates.t");
RD->startDefinition();
- for (auto &&Pair : Privates) {
- auto *VD = Pair.second.Original;
- auto Type = VD->getType();
- Type = Type.getNonReferenceType();
- auto *FD = addFieldToRecordDecl(C, RD, Type);
+ for (const auto &Pair : Privates) {
+ const VarDecl *VD = Pair.second.Original;
+ QualType Type = VD->getType().getNonReferenceType();
+ FieldDecl *FD = addFieldToRecordDecl(C, RD, Type);
if (VD->hasAttrs()) {
for (specific_attr_iterator<AlignedAttr> I(VD->getAttrs().begin()),
E(VD->getAttrs().end());
@@ -3846,7 +4224,7 @@ static RecordDecl *
createKmpTaskTRecordDecl(CodeGenModule &CGM, OpenMPDirectiveKind Kind,
QualType KmpInt32Ty,
QualType KmpRoutineEntryPointerQTy) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
// Build struct kmp_task_t {
// void * shareds;
// kmp_routine_entry_t routine;
@@ -3860,13 +4238,13 @@ createKmpTaskTRecordDecl(CodeGenModule &CGM, OpenMPDirectiveKind Kind,
// kmp_int32 liter;
// void * reductions;
// };
- auto *UD = C.buildImplicitRecord("kmp_cmplrdata_t", TTK_Union);
+ RecordDecl *UD = C.buildImplicitRecord("kmp_cmplrdata_t", TTK_Union);
UD->startDefinition();
addFieldToRecordDecl(C, UD, KmpInt32Ty);
addFieldToRecordDecl(C, UD, KmpRoutineEntryPointerQTy);
UD->completeDefinition();
QualType KmpCmplrdataTy = C.getRecordType(UD);
- auto *RD = C.buildImplicitRecord("kmp_task_t");
+ RecordDecl *RD = C.buildImplicitRecord("kmp_task_t");
RD->startDefinition();
addFieldToRecordDecl(C, RD, C.VoidPtrTy);
addFieldToRecordDecl(C, RD, KmpRoutineEntryPointerQTy);
@@ -3891,22 +4269,21 @@ createKmpTaskTRecordDecl(CodeGenModule &CGM, OpenMPDirectiveKind Kind,
static RecordDecl *
createKmpTaskTWithPrivatesRecordDecl(CodeGenModule &CGM, QualType KmpTaskTQTy,
ArrayRef<PrivateDataTy> Privates) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
// Build struct kmp_task_t_with_privates {
// kmp_task_t task_data;
// .kmp_privates_t. privates;
// };
- auto *RD = C.buildImplicitRecord("kmp_task_t_with_privates");
+ RecordDecl *RD = C.buildImplicitRecord("kmp_task_t_with_privates");
RD->startDefinition();
addFieldToRecordDecl(C, RD, KmpTaskTQTy);
- if (auto *PrivateRD = createPrivatesRecordDecl(CGM, Privates)) {
+ if (const RecordDecl *PrivateRD = createPrivatesRecordDecl(CGM, Privates))
addFieldToRecordDecl(C, RD, C.getRecordType(PrivateRD));
- }
RD->completeDefinition();
return RD;
}
-/// \brief Emit a proxy function which accepts kmp_task_t as the second
+/// Emit a proxy function which accepts kmp_task_t as the second
/// argument.
/// \code
/// kmp_int32 .omp_task_entry.(kmp_int32 gtid, kmp_task_t *tt) {
@@ -3924,7 +4301,7 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
QualType KmpTaskTWithPrivatesQTy, QualType KmpTaskTQTy,
QualType SharedsPtrTy, llvm::Value *TaskFunction,
llvm::Value *TaskPrivatesMap) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl GtidArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, KmpInt32Ty,
ImplicitParamDecl::Other);
@@ -3933,49 +4310,53 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
ImplicitParamDecl::Other);
Args.push_back(&GtidArg);
Args.push_back(&TaskTypeArg);
- auto &TaskEntryFnInfo =
+ const auto &TaskEntryFnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(KmpInt32Ty, Args);
- auto *TaskEntryTy = CGM.getTypes().GetFunctionType(TaskEntryFnInfo);
- auto *TaskEntry =
- llvm::Function::Create(TaskEntryTy, llvm::GlobalValue::InternalLinkage,
- ".omp_task_entry.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, TaskEntry, TaskEntryFnInfo);
+ llvm::FunctionType *TaskEntryTy =
+ CGM.getTypes().GetFunctionType(TaskEntryFnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName({"omp_task_entry", ""});
+ auto *TaskEntry = llvm::Function::Create(
+ TaskEntryTy, llvm::GlobalValue::InternalLinkage, Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskEntry, TaskEntryFnInfo);
+ TaskEntry->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.StartFunction(GlobalDecl(), KmpInt32Ty, TaskEntry, TaskEntryFnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), KmpInt32Ty, TaskEntry, TaskEntryFnInfo, Args,
+ Loc, Loc);
// TaskFunction(gtid, tt->task_data.part_id, &tt->privates, task_privates_map,
// tt,
// For taskloops:
// tt->task_data.lb, tt->task_data.ub, tt->task_data.st, tt->task_data.liter,
// tt->task_data.shareds);
- auto *GtidParam = CGF.EmitLoadOfScalar(
+ llvm::Value *GtidParam = CGF.EmitLoadOfScalar(
CGF.GetAddrOfLocalVar(&GtidArg), /*Volatile=*/false, KmpInt32Ty, Loc);
LValue TDBase = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(&TaskTypeArg),
KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
- auto *KmpTaskTWithPrivatesQTyRD =
+ const auto *KmpTaskTWithPrivatesQTyRD =
cast<RecordDecl>(KmpTaskTWithPrivatesQTy->getAsTagDecl());
LValue Base =
CGF.EmitLValueForField(TDBase, *KmpTaskTWithPrivatesQTyRD->field_begin());
- auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
+ const auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
auto PartIdFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTPartId);
- auto PartIdLVal = CGF.EmitLValueForField(Base, *PartIdFI);
- auto *PartidParam = PartIdLVal.getPointer();
+ LValue PartIdLVal = CGF.EmitLValueForField(Base, *PartIdFI);
+ llvm::Value *PartidParam = PartIdLVal.getPointer();
auto SharedsFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTShareds);
- auto SharedsLVal = CGF.EmitLValueForField(Base, *SharedsFI);
- auto *SharedsParam = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
- CGF.EmitLoadOfLValue(SharedsLVal, Loc).getScalarVal(),
+ LValue SharedsLVal = CGF.EmitLValueForField(Base, *SharedsFI);
+ llvm::Value *SharedsParam = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
+ CGF.EmitLoadOfScalar(SharedsLVal, Loc),
CGF.ConvertTypeForMem(SharedsPtrTy));
auto PrivatesFI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin(), 1);
llvm::Value *PrivatesParam;
if (PrivatesFI != KmpTaskTWithPrivatesQTyRD->field_end()) {
- auto PrivatesLVal = CGF.EmitLValueForField(TDBase, *PrivatesFI);
+ LValue PrivatesLVal = CGF.EmitLValueForField(TDBase, *PrivatesFI);
PrivatesParam = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
PrivatesLVal.getPointer(), CGF.VoidPtrTy);
- } else
+ } else {
PrivatesParam = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
+ }
llvm::Value *CommonArgs[] = {GtidParam, PartidParam, PrivatesParam,
TaskPrivatesMap,
@@ -3987,20 +4368,20 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
std::end(CommonArgs));
if (isOpenMPTaskLoopDirective(Kind)) {
auto LBFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTLowerBound);
- auto LBLVal = CGF.EmitLValueForField(Base, *LBFI);
- auto *LBParam = CGF.EmitLoadOfLValue(LBLVal, Loc).getScalarVal();
+ LValue LBLVal = CGF.EmitLValueForField(Base, *LBFI);
+ llvm::Value *LBParam = CGF.EmitLoadOfScalar(LBLVal, Loc);
auto UBFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTUpperBound);
- auto UBLVal = CGF.EmitLValueForField(Base, *UBFI);
- auto *UBParam = CGF.EmitLoadOfLValue(UBLVal, Loc).getScalarVal();
+ LValue UBLVal = CGF.EmitLValueForField(Base, *UBFI);
+ llvm::Value *UBParam = CGF.EmitLoadOfScalar(UBLVal, Loc);
auto StFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTStride);
- auto StLVal = CGF.EmitLValueForField(Base, *StFI);
- auto *StParam = CGF.EmitLoadOfLValue(StLVal, Loc).getScalarVal();
+ LValue StLVal = CGF.EmitLValueForField(Base, *StFI);
+ llvm::Value *StParam = CGF.EmitLoadOfScalar(StLVal, Loc);
auto LIFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTLastIter);
- auto LILVal = CGF.EmitLValueForField(Base, *LIFI);
- auto *LIParam = CGF.EmitLoadOfLValue(LILVal, Loc).getScalarVal();
+ LValue LILVal = CGF.EmitLValueForField(Base, *LIFI);
+ llvm::Value *LIParam = CGF.EmitLoadOfScalar(LILVal, Loc);
auto RFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTReductions);
- auto RLVal = CGF.EmitLValueForField(Base, *RFI);
- auto *RParam = CGF.EmitLoadOfLValue(RLVal, Loc).getScalarVal();
+ LValue RLVal = CGF.EmitLValueForField(Base, *RFI);
+ llvm::Value *RParam = CGF.EmitLoadOfScalar(RLVal, Loc);
CallArgs.push_back(LBParam);
CallArgs.push_back(UBParam);
CallArgs.push_back(StParam);
@@ -4011,9 +4392,8 @@ emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
CGM.getOpenMPRuntime().emitOutlinedFunctionCall(CGF, Loc, TaskFunction,
CallArgs);
- CGF.EmitStoreThroughLValue(
- RValue::get(CGF.Builder.getInt32(/*C=*/0)),
- CGF.MakeAddrLValue(CGF.ReturnValue, KmpInt32Ty));
+ CGF.EmitStoreThroughLValue(RValue::get(CGF.Builder.getInt32(/*C=*/0)),
+ CGF.MakeAddrLValue(CGF.ReturnValue, KmpInt32Ty));
CGF.FinishFunction();
return TaskEntry;
}
@@ -4023,7 +4403,7 @@ static llvm::Value *emitDestructorsFunction(CodeGenModule &CGM,
QualType KmpInt32Ty,
QualType KmpTaskTWithPrivatesPtrQTy,
QualType KmpTaskTWithPrivatesQTy) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl GtidArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, KmpInt32Ty,
ImplicitParamDecl::Other);
@@ -4032,30 +4412,34 @@ static llvm::Value *emitDestructorsFunction(CodeGenModule &CGM,
ImplicitParamDecl::Other);
Args.push_back(&GtidArg);
Args.push_back(&TaskTypeArg);
- auto &DestructorFnInfo =
+ const auto &DestructorFnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(KmpInt32Ty, Args);
- auto *DestructorFnTy = CGM.getTypes().GetFunctionType(DestructorFnInfo);
+ llvm::FunctionType *DestructorFnTy =
+ CGM.getTypes().GetFunctionType(DestructorFnInfo);
+ std::string Name =
+ CGM.getOpenMPRuntime().getName({"omp_task_destructor", ""});
auto *DestructorFn =
llvm::Function::Create(DestructorFnTy, llvm::GlobalValue::InternalLinkage,
- ".omp_task_destructor.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, DestructorFn,
+ Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), DestructorFn,
DestructorFnInfo);
+ DestructorFn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
CGF.StartFunction(GlobalDecl(), KmpInt32Ty, DestructorFn, DestructorFnInfo,
- Args);
+ Args, Loc, Loc);
LValue Base = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(&TaskTypeArg),
KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
- auto *KmpTaskTWithPrivatesQTyRD =
+ const auto *KmpTaskTWithPrivatesQTyRD =
cast<RecordDecl>(KmpTaskTWithPrivatesQTy->getAsTagDecl());
auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
Base = CGF.EmitLValueForField(Base, *FI);
- for (auto *Field :
+ for (const auto *Field :
cast<RecordDecl>(FI->getType()->getAsTagDecl())->fields()) {
- if (auto DtorKind = Field->getType().isDestructedType()) {
- auto FieldLValue = CGF.EmitLValueForField(Base, Field);
+ if (QualType::DestructionKind DtorKind =
+ Field->getType().isDestructedType()) {
+ LValue FieldLValue = CGF.EmitLValueForField(Base, Field);
CGF.pushDestroy(DtorKind, FieldLValue.getAddress(), Field->getType());
}
}
@@ -4063,7 +4447,7 @@ static llvm::Value *emitDestructorsFunction(CodeGenModule &CGM,
return DestructorFn;
}
-/// \brief Emit a privates mapping function for correct handling of private and
+/// Emit a privates mapping function for correct handling of private and
/// firstprivate variables.
/// \code
/// void .omp_task_privates_map.(const .privates. *noalias privs, <ty1>
@@ -4080,7 +4464,7 @@ emitTaskPrivateMappingFunction(CodeGenModule &CGM, SourceLocation Loc,
ArrayRef<const Expr *> LastprivateVars,
QualType PrivatesQTy,
ArrayRef<PrivateDataTy> Privates) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl TaskPrivatesArg(
C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
@@ -4089,67 +4473,69 @@ emitTaskPrivateMappingFunction(CodeGenModule &CGM, SourceLocation Loc,
Args.push_back(&TaskPrivatesArg);
llvm::DenseMap<const VarDecl *, unsigned> PrivateVarsPos;
unsigned Counter = 1;
- for (auto *E: PrivateVars) {
+ for (const Expr *E : PrivateVars) {
Args.push_back(ImplicitParamDecl::Create(
C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
C.getPointerType(C.getPointerType(E->getType()))
.withConst()
.withRestrict(),
ImplicitParamDecl::Other));
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
PrivateVarsPos[VD] = Counter;
++Counter;
}
- for (auto *E : FirstprivateVars) {
+ for (const Expr *E : FirstprivateVars) {
Args.push_back(ImplicitParamDecl::Create(
C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
C.getPointerType(C.getPointerType(E->getType()))
.withConst()
.withRestrict(),
ImplicitParamDecl::Other));
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
PrivateVarsPos[VD] = Counter;
++Counter;
}
- for (auto *E: LastprivateVars) {
+ for (const Expr *E : LastprivateVars) {
Args.push_back(ImplicitParamDecl::Create(
C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
C.getPointerType(C.getPointerType(E->getType()))
.withConst()
.withRestrict(),
ImplicitParamDecl::Other));
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
PrivateVarsPos[VD] = Counter;
++Counter;
}
- auto &TaskPrivatesMapFnInfo =
+ const auto &TaskPrivatesMapFnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *TaskPrivatesMapTy =
+ llvm::FunctionType *TaskPrivatesMapTy =
CGM.getTypes().GetFunctionType(TaskPrivatesMapFnInfo);
+ std::string Name =
+ CGM.getOpenMPRuntime().getName({"omp_task_privates_map", ""});
auto *TaskPrivatesMap = llvm::Function::Create(
- TaskPrivatesMapTy, llvm::GlobalValue::InternalLinkage,
- ".omp_task_privates_map.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, TaskPrivatesMap,
+ TaskPrivatesMapTy, llvm::GlobalValue::InternalLinkage, Name,
+ &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskPrivatesMap,
TaskPrivatesMapFnInfo);
TaskPrivatesMap->removeFnAttr(llvm::Attribute::NoInline);
TaskPrivatesMap->removeFnAttr(llvm::Attribute::OptimizeNone);
TaskPrivatesMap->addFnAttr(llvm::Attribute::AlwaysInline);
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
CGF.StartFunction(GlobalDecl(), C.VoidTy, TaskPrivatesMap,
- TaskPrivatesMapFnInfo, Args);
+ TaskPrivatesMapFnInfo, Args, Loc, Loc);
// *privi = &.privates.privi;
LValue Base = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(&TaskPrivatesArg),
TaskPrivatesArg.getType()->castAs<PointerType>());
- auto *PrivatesQTyRD = cast<RecordDecl>(PrivatesQTy->getAsTagDecl());
+ const auto *PrivatesQTyRD = cast<RecordDecl>(PrivatesQTy->getAsTagDecl());
Counter = 0;
- for (auto *Field : PrivatesQTyRD->fields()) {
- auto FieldLVal = CGF.EmitLValueForField(Base, Field);
- auto *VD = Args[PrivateVarsPos[Privates[Counter].second.Original]];
- auto RefLVal = CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(VD), VD->getType());
- auto RefLoadLVal = CGF.EmitLoadOfPointerLValue(
+ for (const FieldDecl *Field : PrivatesQTyRD->fields()) {
+ LValue FieldLVal = CGF.EmitLValueForField(Base, Field);
+ const VarDecl *VD = Args[PrivateVarsPos[Privates[Counter].second.Original]];
+ LValue RefLVal =
+ CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(VD), VD->getType());
+ LValue RefLoadLVal = CGF.EmitLoadOfPointerLValue(
RefLVal.getAddress(), RefLVal.getType()->castAs<PointerType>());
CGF.EmitStoreOfScalar(FieldLVal.getPointer(), RefLoadLVal);
++Counter;
@@ -4171,9 +4557,14 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
QualType SharedsTy, QualType SharedsPtrTy,
const OMPTaskDataTy &Data,
ArrayRef<PrivateDataTy> Privates, bool ForDup) {
- auto &C = CGF.getContext();
+ ASTContext &C = CGF.getContext();
auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
LValue PrivatesBase = CGF.EmitLValueForField(TDBase, *FI);
+ OpenMPDirectiveKind Kind = isOpenMPTaskLoopDirective(D.getDirectiveKind())
+ ? OMPD_taskloop
+ : OMPD_task;
+ const CapturedStmt &CS = *D.getCapturedStmt(Kind);
+ CodeGenFunction::CGCapturedStmtInfo CapturesInfo(CS);
LValue SrcBase;
bool IsTargetTask =
isOpenMPTargetDataManagementDirective(D.getDirectiveKind()) ||
@@ -4182,40 +4573,38 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
// PointersArray and SizesArray. The original variables for these arrays are
// not captured and we get their addresses explicitly.
if ((!IsTargetTask && !Data.FirstprivateVars.empty()) ||
- (IsTargetTask && Data.FirstprivateVars.size() > 3)) {
+ (IsTargetTask && KmpTaskSharedsPtr.isValid())) {
SrcBase = CGF.MakeAddrLValue(
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
KmpTaskSharedsPtr, CGF.ConvertTypeForMem(SharedsPtrTy)),
SharedsTy);
}
- OpenMPDirectiveKind Kind = isOpenMPTaskLoopDirective(D.getDirectiveKind())
- ? OMPD_taskloop
- : OMPD_task;
- CodeGenFunction::CGCapturedStmtInfo CapturesInfo(*D.getCapturedStmt(Kind));
FI = cast<RecordDecl>(FI->getType()->getAsTagDecl())->field_begin();
- for (auto &&Pair : Privates) {
- auto *VD = Pair.second.PrivateCopy;
- auto *Init = VD->getAnyInitializer();
+ for (const PrivateDataTy &Pair : Privates) {
+ const VarDecl *VD = Pair.second.PrivateCopy;
+ const Expr *Init = VD->getAnyInitializer();
if (Init && (!ForDup || (isa<CXXConstructExpr>(Init) &&
!CGF.isTrivialInitializer(Init)))) {
LValue PrivateLValue = CGF.EmitLValueForField(PrivatesBase, *FI);
- if (auto *Elem = Pair.second.PrivateElemInit) {
- auto *OriginalVD = Pair.second.Original;
+ if (const VarDecl *Elem = Pair.second.PrivateElemInit) {
+ const VarDecl *OriginalVD = Pair.second.Original;
// Check if the variable is the target-based BasePointersArray,
// PointersArray or SizesArray.
LValue SharedRefLValue;
QualType Type = OriginalVD->getType();
- if (IsTargetTask && isa<ImplicitParamDecl>(OriginalVD) &&
- isa<CapturedDecl>(OriginalVD->getDeclContext()) &&
- cast<CapturedDecl>(OriginalVD->getDeclContext())->getNumParams() ==
- 0 &&
- isa<TranslationUnitDecl>(
- cast<CapturedDecl>(OriginalVD->getDeclContext())
- ->getDeclContext())) {
+ const FieldDecl *SharedField = CapturesInfo.lookup(OriginalVD);
+ if (IsTargetTask && !SharedField) {
+ assert(isa<ImplicitParamDecl>(OriginalVD) &&
+ isa<CapturedDecl>(OriginalVD->getDeclContext()) &&
+ cast<CapturedDecl>(OriginalVD->getDeclContext())
+ ->getNumParams() == 0 &&
+ isa<TranslationUnitDecl>(
+ cast<CapturedDecl>(OriginalVD->getDeclContext())
+ ->getDeclContext()) &&
+ "Expected artificial target data variable.");
SharedRefLValue =
CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(OriginalVD), Type);
} else {
- auto *SharedField = CapturesInfo.lookup(OriginalVD);
SharedRefLValue = CGF.EmitLValueForField(SrcBase, SharedField);
SharedRefLValue = CGF.MakeAddrLValue(
Address(SharedRefLValue.getPointer(), C.getDeclAlign(OriginalVD)),
@@ -4226,8 +4615,7 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
// Initialize firstprivate array.
if (!isa<CXXConstructExpr>(Init) || CGF.isTrivialInitializer(Init)) {
// Perform simple memcpy.
- CGF.EmitAggregateAssign(PrivateLValue.getAddress(),
- SharedRefLValue.getAddress(), Type);
+ CGF.EmitAggregateAssign(PrivateLValue, SharedRefLValue, Type);
} else {
// Initialize firstprivate array using element-by-element
// initialization.
@@ -4258,8 +4646,9 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
CGF.EmitExprAsInit(Init, VD, PrivateLValue,
/*capturedByInit=*/false);
}
- } else
+ } else {
CGF.EmitExprAsInit(Init, VD, PrivateLValue, /*capturedByInit=*/false);
+ }
}
++FI;
}
@@ -4269,11 +4658,13 @@ static void emitPrivatesInit(CodeGenFunction &CGF,
static bool checkInitIsRequired(CodeGenFunction &CGF,
ArrayRef<PrivateDataTy> Privates) {
bool InitRequired = false;
- for (auto &&Pair : Privates) {
- auto *VD = Pair.second.PrivateCopy;
- auto *Init = VD->getAnyInitializer();
+ for (const PrivateDataTy &Pair : Privates) {
+ const VarDecl *VD = Pair.second.PrivateCopy;
+ const Expr *Init = VD->getAnyInitializer();
InitRequired = InitRequired || (Init && isa<CXXConstructExpr>(Init) &&
!CGF.isTrivialInitializer(Init));
+ if (InitRequired)
+ break;
}
return InitRequired;
}
@@ -4297,7 +4688,7 @@ emitTaskDupFunction(CodeGenModule &CGM, SourceLocation Loc,
const RecordDecl *KmpTaskTQTyRD, QualType SharedsTy,
QualType SharedsPtrTy, const OMPTaskDataTy &Data,
ArrayRef<PrivateDataTy> Privates, bool WithLastIter) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl DstArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
KmpTaskTWithPrivatesPtrQTy,
@@ -4310,16 +4701,17 @@ emitTaskDupFunction(CodeGenModule &CGM, SourceLocation Loc,
Args.push_back(&DstArg);
Args.push_back(&SrcArg);
Args.push_back(&LastprivArg);
- auto &TaskDupFnInfo =
+ const auto &TaskDupFnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *TaskDupTy = CGM.getTypes().GetFunctionType(TaskDupFnInfo);
- auto *TaskDup =
- llvm::Function::Create(TaskDupTy, llvm::GlobalValue::InternalLinkage,
- ".omp_task_dup.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, TaskDup, TaskDupFnInfo);
+ llvm::FunctionType *TaskDupTy = CGM.getTypes().GetFunctionType(TaskDupFnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName({"omp_task_dup", ""});
+ auto *TaskDup = llvm::Function::Create(
+ TaskDupTy, llvm::GlobalValue::InternalLinkage, Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskDup, TaskDupFnInfo);
+ TaskDup->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
- CGF.StartFunction(GlobalDecl(), C.VoidTy, TaskDup, TaskDupFnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, TaskDup, TaskDupFnInfo, Args, Loc,
+ Loc);
LValue TDBase = CGF.EmitLoadOfPointerLValue(
CGF.GetAddrOfLocalVar(&DstArg),
@@ -4362,9 +4754,9 @@ emitTaskDupFunction(CodeGenModule &CGM, SourceLocation Loc,
static bool
checkDestructorsRequired(const RecordDecl *KmpTaskTWithPrivatesQTyRD) {
bool NeedsCleanup = false;
- auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
- auto *PrivateRD = cast<RecordDecl>(FI->getType()->getAsTagDecl());
- for (auto *FD : PrivateRD->fields()) {
+ auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin(), 1);
+ const auto *PrivateRD = cast<RecordDecl>(FI->getType()->getAsTagDecl());
+ for (const FieldDecl *FD : PrivateRD->fields()) {
NeedsCleanup = NeedsCleanup || FD->getType().isDestructedType();
if (NeedsCleanup)
break;
@@ -4377,41 +4769,41 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
const OMPExecutableDirective &D,
llvm::Value *TaskFunction, QualType SharedsTy,
Address Shareds, const OMPTaskDataTy &Data) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
llvm::SmallVector<PrivateDataTy, 4> Privates;
// Aggregate privates and sort them by the alignment.
auto I = Data.PrivateCopies.begin();
- for (auto *E : Data.PrivateVars) {
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
- Privates.push_back(std::make_pair(
+ for (const Expr *E : Data.PrivateVars) {
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ Privates.emplace_back(
C.getDeclAlign(VD),
PrivateHelpersTy(VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
- /*PrivateElemInit=*/nullptr)));
+ /*PrivateElemInit=*/nullptr));
++I;
}
I = Data.FirstprivateCopies.begin();
auto IElemInitRef = Data.FirstprivateInits.begin();
- for (auto *E : Data.FirstprivateVars) {
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
- Privates.push_back(std::make_pair(
+ for (const Expr *E : Data.FirstprivateVars) {
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ Privates.emplace_back(
C.getDeclAlign(VD),
PrivateHelpersTy(
VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
- cast<VarDecl>(cast<DeclRefExpr>(*IElemInitRef)->getDecl()))));
+ cast<VarDecl>(cast<DeclRefExpr>(*IElemInitRef)->getDecl())));
++I;
++IElemInitRef;
}
I = Data.LastprivateCopies.begin();
- for (auto *E : Data.LastprivateVars) {
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
- Privates.push_back(std::make_pair(
+ for (const Expr *E : Data.LastprivateVars) {
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
+ Privates.emplace_back(
C.getDeclAlign(VD),
PrivateHelpersTy(VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
- /*PrivateElemInit=*/nullptr)));
+ /*PrivateElemInit=*/nullptr));
++I;
}
std::stable_sort(Privates.begin(), Privates.end(), stable_sort_comparator);
- auto KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
+ QualType KmpInt32Ty = C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
// Build type kmp_routine_entry_t (if not built yet).
emitKmpRoutineEntryT(KmpInt32Ty);
// Build type kmp_task_t (if not built yet).
@@ -4432,21 +4824,23 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
}
KmpTaskTQTy = SavedKmpTaskTQTy;
}
- auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
+ const auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
// Build particular struct kmp_task_t for the given task.
- auto *KmpTaskTWithPrivatesQTyRD =
+ const RecordDecl *KmpTaskTWithPrivatesQTyRD =
createKmpTaskTWithPrivatesRecordDecl(CGM, KmpTaskTQTy, Privates);
- auto KmpTaskTWithPrivatesQTy = C.getRecordType(KmpTaskTWithPrivatesQTyRD);
+ QualType KmpTaskTWithPrivatesQTy = C.getRecordType(KmpTaskTWithPrivatesQTyRD);
QualType KmpTaskTWithPrivatesPtrQTy =
C.getPointerType(KmpTaskTWithPrivatesQTy);
- auto *KmpTaskTWithPrivatesTy = CGF.ConvertType(KmpTaskTWithPrivatesQTy);
- auto *KmpTaskTWithPrivatesPtrTy = KmpTaskTWithPrivatesTy->getPointerTo();
- auto *KmpTaskTWithPrivatesTySize = CGF.getTypeSize(KmpTaskTWithPrivatesQTy);
+ llvm::Type *KmpTaskTWithPrivatesTy = CGF.ConvertType(KmpTaskTWithPrivatesQTy);
+ llvm::Type *KmpTaskTWithPrivatesPtrTy =
+ KmpTaskTWithPrivatesTy->getPointerTo();
+ llvm::Value *KmpTaskTWithPrivatesTySize =
+ CGF.getTypeSize(KmpTaskTWithPrivatesQTy);
QualType SharedsPtrTy = C.getPointerType(SharedsTy);
// Emit initial values for private copies (if any).
llvm::Value *TaskPrivatesMap = nullptr;
- auto *TaskPrivatesMapTy =
+ llvm::Type *TaskPrivatesMapTy =
std::next(cast<llvm::Function>(TaskFunction)->arg_begin(), 3)->getType();
if (!Privates.empty()) {
auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
@@ -4461,7 +4855,7 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
}
// Build a proxy function kmp_int32 .omp_task_entry.(kmp_int32 gtid,
// kmp_task_t *tt);
- auto *TaskEntry = emitProxyTaskFunction(
+ llvm::Value *TaskEntry = emitProxyTaskFunction(
CGM, Loc, D.getDirectiveKind(), KmpInt32Ty, KmpTaskTWithPrivatesPtrQTy,
KmpTaskTWithPrivatesQTy, KmpTaskTQTy, SharedsPtrTy, TaskFunction,
TaskPrivatesMap);
@@ -4487,23 +4881,24 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
}
if (Data.Priority.getInt())
Flags = Flags | PriorityFlag;
- auto *TaskFlags =
+ llvm::Value *TaskFlags =
Data.Final.getPointer()
? CGF.Builder.CreateSelect(Data.Final.getPointer(),
CGF.Builder.getInt32(FinalFlag),
CGF.Builder.getInt32(/*C=*/0))
: CGF.Builder.getInt32(Data.Final.getInt() ? FinalFlag : 0);
TaskFlags = CGF.Builder.CreateOr(TaskFlags, CGF.Builder.getInt32(Flags));
- auto *SharedsSize = CGM.getSize(C.getTypeSizeInChars(SharedsTy));
+ llvm::Value *SharedsSize = CGM.getSize(C.getTypeSizeInChars(SharedsTy));
llvm::Value *AllocArgs[] = {emitUpdateLocation(CGF, Loc),
getThreadID(CGF, Loc), TaskFlags,
KmpTaskTWithPrivatesTySize, SharedsSize,
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
TaskEntry, KmpRoutineEntryPtrTy)};
- auto *NewTask = CGF.EmitRuntimeCall(
+ llvm::Value *NewTask = CGF.EmitRuntimeCall(
createRuntimeFunction(OMPRTL__kmpc_omp_task_alloc), AllocArgs);
- auto *NewTaskNewTaskTTy = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
- NewTask, KmpTaskTWithPrivatesPtrTy);
+ llvm::Value *NewTaskNewTaskTTy =
+ CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
+ NewTask, KmpTaskTWithPrivatesPtrTy);
LValue Base = CGF.MakeNaturalAlignAddrLValue(NewTaskNewTaskTTy,
KmpTaskTWithPrivatesQTy);
LValue TDBase =
@@ -4519,7 +4914,9 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
KmpTaskTShareds)),
Loc),
CGF.getNaturalTypeAlignment(SharedsTy));
- CGF.EmitAggregateCopy(KmpTaskSharedsPtr, Shareds, SharedsTy);
+ LValue Dest = CGF.MakeAddrLValue(KmpTaskSharedsPtr, SharedsTy);
+ LValue Src = CGF.MakeAddrLValue(Shareds, SharedsTy);
+ CGF.EmitAggregateCopy(Dest, Src, SharedsTy, AggValueSlot::DoesNotOverlap);
}
// Emit initial values for private copies (if any).
TaskResultTy Result;
@@ -4539,7 +4936,8 @@ CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
enum { Priority = 0, Destructors = 1 };
// Provide pointer to function with destructors for privates.
auto FI = std::next(KmpTaskTQTyRD->field_begin(), Data1);
- auto *KmpCmplrdataUD = (*FI)->getType()->getAsUnionType()->getDecl();
+ const RecordDecl *KmpCmplrdataUD =
+ (*FI)->getType()->getAsUnionType()->getDecl();
if (NeedsCleanup) {
llvm::Value *DestructorFn = emitDestructorsFunction(
CGM, Loc, KmpInt32Ty, KmpTaskTWithPrivatesPtrQTy,
@@ -4582,8 +4980,8 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
llvm::Value *TaskEntry = Result.TaskEntry;
llvm::Value *NewTaskNewTaskTTy = Result.NewTaskNewTaskTTy;
LValue TDBase = Result.TDBase;
- RecordDecl *KmpTaskTQTyRD = Result.KmpTaskTQTyRD;
- auto &C = CGM.getContext();
+ const RecordDecl *KmpTaskTQTyRD = Result.KmpTaskTQTyRD;
+ ASTContext &C = CGM.getContext();
// Process list of dependences.
Address DependenciesArray = Address::invalid();
unsigned NumDependencies = Data.Dependences.size();
@@ -4603,8 +5001,9 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
addFieldToRecordDecl(C, KmpDependInfoRD, FlagsTy);
KmpDependInfoRD->completeDefinition();
KmpDependInfoTy = C.getRecordType(KmpDependInfoRD);
- } else
+ } else {
KmpDependInfoRD = cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
+ }
CharUnits DependencySize = C.getTypeSizeInChars(KmpDependInfoTy);
// Define type kmp_depend_info[<Dependences.size()>];
QualType KmpDependInfoArrayTy = C.getConstantArrayType(
@@ -4613,12 +5012,13 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
// kmp_depend_info[<Dependences.size()>] deps;
DependenciesArray =
CGF.CreateMemTemp(KmpDependInfoArrayTy, ".dep.arr.addr");
- for (unsigned i = 0; i < NumDependencies; ++i) {
- const Expr *E = Data.Dependences[i].second;
- auto Addr = CGF.EmitLValue(E);
+ for (unsigned I = 0; I < NumDependencies; ++I) {
+ const Expr *E = Data.Dependences[I].second;
+ LValue Addr = CGF.EmitLValue(E);
llvm::Value *Size;
QualType Ty = E->getType();
- if (auto *ASE = dyn_cast<OMPArraySectionExpr>(E->IgnoreParenImpCasts())) {
+ if (const auto *ASE =
+ dyn_cast<OMPArraySectionExpr>(E->IgnoreParenImpCasts())) {
LValue UpAddrLVal =
CGF.EmitOMPArraySectionExpr(ASE, /*LowerBound=*/false);
llvm::Value *UpAddr =
@@ -4627,24 +5027,25 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
CGF.Builder.CreatePtrToInt(Addr.getPointer(), CGM.SizeTy);
llvm::Value *UpIntPtr = CGF.Builder.CreatePtrToInt(UpAddr, CGM.SizeTy);
Size = CGF.Builder.CreateNUWSub(UpIntPtr, LowIntPtr);
- } else
+ } else {
Size = CGF.getTypeSize(Ty);
- auto Base = CGF.MakeAddrLValue(
- CGF.Builder.CreateConstArrayGEP(DependenciesArray, i, DependencySize),
+ }
+ LValue Base = CGF.MakeAddrLValue(
+ CGF.Builder.CreateConstArrayGEP(DependenciesArray, I, DependencySize),
KmpDependInfoTy);
// deps[i].base_addr = &<Dependences[i].second>;
- auto BaseAddrLVal = CGF.EmitLValueForField(
+ LValue BaseAddrLVal = CGF.EmitLValueForField(
Base, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
CGF.EmitStoreOfScalar(
CGF.Builder.CreatePtrToInt(Addr.getPointer(), CGF.IntPtrTy),
BaseAddrLVal);
// deps[i].len = sizeof(<Dependences[i].second>);
- auto LenLVal = CGF.EmitLValueForField(
+ LValue LenLVal = CGF.EmitLValueForField(
Base, *std::next(KmpDependInfoRD->field_begin(), Len));
CGF.EmitStoreOfScalar(Size, LenLVal);
// deps[i].flags = <Dependences[i].first>;
RTLDependenceKindTy DepKind;
- switch (Data.Dependences[i].first) {
+ switch (Data.Dependences[I].first) {
case OMPC_DEPEND_in:
DepKind = DepIn;
break;
@@ -4658,7 +5059,7 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
case OMPC_DEPEND_unknown:
llvm_unreachable("Unknown task dependence type");
}
- auto FlagsLVal = CGF.EmitLValueForField(
+ LValue FlagsLVal = CGF.EmitLValueForField(
Base, *std::next(KmpDependInfoRD->field_begin(), Flags));
CGF.EmitStoreOfScalar(llvm::ConstantInt::get(LLVMFlagsTy, DepKind),
FlagsLVal);
@@ -4668,14 +5069,14 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
CGF.VoidPtrTy);
}
- // NOTE: routine and part_id fields are intialized by __kmpc_omp_task_alloc()
+ // NOTE: routine and part_id fields are initialized by __kmpc_omp_task_alloc()
// libcall.
// Build kmp_int32 __kmpc_omp_task_with_deps(ident_t *, kmp_int32 gtid,
// kmp_task_t *new_task, kmp_int32 ndeps, kmp_depend_info_t *dep_list,
// kmp_int32 ndeps_noalias, kmp_depend_info_t *noalias_dep_list) if dependence
// list is not empty
- auto *ThreadID = getThreadID(CGF, Loc);
- auto *UpLoc = emitUpdateLocation(CGF, Loc);
+ llvm::Value *ThreadID = getThreadID(CGF, Loc);
+ llvm::Value *UpLoc = emitUpdateLocation(CGF, Loc);
llvm::Value *TaskArgs[] = { UpLoc, ThreadID, NewTask };
llvm::Value *DepTaskArgs[7];
if (NumDependencies) {
@@ -4692,7 +5093,7 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
&DepTaskArgs](CodeGenFunction &CGF, PrePostActionTy &) {
if (!Data.Tied) {
auto PartIdFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTPartId);
- auto PartIdLVal = CGF.EmitLValueForField(TDBase, *PartIdFI);
+ LValue PartIdLVal = CGF.EmitLValueForField(TDBase, *PartIdFI);
CGF.EmitStoreOfScalar(CGF.Builder.getInt32(0), PartIdLVal);
}
if (NumDependencies) {
@@ -4720,7 +5121,7 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
auto &&ElseCodeGen = [&TaskArgs, ThreadID, NewTaskNewTaskTTy, TaskEntry,
NumDependencies, &DepWaitTaskArgs,
Loc](CodeGenFunction &CGF, PrePostActionTy &) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
CodeGenFunction::RunCleanupsScope LocalScope(CGF);
// Build void __kmpc_omp_wait_deps(ident_t *, kmp_int32 gtid,
// kmp_int32 ndeps, kmp_depend_info_t *dep_list, kmp_int32
@@ -4750,9 +5151,9 @@ void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
RCG(CGF);
};
- if (IfCond)
+ if (IfCond) {
emitOMPIfClause(CGF, IfCond, ThenCodeGen, ElseCodeGen);
- else {
+ } else {
RegionCodeGenTy ThenRCG(ThenCodeGen);
ThenRCG(CGF);
}
@@ -4768,7 +5169,7 @@ void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
return;
TaskResultTy Result =
emitTaskInit(CGF, Loc, D, TaskFunction, SharedsTy, Shareds, Data);
- // NOTE: routine and part_id fields are intialized by __kmpc_omp_task_alloc()
+ // NOTE: routine and part_id fields are initialized by __kmpc_omp_task_alloc()
// libcall.
// Call to void __kmpc_taskloop(ident_t *loc, int gtid, kmp_task_t *task, int
// if_val, kmp_uint64 *lb, kmp_uint64 *ub, kmp_int64 st, int nogroup, int
@@ -4779,27 +5180,28 @@ void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
if (IfCond) {
IfVal = CGF.Builder.CreateIntCast(CGF.EvaluateExprAsBool(IfCond), CGF.IntTy,
/*isSigned=*/true);
- } else
+ } else {
IfVal = llvm::ConstantInt::getSigned(CGF.IntTy, /*V=*/1);
+ }
LValue LBLVal = CGF.EmitLValueForField(
Result.TDBase,
*std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTLowerBound));
- auto *LBVar =
+ const auto *LBVar =
cast<VarDecl>(cast<DeclRefExpr>(D.getLowerBoundVariable())->getDecl());
CGF.EmitAnyExprToMem(LBVar->getInit(), LBLVal.getAddress(), LBLVal.getQuals(),
/*IsInitializer=*/true);
LValue UBLVal = CGF.EmitLValueForField(
Result.TDBase,
*std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTUpperBound));
- auto *UBVar =
+ const auto *UBVar =
cast<VarDecl>(cast<DeclRefExpr>(D.getUpperBoundVariable())->getDecl());
CGF.EmitAnyExprToMem(UBVar->getInit(), UBLVal.getAddress(), UBLVal.getQuals(),
/*IsInitializer=*/true);
LValue StLVal = CGF.EmitLValueForField(
Result.TDBase,
*std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTStride));
- auto *StVar =
+ const auto *StVar =
cast<VarDecl>(cast<DeclRefExpr>(D.getStrideVariable())->getDecl());
CGF.EmitAnyExprToMem(StVar->getInit(), StLVal.getAddress(), StLVal.getQuals(),
/*IsInitializer=*/true);
@@ -4807,9 +5209,9 @@ void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
LValue RedLVal = CGF.EmitLValueForField(
Result.TDBase,
*std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTReductions));
- if (Data.Reductions)
+ if (Data.Reductions) {
CGF.EmitStoreOfScalar(Data.Reductions, RedLVal);
- else {
+ } else {
CGF.EmitNullInitialization(RedLVal.getAddress(),
CGF.getContext().VoidPtrTy);
}
@@ -4821,7 +5223,7 @@ void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
IfVal,
LBLVal.getPointer(),
UBLVal.getPointer(),
- CGF.EmitLoadOfScalar(StLVal, SourceLocation()),
+ CGF.EmitLoadOfScalar(StLVal, Loc),
llvm::ConstantInt::getNullValue(
CGF.IntTy), // Always 0 because taskgroup emitted by the compiler
llvm::ConstantInt::getSigned(
@@ -4838,7 +5240,7 @@ void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
CGF.EmitRuntimeCall(createRuntimeFunction(OMPRTL__kmpc_taskloop), TaskArgs);
}
-/// \brief Emit reduction operation for each element of array (required for
+/// Emit reduction operation for each element of array (required for
/// array sections) LHS op = RHS.
/// \param Type Type of array.
/// \param LHSVar Variable on the left side of the reduction operation
@@ -4860,22 +5262,22 @@ static void EmitOMPAggregateReduction(
Address RHSAddr = CGF.GetAddrOfLocalVar(RHSVar);
// Drill down to the base element type on both arrays.
- auto ArrayTy = Type->getAsArrayTypeUnsafe();
- auto NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, LHSAddr);
+ const ArrayType *ArrayTy = Type->getAsArrayTypeUnsafe();
+ llvm::Value *NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, LHSAddr);
- auto RHSBegin = RHSAddr.getPointer();
- auto LHSBegin = LHSAddr.getPointer();
+ llvm::Value *RHSBegin = RHSAddr.getPointer();
+ llvm::Value *LHSBegin = LHSAddr.getPointer();
// Cast from pointer to array type to pointer to single element.
- auto LHSEnd = CGF.Builder.CreateGEP(LHSBegin, NumElements);
+ llvm::Value *LHSEnd = CGF.Builder.CreateGEP(LHSBegin, NumElements);
// The basic structure here is a while-do loop.
- auto BodyBB = CGF.createBasicBlock("omp.arraycpy.body");
- auto DoneBB = CGF.createBasicBlock("omp.arraycpy.done");
- auto IsEmpty =
+ llvm::BasicBlock *BodyBB = CGF.createBasicBlock("omp.arraycpy.body");
+ llvm::BasicBlock *DoneBB = CGF.createBasicBlock("omp.arraycpy.done");
+ llvm::Value *IsEmpty =
CGF.Builder.CreateICmpEQ(LHSBegin, LHSEnd, "omp.arraycpy.isempty");
CGF.Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);
// Enter the loop body, making that address the current address.
- auto EntryBB = CGF.Builder.GetInsertBlock();
+ llvm::BasicBlock *EntryBB = CGF.Builder.GetInsertBlock();
CGF.EmitBlock(BodyBB);
CharUnits ElementSize = CGF.getContext().getTypeSizeInChars(ElementTy);
@@ -4896,19 +5298,19 @@ static void EmitOMPAggregateReduction(
// Emit copy.
CodeGenFunction::OMPPrivateScope Scope(CGF);
- Scope.addPrivate(LHSVar, [=]() -> Address { return LHSElementCurrent; });
- Scope.addPrivate(RHSVar, [=]() -> Address { return RHSElementCurrent; });
+ Scope.addPrivate(LHSVar, [=]() { return LHSElementCurrent; });
+ Scope.addPrivate(RHSVar, [=]() { return RHSElementCurrent; });
Scope.Privatize();
RedOpGen(CGF, XExpr, EExpr, UpExpr);
Scope.ForceCleanup();
// Shift the address forward by one element.
- auto LHSElementNext = CGF.Builder.CreateConstGEP1_32(
+ llvm::Value *LHSElementNext = CGF.Builder.CreateConstGEP1_32(
LHSElementPHI, /*Idx0=*/1, "omp.arraycpy.dest.element");
- auto RHSElementNext = CGF.Builder.CreateConstGEP1_32(
+ llvm::Value *RHSElementNext = CGF.Builder.CreateConstGEP1_32(
RHSElementPHI, /*Idx0=*/1, "omp.arraycpy.src.element");
// Check whether we've reached the end.
- auto Done =
+ llvm::Value *Done =
CGF.Builder.CreateICmpEQ(LHSElementNext, LHSEnd, "omp.arraycpy.done");
CGF.Builder.CreateCondBr(Done, DoneBB, BodyBB);
LHSElementPHI->addIncoming(LHSElementNext, CGF.Builder.GetInsertBlock());
@@ -4923,11 +5325,12 @@ static void EmitOMPAggregateReduction(
/// UDR combiner function.
static void emitReductionCombiner(CodeGenFunction &CGF,
const Expr *ReductionOp) {
- if (auto *CE = dyn_cast<CallExpr>(ReductionOp))
- if (auto *OVE = dyn_cast<OpaqueValueExpr>(CE->getCallee()))
- if (auto *DRE =
+ if (const auto *CE = dyn_cast<CallExpr>(ReductionOp))
+ if (const auto *OVE = dyn_cast<OpaqueValueExpr>(CE->getCallee()))
+ if (const auto *DRE =
dyn_cast<DeclRefExpr>(OVE->getSourceExpr()->IgnoreImpCasts()))
- if (auto *DRD = dyn_cast<OMPDeclareReductionDecl>(DRE->getDecl())) {
+ if (const auto *DRD =
+ dyn_cast<OMPDeclareReductionDecl>(DRE->getDecl())) {
std::pair<llvm::Function *, llvm::Function *> Reduction =
CGF.CGM.getOpenMPRuntime().getUserDefinedReduction(DRD);
RValue Func = RValue::get(Reduction.first);
@@ -4939,24 +5342,29 @@ static void emitReductionCombiner(CodeGenFunction &CGF,
}
llvm::Value *CGOpenMPRuntime::emitReductionFunction(
- CodeGenModule &CGM, llvm::Type *ArgsType, ArrayRef<const Expr *> Privates,
- ArrayRef<const Expr *> LHSExprs, ArrayRef<const Expr *> RHSExprs,
- ArrayRef<const Expr *> ReductionOps) {
- auto &C = CGM.getContext();
+ CodeGenModule &CGM, SourceLocation Loc, llvm::Type *ArgsType,
+ ArrayRef<const Expr *> Privates, ArrayRef<const Expr *> LHSExprs,
+ ArrayRef<const Expr *> RHSExprs, ArrayRef<const Expr *> ReductionOps) {
+ ASTContext &C = CGM.getContext();
// void reduction_func(void *LHSArg, void *RHSArg);
FunctionArgList Args;
- ImplicitParamDecl LHSArg(C, C.VoidPtrTy, ImplicitParamDecl::Other);
- ImplicitParamDecl RHSArg(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl LHSArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
+ ImplicitParamDecl RHSArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
Args.push_back(&LHSArg);
Args.push_back(&RHSArg);
- auto &CGFI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *Fn = llvm::Function::Create(
- CGM.getTypes().GetFunctionType(CGFI), llvm::GlobalValue::InternalLinkage,
- ".omp.reduction.reduction_func", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, CGFI);
+ const auto &CGFI =
+ CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
+ std::string Name = getName({"omp", "reduction", "reduction_func"});
+ auto *Fn = llvm::Function::Create(CGM.getTypes().GetFunctionType(CGFI),
+ llvm::GlobalValue::InternalLinkage, Name,
+ &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, CGFI);
+ Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, CGFI, Args, Loc, Loc);
// Dst = (void*[n])(LHSArg);
// Src = (void*[n])(RHSArg);
@@ -4974,12 +5382,14 @@ llvm::Value *CGOpenMPRuntime::emitReductionFunction(
auto IPriv = Privates.begin();
unsigned Idx = 0;
for (unsigned I = 0, E = ReductionOps.size(); I < E; ++I, ++IPriv, ++Idx) {
- auto RHSVar = cast<VarDecl>(cast<DeclRefExpr>(RHSExprs[I])->getDecl());
- Scope.addPrivate(RHSVar, [&]() -> Address {
+ const auto *RHSVar =
+ cast<VarDecl>(cast<DeclRefExpr>(RHSExprs[I])->getDecl());
+ Scope.addPrivate(RHSVar, [&CGF, RHS, Idx, RHSVar]() {
return emitAddrOfVarFromArray(CGF, RHS, Idx, RHSVar);
});
- auto LHSVar = cast<VarDecl>(cast<DeclRefExpr>(LHSExprs[I])->getDecl());
- Scope.addPrivate(LHSVar, [&]() -> Address {
+ const auto *LHSVar =
+ cast<VarDecl>(cast<DeclRefExpr>(LHSExprs[I])->getDecl());
+ Scope.addPrivate(LHSVar, [&CGF, LHS, Idx, LHSVar]() {
return emitAddrOfVarFromArray(CGF, LHS, Idx, LHSVar);
});
QualType PrivTy = (*IPriv)->getType();
@@ -4989,8 +5399,9 @@ llvm::Value *CGOpenMPRuntime::emitReductionFunction(
Address Elem =
CGF.Builder.CreateConstArrayGEP(LHS, Idx, CGF.getPointerSize());
llvm::Value *Ptr = CGF.Builder.CreateLoad(Elem);
- auto *VLA = CGF.getContext().getAsVariableArrayType(PrivTy);
- auto *OVE = cast<OpaqueValueExpr>(VLA->getSizeExpr());
+ const VariableArrayType *VLA =
+ CGF.getContext().getAsVariableArrayType(PrivTy);
+ const auto *OVE = cast<OpaqueValueExpr>(VLA->getSizeExpr());
CodeGenFunction::OpaqueValueMapping OpaqueMap(
CGF, OVE, RValue::get(CGF.Builder.CreatePtrToInt(Ptr, CGF.SizeTy)));
CGF.EmitVariablyModifiedType(PrivTy);
@@ -5000,19 +5411,20 @@ llvm::Value *CGOpenMPRuntime::emitReductionFunction(
IPriv = Privates.begin();
auto ILHS = LHSExprs.begin();
auto IRHS = RHSExprs.begin();
- for (auto *E : ReductionOps) {
+ for (const Expr *E : ReductionOps) {
if ((*IPriv)->getType()->isArrayType()) {
// Emit reduction for array section.
- auto *LHSVar = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
- auto *RHSVar = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
+ const auto *LHSVar = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
+ const auto *RHSVar = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
EmitOMPAggregateReduction(
CGF, (*IPriv)->getType(), LHSVar, RHSVar,
[=](CodeGenFunction &CGF, const Expr *, const Expr *, const Expr *) {
emitReductionCombiner(CGF, E);
});
- } else
+ } else {
// Emit reduction for array subscript or single variable.
emitReductionCombiner(CGF, E);
+ }
++IPriv;
++ILHS;
++IRHS;
@@ -5029,16 +5441,17 @@ void CGOpenMPRuntime::emitSingleReductionCombiner(CodeGenFunction &CGF,
const DeclRefExpr *RHS) {
if (PrivateRef->getType()->isArrayType()) {
// Emit reduction for array section.
- auto *LHSVar = cast<VarDecl>(LHS->getDecl());
- auto *RHSVar = cast<VarDecl>(RHS->getDecl());
+ const auto *LHSVar = cast<VarDecl>(LHS->getDecl());
+ const auto *RHSVar = cast<VarDecl>(RHS->getDecl());
EmitOMPAggregateReduction(
CGF, PrivateRef->getType(), LHSVar, RHSVar,
[=](CodeGenFunction &CGF, const Expr *, const Expr *, const Expr *) {
emitReductionCombiner(CGF, ReductionOp);
});
- } else
+ } else {
// Emit reduction for array subscript or single variable.
emitReductionCombiner(CGF, ReductionOp);
+ }
}
void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
@@ -5088,14 +5501,14 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
// <LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]);
// ...
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
if (SimpleReduction) {
CodeGenFunction::RunCleanupsScope Scope(CGF);
auto IPriv = Privates.begin();
auto ILHS = LHSExprs.begin();
auto IRHS = RHSExprs.begin();
- for (auto *E : ReductionOps) {
+ for (const Expr *E : ReductionOps) {
emitSingleReductionCombiner(CGF, E, *IPriv, cast<DeclRefExpr>(*ILHS),
cast<DeclRefExpr>(*IRHS));
++IPriv;
@@ -5108,7 +5521,7 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
// 1. Build a list of reduction variables.
// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
auto Size = RHSExprs.size();
- for (auto *E : Privates) {
+ for (const Expr *E : Privates) {
if (E->getType()->isVariablyModifiedType())
// Reserve place for array size.
++Size;
@@ -5136,7 +5549,7 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
llvm::Value *Size = CGF.Builder.CreateIntCast(
CGF.getVLASize(
CGF.getContext().getAsVariableArrayType((*IPriv)->getType()))
- .first,
+ .NumElts,
CGF.SizeTy, /*isSigned=*/false);
CGF.Builder.CreateStore(CGF.Builder.CreateIntToPtr(Size, CGF.VoidPtrTy),
Elem);
@@ -5144,19 +5557,20 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
}
// 2. Emit reduce_func().
- auto *ReductionFn = emitReductionFunction(
- CGM, CGF.ConvertTypeForMem(ReductionArrayTy)->getPointerTo(), Privates,
- LHSExprs, RHSExprs, ReductionOps);
+ llvm::Value *ReductionFn = emitReductionFunction(
+ CGM, Loc, CGF.ConvertTypeForMem(ReductionArrayTy)->getPointerTo(),
+ Privates, LHSExprs, RHSExprs, ReductionOps);
// 3. Create static kmp_critical_name lock = { 0 };
- auto *Lock = getCriticalRegionLock(".reduction");
+ std::string Name = getName({"reduction"});
+ llvm::Value *Lock = getCriticalRegionLock(Name);
// 4. Build res = __kmpc_reduce{_nowait}(<loc>, <gtid>, <n>, sizeof(RedList),
// RedList, reduce_func, &<lock>);
- auto *IdentTLoc = emitUpdateLocation(CGF, Loc, OMP_ATOMIC_REDUCE);
- auto *ThreadId = getThreadID(CGF, Loc);
- auto *ReductionArrayTySize = CGF.getTypeSize(ReductionArrayTy);
- auto *RL = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
+ llvm::Value *IdentTLoc = emitUpdateLocation(CGF, Loc, OMP_ATOMIC_REDUCE);
+ llvm::Value *ThreadId = getThreadID(CGF, Loc);
+ llvm::Value *ReductionArrayTySize = CGF.getTypeSize(ReductionArrayTy);
+ llvm::Value *RL = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
ReductionList.getPointer(), CGF.VoidPtrTy);
llvm::Value *Args[] = {
IdentTLoc, // ident_t *<loc>
@@ -5167,14 +5581,15 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
ReductionFn, // void (*) (void *, void *) <reduce_func>
Lock // kmp_critical_name *&<lock>
};
- auto Res = CGF.EmitRuntimeCall(
+ llvm::Value *Res = CGF.EmitRuntimeCall(
createRuntimeFunction(WithNowait ? OMPRTL__kmpc_reduce_nowait
: OMPRTL__kmpc_reduce),
Args);
// 5. Build switch(res)
- auto *DefaultBB = CGF.createBasicBlock(".omp.reduction.default");
- auto *SwInst = CGF.Builder.CreateSwitch(Res, DefaultBB, /*NumCases=*/2);
+ llvm::BasicBlock *DefaultBB = CGF.createBasicBlock(".omp.reduction.default");
+ llvm::SwitchInst *SwInst =
+ CGF.Builder.CreateSwitch(Res, DefaultBB, /*NumCases=*/2);
// 6. Build case 1:
// ...
@@ -5182,7 +5597,7 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
// ...
// __kmpc_end_reduce{_nowait}(<loc>, <gtid>, &<lock>);
// break;
- auto *Case1BB = CGF.createBasicBlock(".omp.reduction.case1");
+ llvm::BasicBlock *Case1BB = CGF.createBasicBlock(".omp.reduction.case1");
SwInst->addCase(CGF.Builder.getInt32(1), Case1BB);
CGF.EmitBlock(Case1BB);
@@ -5192,13 +5607,13 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
ThreadId, // i32 <gtid>
Lock // kmp_critical_name *&<lock>
};
- auto &&CodeGen = [&Privates, &LHSExprs, &RHSExprs, &ReductionOps](
- CodeGenFunction &CGF, PrePostActionTy &Action) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
+ auto &&CodeGen = [Privates, LHSExprs, RHSExprs, ReductionOps](
+ CodeGenFunction &CGF, PrePostActionTy &Action) {
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
auto IPriv = Privates.begin();
auto ILHS = LHSExprs.begin();
auto IRHS = RHSExprs.begin();
- for (auto *E : ReductionOps) {
+ for (const Expr *E : ReductionOps) {
RT.emitSingleReductionCombiner(CGF, E, *IPriv, cast<DeclRefExpr>(*ILHS),
cast<DeclRefExpr>(*IRHS));
++IPriv;
@@ -5222,44 +5637,44 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
// Atomic(<LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]));
// ...
// break;
- auto *Case2BB = CGF.createBasicBlock(".omp.reduction.case2");
+ llvm::BasicBlock *Case2BB = CGF.createBasicBlock(".omp.reduction.case2");
SwInst->addCase(CGF.Builder.getInt32(2), Case2BB);
CGF.EmitBlock(Case2BB);
- auto &&AtomicCodeGen = [Loc, &Privates, &LHSExprs, &RHSExprs, &ReductionOps](
- CodeGenFunction &CGF, PrePostActionTy &Action) {
+ auto &&AtomicCodeGen = [Loc, Privates, LHSExprs, RHSExprs, ReductionOps](
+ CodeGenFunction &CGF, PrePostActionTy &Action) {
auto ILHS = LHSExprs.begin();
auto IRHS = RHSExprs.begin();
auto IPriv = Privates.begin();
- for (auto *E : ReductionOps) {
+ for (const Expr *E : ReductionOps) {
const Expr *XExpr = nullptr;
const Expr *EExpr = nullptr;
const Expr *UpExpr = nullptr;
BinaryOperatorKind BO = BO_Comma;
- if (auto *BO = dyn_cast<BinaryOperator>(E)) {
+ if (const auto *BO = dyn_cast<BinaryOperator>(E)) {
if (BO->getOpcode() == BO_Assign) {
XExpr = BO->getLHS();
UpExpr = BO->getRHS();
}
}
// Try to emit update expression as a simple atomic.
- auto *RHSExpr = UpExpr;
+ const Expr *RHSExpr = UpExpr;
if (RHSExpr) {
// Analyze RHS part of the whole expression.
- if (auto *ACO = dyn_cast<AbstractConditionalOperator>(
+ if (const auto *ACO = dyn_cast<AbstractConditionalOperator>(
RHSExpr->IgnoreParenImpCasts())) {
// If this is a conditional operator, analyze its condition for
// min/max reduction operator.
RHSExpr = ACO->getCond();
}
- if (auto *BORHS =
+ if (const auto *BORHS =
dyn_cast<BinaryOperator>(RHSExpr->IgnoreParenImpCasts())) {
EExpr = BORHS->getRHS();
BO = BORHS->getOpcode();
}
}
if (XExpr) {
- auto *VD = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
+ const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
auto &&AtomicRedGen = [BO, VD,
Loc](CodeGenFunction &CGF, const Expr *XExpr,
const Expr *EExpr, const Expr *UpExpr) {
@@ -5273,7 +5688,7 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
[&CGF, UpExpr, VD, Loc](RValue XRValue) {
CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
PrivateScope.addPrivate(
- VD, [&CGF, VD, XRValue, Loc]() -> Address {
+ VD, [&CGF, VD, XRValue, Loc]() {
Address LHSTemp = CGF.CreateMemTemp(VD->getType());
CGF.emitOMPSimpleStore(
CGF.MakeAddrLValue(LHSTemp, VD->getType()), XRValue,
@@ -5286,19 +5701,22 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
};
if ((*IPriv)->getType()->isArrayType()) {
// Emit atomic reduction for array section.
- auto *RHSVar = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
+ const auto *RHSVar =
+ cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
EmitOMPAggregateReduction(CGF, (*IPriv)->getType(), VD, RHSVar,
AtomicRedGen, XExpr, EExpr, UpExpr);
- } else
+ } else {
// Emit atomic reduction for array subscript or single variable.
AtomicRedGen(CGF, XExpr, EExpr, UpExpr);
+ }
} else {
// Emit as a critical region.
auto &&CritRedGen = [E, Loc](CodeGenFunction &CGF, const Expr *,
- const Expr *, const Expr *) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
+ const Expr *, const Expr *) {
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
+ std::string Name = RT.getName({"atomic_reduction"});
RT.emitCriticalRegion(
- CGF, ".atomic_reduction",
+ CGF, Name,
[=](CodeGenFunction &CGF, PrePostActionTy &Action) {
Action.Enter(CGF);
emitReductionCombiner(CGF, E);
@@ -5306,12 +5724,15 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
Loc);
};
if ((*IPriv)->getType()->isArrayType()) {
- auto *LHSVar = cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
- auto *RHSVar = cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
+ const auto *LHSVar =
+ cast<VarDecl>(cast<DeclRefExpr>(*ILHS)->getDecl());
+ const auto *RHSVar =
+ cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
EmitOMPAggregateReduction(CGF, (*IPriv)->getType(), LHSVar, RHSVar,
CritRedGen);
- } else
+ } else {
CritRedGen(CGF, nullptr, nullptr, nullptr);
+ }
}
++ILHS;
++IRHS;
@@ -5331,20 +5752,29 @@ void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
EndArgs);
AtomicRCG.setAction(Action);
AtomicRCG(CGF);
- } else
+ } else {
AtomicRCG(CGF);
+ }
CGF.EmitBranch(DefaultBB);
CGF.EmitBlock(DefaultBB, /*IsFinished=*/true);
}
/// Generates unique name for artificial threadprivate variables.
-/// Format is: <Prefix> "." <Loc_raw_encoding> "_" <N>
-static std::string generateUniqueName(StringRef Prefix, SourceLocation Loc,
- unsigned N) {
+/// Format is: <Prefix> "." <Decl_mangled_name> "_" "<Decl_start_loc_raw_enc>"
+static std::string generateUniqueName(CodeGenModule &CGM, StringRef Prefix,
+ const Expr *Ref) {
SmallString<256> Buffer;
llvm::raw_svector_ostream Out(Buffer);
- Out << Prefix << "." << Loc.getRawEncoding() << "_" << N;
+ const clang::DeclRefExpr *DE;
+ const VarDecl *D = ::getBaseDecl(Ref, DE);
+ if (!D)
+ D = cast<VarDecl>(cast<DeclRefExpr>(Ref)->getDecl());
+ D = D->getCanonicalDecl();
+ std::string Name = CGM.getOpenMPRuntime().getName(
+ {D->isLocalVarDeclOrParm() ? D->getName() : CGM.getMangledName(D)});
+ Out << Prefix << Name << "_"
+ << D->getCanonicalDecl()->getLocStart().getRawEncoding();
return Out.str();
}
@@ -5359,19 +5789,21 @@ static std::string generateUniqueName(StringRef Prefix, SourceLocation Loc,
static llvm::Value *emitReduceInitFunction(CodeGenModule &CGM,
SourceLocation Loc,
ReductionCodeGen &RCG, unsigned N) {
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
- ImplicitParamDecl Param(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl Param(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
Args.emplace_back(&Param);
- auto &FnInfo =
+ const auto &FnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName({"red_init", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
- ".red_init.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, FnInfo);
+ Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+ Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
Address PrivateAddr = CGF.EmitLoadOfPointer(
CGF.GetAddrOfLocalVar(&Param),
C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
@@ -5381,10 +5813,9 @@ static llvm::Value *emitReduceInitFunction(CodeGenModule &CGM,
if (RCG.getSizes(N).second) {
Address SizeAddr = CGM.getOpenMPRuntime().getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().getSizeType(),
- generateUniqueName("reduction_size", Loc, N));
- Size =
- CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
- CGM.getContext().getSizeType(), SourceLocation());
+ generateUniqueName(CGM, "reduction_size", RCG.getRefExpr(N)));
+ Size = CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
+ CGM.getContext().getSizeType(), Loc);
}
RCG.emitAggregateType(CGF, N, Size);
LValue SharedLVal;
@@ -5395,7 +5826,10 @@ static llvm::Value *emitReduceInitFunction(CodeGenModule &CGM,
Address SharedAddr =
CGM.getOpenMPRuntime().getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().VoidPtrTy,
- generateUniqueName("reduction", Loc, N));
+ generateUniqueName(CGM, "reduction", RCG.getRefExpr(N)));
+ SharedAddr = CGF.EmitLoadOfPointer(
+ SharedAddr,
+ CGM.getContext().VoidPtrTy.castAs<PointerType>()->getTypePtr());
SharedLVal = CGF.MakeAddrLValue(SharedAddr, CGM.getContext().VoidPtrTy);
} else {
SharedLVal = CGF.MakeNaturalAlignAddrLValue(
@@ -5427,40 +5861,42 @@ static llvm::Value *emitReduceCombFunction(CodeGenModule &CGM,
const Expr *ReductionOp,
const Expr *LHS, const Expr *RHS,
const Expr *PrivateRef) {
- auto &C = CGM.getContext();
- auto *LHSVD = cast<VarDecl>(cast<DeclRefExpr>(LHS)->getDecl());
- auto *RHSVD = cast<VarDecl>(cast<DeclRefExpr>(RHS)->getDecl());
+ ASTContext &C = CGM.getContext();
+ const auto *LHSVD = cast<VarDecl>(cast<DeclRefExpr>(LHS)->getDecl());
+ const auto *RHSVD = cast<VarDecl>(cast<DeclRefExpr>(RHS)->getDecl());
FunctionArgList Args;
- ImplicitParamDecl ParamInOut(C, C.VoidPtrTy, ImplicitParamDecl::Other);
- ImplicitParamDecl ParamIn(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl ParamInOut(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
+ C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl ParamIn(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
Args.emplace_back(&ParamInOut);
Args.emplace_back(&ParamIn);
- auto &FnInfo =
+ const auto &FnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName({"red_comb", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
- ".red_comb.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, FnInfo);
+ Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+ Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
llvm::Value *Size = nullptr;
// If the size of the reduction item is non-constant, load it from global
// threadprivate variable.
if (RCG.getSizes(N).second) {
Address SizeAddr = CGM.getOpenMPRuntime().getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().getSizeType(),
- generateUniqueName("reduction_size", Loc, N));
- Size =
- CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
- CGM.getContext().getSizeType(), SourceLocation());
+ generateUniqueName(CGM, "reduction_size", RCG.getRefExpr(N)));
+ Size = CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
+ CGM.getContext().getSizeType(), Loc);
}
RCG.emitAggregateType(CGF, N, Size);
// Remap lhs and rhs variables to the addresses of the function arguments.
// %lhs = bitcast void* %arg0 to <type>*
// %rhs = bitcast void* %arg1 to <type>*
CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
- PrivateScope.addPrivate(LHSVD, [&C, &CGF, &ParamInOut, LHSVD]() -> Address {
+ PrivateScope.addPrivate(LHSVD, [&C, &CGF, &ParamInOut, LHSVD]() {
// Pull out the pointer to the variable.
Address PtrAddr = CGF.EmitLoadOfPointer(
CGF.GetAddrOfLocalVar(&ParamInOut),
@@ -5468,7 +5904,7 @@ static llvm::Value *emitReduceCombFunction(CodeGenModule &CGM,
return CGF.Builder.CreateElementBitCast(
PtrAddr, CGF.ConvertTypeForMem(LHSVD->getType()));
});
- PrivateScope.addPrivate(RHSVD, [&C, &CGF, &ParamIn, RHSVD]() -> Address {
+ PrivateScope.addPrivate(RHSVD, [&C, &CGF, &ParamIn, RHSVD]() {
// Pull out the pointer to the variable.
Address PtrAddr = CGF.EmitLoadOfPointer(
CGF.GetAddrOfLocalVar(&ParamIn),
@@ -5500,19 +5936,21 @@ static llvm::Value *emitReduceFiniFunction(CodeGenModule &CGM,
ReductionCodeGen &RCG, unsigned N) {
if (!RCG.needCleanups(N))
return nullptr;
- auto &C = CGM.getContext();
+ ASTContext &C = CGM.getContext();
FunctionArgList Args;
- ImplicitParamDecl Param(C, C.VoidPtrTy, ImplicitParamDecl::Other);
+ ImplicitParamDecl Param(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
+ ImplicitParamDecl::Other);
Args.emplace_back(&Param);
- auto &FnInfo =
+ const auto &FnInfo =
CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
- auto *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
+ std::string Name = CGM.getOpenMPRuntime().getName({"red_fini", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
- ".red_fini.", &CGM.getModule());
- CGM.SetInternalFunctionAttributes(/*D=*/nullptr, Fn, FnInfo);
+ Name, &CGM.getModule());
+ CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
+ Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
- CGF.disableDebugInfo();
- CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args);
+ CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
Address PrivateAddr = CGF.EmitLoadOfPointer(
CGF.GetAddrOfLocalVar(&Param),
C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
@@ -5522,10 +5960,9 @@ static llvm::Value *emitReduceFiniFunction(CodeGenModule &CGM,
if (RCG.getSizes(N).second) {
Address SizeAddr = CGM.getOpenMPRuntime().getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().getSizeType(),
- generateUniqueName("reduction_size", Loc, N));
- Size =
- CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
- CGM.getContext().getSizeType(), SourceLocation());
+ generateUniqueName(CGM, "reduction_size", RCG.getRefExpr(N)));
+ Size = CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
+ CGM.getContext().getSizeType(), Loc);
}
RCG.emitAggregateType(CGF, N, Size);
// Emit the finalizer body:
@@ -5551,7 +5988,7 @@ llvm::Value *CGOpenMPRuntime::emitTaskReductionInit(
// kmp_task_red_flags_t flags; // flags for additional info from compiler
// } kmp_task_red_input_t;
ASTContext &C = CGM.getContext();
- auto *RD = C.buildImplicitRecord("kmp_task_red_input_t");
+ RecordDecl *RD = C.buildImplicitRecord("kmp_task_red_input_t");
RD->startDefinition();
const FieldDecl *SharedFD = addFieldToRecordDecl(C, RD, C.VoidPtrTy);
const FieldDecl *SizeFD = addFieldToRecordDecl(C, RD, C.getSizeType());
@@ -5652,14 +6089,14 @@ void CGOpenMPRuntime::emitTaskReductionFixups(CodeGenFunction &CGF,
/*isSigned=*/false);
Address SizeAddr = getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().getSizeType(),
- generateUniqueName("reduction_size", Loc, N));
+ generateUniqueName(CGM, "reduction_size", RCG.getRefExpr(N)));
CGF.Builder.CreateStore(SizeVal, SizeAddr, /*IsVolatile=*/false);
}
// Store address of the original reduction item if custom initializer is used.
if (RCG.usesReductionInitializer(N)) {
Address SharedAddr = getAddrOfArtificialThreadPrivate(
CGF, CGM.getContext().VoidPtrTy,
- generateUniqueName("reduction", Loc, N));
+ generateUniqueName(CGM, "reduction", RCG.getRefExpr(N)));
CGF.Builder.CreateStore(
CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
RCG.getSharedLValue(N).getPointer(), CGM.VoidPtrTy),
@@ -5749,18 +6186,18 @@ void CGOpenMPRuntime::emitCancellationPointCall(
emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
CGF.Builder.getInt32(getCancellationKind(CancelRegion))};
// Ignore return result until untied tasks are supported.
- auto *Result = CGF.EmitRuntimeCall(
+ llvm::Value *Result = CGF.EmitRuntimeCall(
createRuntimeFunction(OMPRTL__kmpc_cancellationpoint), Args);
// if (__kmpc_cancellationpoint()) {
// exit from construct;
// }
- auto *ExitBB = CGF.createBasicBlock(".cancel.exit");
- auto *ContBB = CGF.createBasicBlock(".cancel.continue");
- auto *Cmp = CGF.Builder.CreateIsNotNull(Result);
+ llvm::BasicBlock *ExitBB = CGF.createBasicBlock(".cancel.exit");
+ llvm::BasicBlock *ContBB = CGF.createBasicBlock(".cancel.continue");
+ llvm::Value *Cmp = CGF.Builder.CreateIsNotNull(Result);
CGF.Builder.CreateCondBr(Cmp, ExitBB, ContBB);
CGF.EmitBlock(ExitBB);
// exit from construct;
- auto CancelDest =
+ CodeGenFunction::JumpDest CancelDest =
CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
CGF.EmitBranchThroughCleanup(CancelDest);
CGF.EmitBlock(ContBB, /*IsFinished=*/true);
@@ -5779,70 +6216,42 @@ void CGOpenMPRuntime::emitCancelCall(CodeGenFunction &CGF, SourceLocation Loc,
dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo)) {
auto &&ThenGen = [Loc, CancelRegion, OMPRegionInfo](CodeGenFunction &CGF,
PrePostActionTy &) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
+ CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
llvm::Value *Args[] = {
RT.emitUpdateLocation(CGF, Loc), RT.getThreadID(CGF, Loc),
CGF.Builder.getInt32(getCancellationKind(CancelRegion))};
// Ignore return result until untied tasks are supported.
- auto *Result = CGF.EmitRuntimeCall(
+ llvm::Value *Result = CGF.EmitRuntimeCall(
RT.createRuntimeFunction(OMPRTL__kmpc_cancel), Args);
// if (__kmpc_cancel()) {
// exit from construct;
// }
- auto *ExitBB = CGF.createBasicBlock(".cancel.exit");
- auto *ContBB = CGF.createBasicBlock(".cancel.continue");
- auto *Cmp = CGF.Builder.CreateIsNotNull(Result);
+ llvm::BasicBlock *ExitBB = CGF.createBasicBlock(".cancel.exit");
+ llvm::BasicBlock *ContBB = CGF.createBasicBlock(".cancel.continue");
+ llvm::Value *Cmp = CGF.Builder.CreateIsNotNull(Result);
CGF.Builder.CreateCondBr(Cmp, ExitBB, ContBB);
CGF.EmitBlock(ExitBB);
// exit from construct;
- auto CancelDest =
+ CodeGenFunction::JumpDest CancelDest =
CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
CGF.EmitBranchThroughCleanup(CancelDest);
CGF.EmitBlock(ContBB, /*IsFinished=*/true);
};
- if (IfCond)
+ if (IfCond) {
emitOMPIfClause(CGF, IfCond, ThenGen,
[](CodeGenFunction &, PrePostActionTy &) {});
- else {
+ } else {
RegionCodeGenTy ThenRCG(ThenGen);
ThenRCG(CGF);
}
}
}
-/// \brief Obtain information that uniquely identifies a target entry. This
-/// consists of the file and device IDs as well as line number associated with
-/// the relevant entry source location.
-static void getTargetEntryUniqueInfo(ASTContext &C, SourceLocation Loc,
- unsigned &DeviceID, unsigned &FileID,
- unsigned &LineNum) {
-
- auto &SM = C.getSourceManager();
-
- // The loc should be always valid and have a file ID (the user cannot use
- // #pragma directives in macros)
-
- assert(Loc.isValid() && "Source location is expected to be always valid.");
- assert(Loc.isFileID() && "Source location is expected to refer to a file.");
-
- PresumedLoc PLoc = SM.getPresumedLoc(Loc);
- assert(PLoc.isValid() && "Source location is expected to be always valid.");
-
- llvm::sys::fs::UniqueID ID;
- if (llvm::sys::fs::getUniqueID(PLoc.getFilename(), ID))
- llvm_unreachable("Source file with target region no longer exists!");
-
- DeviceID = ID.getDevice();
- FileID = ID.getFile();
- LineNum = PLoc.getLine();
-}
-
void CGOpenMPRuntime::emitTargetOutlinedFunction(
const OMPExecutableDirective &D, StringRef ParentName,
llvm::Function *&OutlinedFn, llvm::Constant *&OutlinedFnID,
bool IsOffloadEntry, const RegionCodeGenTy &CodeGen) {
assert(!ParentName.empty() && "Invalid target region parent name!");
-
emitTargetOutlinedFunctionHelper(D, ParentName, OutlinedFn, OutlinedFnID,
IsOffloadEntry, CodeGen);
}
@@ -5872,7 +6281,7 @@ void CGOpenMPRuntime::emitTargetOutlinedFunctionHelper(
<< llvm::format("_%x_", FileID) << ParentName << "_l" << Line;
}
- const CapturedStmt &CS = *cast<CapturedStmt>(D.getAssociatedStmt());
+ const CapturedStmt &CS = *D.getCapturedStmt(OMPD_target);
CodeGenFunction CGF(CGM, true);
CGOpenMPTargetRegionInfo CGInfo(CS, CodeGen, EntryFnName);
@@ -5898,22 +6307,25 @@ void CGOpenMPRuntime::emitTargetOutlinedFunctionHelper(
if (CGM.getLangOpts().OpenMPIsDevice) {
OutlinedFnID = llvm::ConstantExpr::getBitCast(OutlinedFn, CGM.Int8PtrTy);
- OutlinedFn->setLinkage(llvm::GlobalValue::ExternalLinkage);
- } else
+ OutlinedFn->setLinkage(llvm::GlobalValue::WeakAnyLinkage);
+ OutlinedFn->setDSOLocal(false);
+ } else {
+ std::string Name = getName({EntryFnName, "region_id"});
OutlinedFnID = new llvm::GlobalVariable(
CGM.getModule(), CGM.Int8Ty, /*isConstant=*/true,
- llvm::GlobalValue::PrivateLinkage,
- llvm::Constant::getNullValue(CGM.Int8Ty), ".omp_offload.region_id");
+ llvm::GlobalValue::WeakAnyLinkage,
+ llvm::Constant::getNullValue(CGM.Int8Ty), Name);
+ }
// Register the information for the entry associated with this target region.
OffloadEntriesInfoManager.registerTargetRegionEntryInfo(
DeviceID, FileID, ParentName, Line, OutlinedFn, OutlinedFnID,
- /*Flags=*/0);
+ OffloadEntriesInfoManagerTy::OMPTargetRegionEntryTargetRegion);
}
/// discard all CompoundStmts intervening between two constructs
static const Stmt *ignoreCompoundStmts(const Stmt *Body) {
- while (auto *CS = dyn_cast_or_null<CompoundStmt>(Body))
+ while (const auto *CS = dyn_cast_or_null<CompoundStmt>(Body))
Body = CS->body_front();
return Body;
@@ -5931,12 +6343,11 @@ static llvm::Value *
emitNumTeamsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
CodeGenFunction &CGF,
const OMPExecutableDirective &D) {
-
assert(!CGF.getLangOpts().OpenMPIsDevice && "Clauses associated with the "
"teams directive expected to be "
"emitted only for the host!");
- auto &Bld = CGF.Builder;
+ CGBuilderTy &Bld = CGF.Builder;
// If the target directive is combined with a teams directive:
// Return the value in the num_teams clause, if any.
@@ -5944,8 +6355,8 @@ emitNumTeamsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
if (isOpenMPTeamsDirective(D.getDirectiveKind())) {
if (const auto *NumTeamsClause = D.getSingleClause<OMPNumTeamsClause>()) {
CodeGenFunction::RunCleanupsScope NumTeamsScope(CGF);
- auto NumTeams = CGF.EmitScalarExpr(NumTeamsClause->getNumTeams(),
- /*IgnoreResultAssign*/ true);
+ llvm::Value *NumTeams = CGF.EmitScalarExpr(NumTeamsClause->getNumTeams(),
+ /*IgnoreResultAssign*/ true);
return Bld.CreateIntCast(NumTeams, CGF.Int32Ty,
/*IsSigned=*/true);
}
@@ -5965,12 +6376,12 @@ emitNumTeamsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
// the expression is captured in the enclosing target environment when the
// teams directive is not combined with target.
- const CapturedStmt &CS = *cast<CapturedStmt>(D.getAssociatedStmt());
+ const CapturedStmt &CS = *D.getCapturedStmt(OMPD_target);
- if (auto *TeamsDir = dyn_cast_or_null<OMPExecutableDirective>(
+ if (const auto *TeamsDir = dyn_cast_or_null<OMPExecutableDirective>(
ignoreCompoundStmts(CS.getCapturedStmt()))) {
if (isOpenMPTeamsDirective(TeamsDir->getDirectiveKind())) {
- if (auto *NTE = TeamsDir->getSingleClause<OMPNumTeamsClause>()) {
+ if (const auto *NTE = TeamsDir->getSingleClause<OMPNumTeamsClause>()) {
CGOpenMPInnerExprInfo CGInfo(CGF, CS);
CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
llvm::Value *NumTeams = CGF.EmitScalarExpr(NTE->getNumTeams());
@@ -6000,12 +6411,11 @@ static llvm::Value *
emitNumThreadsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
CodeGenFunction &CGF,
const OMPExecutableDirective &D) {
-
assert(!CGF.getLangOpts().OpenMPIsDevice && "Clauses associated with the "
"teams directive expected to be "
"emitted only for the host!");
- auto &Bld = CGF.Builder;
+ CGBuilderTy &Bld = CGF.Builder;
//
// If the target directive is combined with a teams directive:
@@ -6030,8 +6440,9 @@ emitNumThreadsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
if (const auto *ThreadLimitClause =
D.getSingleClause<OMPThreadLimitClause>()) {
CodeGenFunction::RunCleanupsScope ThreadLimitScope(CGF);
- auto ThreadLimit = CGF.EmitScalarExpr(ThreadLimitClause->getThreadLimit(),
- /*IgnoreResultAssign*/ true);
+ llvm::Value *ThreadLimit =
+ CGF.EmitScalarExpr(ThreadLimitClause->getThreadLimit(),
+ /*IgnoreResultAssign*/ true);
ThreadLimitVal = Bld.CreateIntCast(ThreadLimit, CGF.Int32Ty,
/*IsSigned=*/true);
}
@@ -6068,12 +6479,12 @@ emitNumThreadsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
// the expression is captured in the enclosing target environment when the
// teams directive is not combined with target.
- const CapturedStmt &CS = *cast<CapturedStmt>(D.getAssociatedStmt());
+ const CapturedStmt &CS = *D.getCapturedStmt(OMPD_target);
- if (auto *TeamsDir = dyn_cast_or_null<OMPExecutableDirective>(
+ if (const auto *TeamsDir = dyn_cast_or_null<OMPExecutableDirective>(
ignoreCompoundStmts(CS.getCapturedStmt()))) {
if (isOpenMPTeamsDirective(TeamsDir->getDirectiveKind())) {
- if (auto *TLE = TeamsDir->getSingleClause<OMPThreadLimitClause>()) {
+ if (const auto *TLE = TeamsDir->getSingleClause<OMPThreadLimitClause>()) {
CGOpenMPInnerExprInfo CGInfo(CGF, CS);
CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
llvm::Value *ThreadLimit = CGF.EmitScalarExpr(TLE->getThreadLimit());
@@ -6092,42 +6503,50 @@ emitNumThreadsForTargetDirective(CGOpenMPRuntime &OMPRuntime,
}
namespace {
-// \brief Utility to handle information from clauses associated with a given
+LLVM_ENABLE_BITMASK_ENUMS_IN_NAMESPACE();
+
+// Utility to handle information from clauses associated with a given
// construct that use mappable expressions (e.g. 'map' clause, 'to' clause).
// It provides a convenient interface to obtain the information and generate
// code for that information.
class MappableExprsHandler {
public:
- /// \brief Values for bit flags used to specify the mapping type for
+ /// Values for bit flags used to specify the mapping type for
/// offloading.
- enum OpenMPOffloadMappingFlags {
- /// \brief Allocate memory on the device and move data from host to device.
+ enum OpenMPOffloadMappingFlags : uint64_t {
+ /// No flags
+ OMP_MAP_NONE = 0x0,
+ /// Allocate memory on the device and move data from host to device.
OMP_MAP_TO = 0x01,
- /// \brief Allocate memory on the device and move data from device to host.
+ /// Allocate memory on the device and move data from device to host.
OMP_MAP_FROM = 0x02,
- /// \brief Always perform the requested mapping action on the element, even
+ /// Always perform the requested mapping action on the element, even
/// if it was already mapped before.
OMP_MAP_ALWAYS = 0x04,
- /// \brief Delete the element from the device environment, ignoring the
+ /// Delete the element from the device environment, ignoring the
/// current reference count associated with the element.
OMP_MAP_DELETE = 0x08,
- /// \brief The element being mapped is a pointer-pointee pair; both the
+ /// The element being mapped is a pointer-pointee pair; both the
/// pointer and the pointee should be mapped.
OMP_MAP_PTR_AND_OBJ = 0x10,
- /// \brief This flags signals that the base address of an entry should be
+ /// This flags signals that the base address of an entry should be
/// passed to the target kernel as an argument.
OMP_MAP_TARGET_PARAM = 0x20,
- /// \brief Signal that the runtime library has to return the device pointer
+ /// Signal that the runtime library has to return the device pointer
/// in the current position for the data being mapped. Used when we have the
/// use_device_ptr clause.
OMP_MAP_RETURN_PARAM = 0x40,
- /// \brief This flag signals that the reference being passed is a pointer to
+ /// This flag signals that the reference being passed is a pointer to
/// private data.
OMP_MAP_PRIVATE = 0x80,
- /// \brief Pass the element to the device by value.
+ /// Pass the element to the device by value.
OMP_MAP_LITERAL = 0x100,
/// Implicit map
OMP_MAP_IMPLICIT = 0x200,
+ /// The 16 MSBs of the flags indicate whether the entry is member of some
+ /// struct/class.
+ OMP_MAP_MEMBER_OF = 0xffff000000000000,
+ LLVM_MARK_AS_BITMASK_ENUM(/* LargestFlag = */ OMP_MAP_MEMBER_OF),
};
/// Class that associates information with a base pointer to be passed to the
@@ -6147,21 +6566,60 @@ public:
void setDevicePtrDecl(const ValueDecl *D) { DevPtrDecl = D; }
};
- typedef SmallVector<BasePointerInfo, 16> MapBaseValuesArrayTy;
- typedef SmallVector<llvm::Value *, 16> MapValuesArrayTy;
- typedef SmallVector<uint64_t, 16> MapFlagsArrayTy;
+ using MapBaseValuesArrayTy = SmallVector<BasePointerInfo, 4>;
+ using MapValuesArrayTy = SmallVector<llvm::Value *, 4>;
+ using MapFlagsArrayTy = SmallVector<OpenMPOffloadMappingFlags, 4>;
+
+ /// Map between a struct and the its lowest & highest elements which have been
+ /// mapped.
+ /// [ValueDecl *] --> {LE(FieldIndex, Pointer),
+ /// HE(FieldIndex, Pointer)}
+ struct StructRangeInfoTy {
+ std::pair<unsigned /*FieldIndex*/, Address /*Pointer*/> LowestElem = {
+ 0, Address::invalid()};
+ std::pair<unsigned /*FieldIndex*/, Address /*Pointer*/> HighestElem = {
+ 0, Address::invalid()};
+ Address Base = Address::invalid();
+ };
private:
- /// \brief Directive from where the map clauses were extracted.
+ /// Kind that defines how a device pointer has to be returned.
+ struct MapInfo {
+ OMPClauseMappableExprCommon::MappableExprComponentListRef Components;
+ OpenMPMapClauseKind MapType = OMPC_MAP_unknown;
+ OpenMPMapClauseKind MapTypeModifier = OMPC_MAP_unknown;
+ bool ReturnDevicePointer = false;
+ bool IsImplicit = false;
+
+ MapInfo() = default;
+ MapInfo(
+ OMPClauseMappableExprCommon::MappableExprComponentListRef Components,
+ OpenMPMapClauseKind MapType, OpenMPMapClauseKind MapTypeModifier,
+ bool ReturnDevicePointer, bool IsImplicit)
+ : Components(Components), MapType(MapType),
+ MapTypeModifier(MapTypeModifier),
+ ReturnDevicePointer(ReturnDevicePointer), IsImplicit(IsImplicit) {}
+ };
+
+ /// If use_device_ptr is used on a pointer which is a struct member and there
+ /// is no map information about it, then emission of that entry is deferred
+ /// until the whole struct has been processed.
+ struct DeferredDevicePtrEntryTy {
+ const Expr *IE = nullptr;
+ const ValueDecl *VD = nullptr;
+
+ DeferredDevicePtrEntryTy(const Expr *IE, const ValueDecl *VD)
+ : IE(IE), VD(VD) {}
+ };
+
+ /// Directive from where the map clauses were extracted.
const OMPExecutableDirective &CurDir;
- /// \brief Function the directive is being generated for.
+ /// Function the directive is being generated for.
CodeGenFunction &CGF;
- /// \brief Set of all first private variables in the current directive.
+ /// Set of all first private variables in the current directive.
llvm::SmallPtrSet<const VarDecl *, 8> FirstPrivateDecls;
- /// Set of all reduction variables in the current directive.
- llvm::SmallPtrSet<const VarDecl *, 8> ReductionDecls;
/// Map between device pointer declarations and their expression components.
/// The key value for declarations in 'this' is null.
@@ -6171,10 +6629,10 @@ private:
DevPointersMap;
llvm::Value *getExprTypeSize(const Expr *E) const {
- auto ExprTy = E->getType().getCanonicalType();
+ QualType ExprTy = E->getType().getCanonicalType();
// Reference types are ignored for mapping purposes.
- if (auto *RefTy = ExprTy->getAs<ReferenceType>())
+ if (const auto *RefTy = ExprTy->getAs<ReferenceType>())
ExprTy = RefTy->getPointeeType().getCanonicalType();
// Given that an array section is considered a built-in type, we need to
@@ -6191,10 +6649,10 @@ private:
return CGF.getTypeSize(BaseTy);
llvm::Value *ElemSize;
- if (auto *PTy = BaseTy->getAs<PointerType>())
+ if (const auto *PTy = BaseTy->getAs<PointerType>()) {
ElemSize = CGF.getTypeSize(PTy->getPointeeType().getCanonicalType());
- else {
- auto *ATy = cast<ArrayType>(BaseTy.getTypePtr());
+ } else {
+ const auto *ATy = cast<ArrayType>(BaseTy.getTypePtr());
assert(ATy && "Expecting array type if not a pointer type.");
ElemSize = CGF.getTypeSize(ATy->getElementType().getCanonicalType());
}
@@ -6204,7 +6662,7 @@ private:
if (!OAE->getLength())
return ElemSize;
- auto *LengthVal = CGF.EmitScalarExpr(OAE->getLength());
+ llvm::Value *LengthVal = CGF.EmitScalarExpr(OAE->getLength());
LengthVal =
CGF.Builder.CreateIntCast(LengthVal, CGF.SizeTy, /*isSigned=*/false);
return CGF.Builder.CreateNUWMul(LengthVal, ElemSize);
@@ -6212,14 +6670,16 @@ private:
return CGF.getTypeSize(ExprTy);
}
- /// \brief Return the corresponding bits for a given map clause modifier. Add
+ /// Return the corresponding bits for a given map clause modifier. Add
/// a flag marking the map as a pointer if requested. Add a flag marking the
/// map as the first one of a series of maps that relate to the same map
/// expression.
- uint64_t getMapTypeBits(OpenMPMapClauseKind MapType,
- OpenMPMapClauseKind MapTypeModifier, bool AddPtrFlag,
- bool AddIsTargetParamFlag) const {
- uint64_t Bits = 0u;
+ OpenMPOffloadMappingFlags getMapTypeBits(OpenMPMapClauseKind MapType,
+ OpenMPMapClauseKind MapTypeModifier,
+ bool IsImplicit, bool AddPtrFlag,
+ bool AddIsTargetParamFlag) const {
+ OpenMPOffloadMappingFlags Bits =
+ IsImplicit ? OMP_MAP_IMPLICIT : OMP_MAP_NONE;
switch (MapType) {
case OMPC_MAP_alloc:
case OMPC_MAP_release:
@@ -6229,20 +6689,20 @@ private:
// type modifiers.
break;
case OMPC_MAP_to:
- Bits = OMP_MAP_TO;
+ Bits |= OMP_MAP_TO;
break;
case OMPC_MAP_from:
- Bits = OMP_MAP_FROM;
+ Bits |= OMP_MAP_FROM;
break;
case OMPC_MAP_tofrom:
- Bits = OMP_MAP_TO | OMP_MAP_FROM;
+ Bits |= OMP_MAP_TO | OMP_MAP_FROM;
break;
case OMPC_MAP_delete:
- Bits = OMP_MAP_DELETE;
+ Bits |= OMP_MAP_DELETE;
break;
- default:
+ case OMPC_MAP_always:
+ case OMPC_MAP_unknown:
llvm_unreachable("Unexpected map type!");
- break;
}
if (AddPtrFlag)
Bits |= OMP_MAP_PTR_AND_OBJ;
@@ -6253,10 +6713,10 @@ private:
return Bits;
}
- /// \brief Return true if the provided expression is a final array section. A
+ /// Return true if the provided expression is a final array section. A
/// final array section, is one whose length can't be proved to be one.
bool isFinalArraySectionExpression(const Expr *E) const {
- auto *OASE = dyn_cast<OMPArraySectionExpr>(E);
+ const auto *OASE = dyn_cast<OMPArraySectionExpr>(E);
// It is not an array section and therefore not a unity-size one.
if (!OASE)
@@ -6266,16 +6726,16 @@ private:
if (OASE->getColonLoc().isInvalid())
return false;
- auto *Length = OASE->getLength();
+ const Expr *Length = OASE->getLength();
// If we don't have a length we have to check if the array has size 1
// for this dimension. Also, we should always expect a length if the
// base type is pointer.
if (!Length) {
- auto BaseQTy = OMPArraySectionExpr::getBaseOriginalType(
- OASE->getBase()->IgnoreParenImpCasts())
- .getCanonicalType();
- if (auto *ATy = dyn_cast<ConstantArrayType>(BaseQTy.getTypePtr()))
+ QualType BaseQTy = OMPArraySectionExpr::getBaseOriginalType(
+ OASE->getBase()->IgnoreParenImpCasts())
+ .getCanonicalType();
+ if (const auto *ATy = dyn_cast<ConstantArrayType>(BaseQTy.getTypePtr()))
return ATy->getSize().getSExtValue() != 1;
// If we don't have a constant dimension length, we have to consider
// the current section as having any size, so it is not necessarily
@@ -6291,7 +6751,7 @@ private:
return ConstLength.getSExtValue() != 1;
}
- /// \brief Generate the base pointers, section pointers, sizes and map type
+ /// Generate the base pointers, section pointers, sizes and map type
/// bits for the provided map type, map modifier, and expression components.
/// \a IsFirstComponent should be set to true if the provided set of
/// components is the first associated with a capture.
@@ -6300,10 +6760,10 @@ private:
OMPClauseMappableExprCommon::MappableExprComponentListRef Components,
MapBaseValuesArrayTy &BasePointers, MapValuesArrayTy &Pointers,
MapValuesArrayTy &Sizes, MapFlagsArrayTy &Types,
- bool IsFirstComponentList, bool IsImplicit) const {
-
+ StructRangeInfoTy &PartialStruct, bool IsFirstComponentList,
+ bool IsImplicit) const {
// The following summarizes what has to be generated for each map and the
- // types bellow. The generated information is expressed in this order:
+ // types below. The generated information is expressed in this order:
// base pointer, section pointer, size, flags
// (to add to the ones that come from the map type and modifier).
//
@@ -6326,96 +6786,141 @@ private:
// S2 *ps;
//
// map(d)
- // &d, &d, sizeof(double), noflags
+ // &d, &d, sizeof(double), TARGET_PARAM | TO | FROM
//
// map(i)
- // &i, &i, 100*sizeof(int), noflags
+ // &i, &i, 100*sizeof(int), TARGET_PARAM | TO | FROM
//
// map(i[1:23])
- // &i(=&i[0]), &i[1], 23*sizeof(int), noflags
+ // &i(=&i[0]), &i[1], 23*sizeof(int), TARGET_PARAM | TO | FROM
//
// map(p)
- // &p, &p, sizeof(float*), noflags
+ // &p, &p, sizeof(float*), TARGET_PARAM | TO | FROM
//
// map(p[1:24])
- // p, &p[1], 24*sizeof(float), noflags
+ // p, &p[1], 24*sizeof(float), TARGET_PARAM | TO | FROM
//
// map(s)
- // &s, &s, sizeof(S2), noflags
+ // &s, &s, sizeof(S2), TARGET_PARAM | TO | FROM
//
// map(s.i)
- // &s, &(s.i), sizeof(int), noflags
+ // &s, &(s.i), sizeof(int), TARGET_PARAM | TO | FROM
//
// map(s.s.f)
- // &s, &(s.i.f), 50*sizeof(int), noflags
+ // &s, &(s.s.f[0]), 50*sizeof(float), TARGET_PARAM | TO | FROM
//
// map(s.p)
- // &s, &(s.p), sizeof(double*), noflags
+ // &s, &(s.p), sizeof(double*), TARGET_PARAM | TO | FROM
//
- // map(s.p[:22], s.a s.b)
- // &s, &(s.p), sizeof(double*), noflags
- // &(s.p), &(s.p[0]), 22*sizeof(double), ptr_flag
+ // map(to: s.p[:22])
+ // &s, &(s.p), sizeof(double*), TARGET_PARAM (*)
+ // &s, &(s.p), sizeof(double*), MEMBER_OF(1) (**)
+ // &(s.p), &(s.p[0]), 22*sizeof(double),
+ // MEMBER_OF(1) | PTR_AND_OBJ | TO (***)
+ // (*) alloc space for struct members, only this is a target parameter
+ // (**) map the pointer (nothing to be mapped in this example) (the compiler
+ // optimizes this entry out, same in the examples below)
+ // (***) map the pointee (map: to)
//
// map(s.ps)
- // &s, &(s.ps), sizeof(S2*), noflags
+ // &s, &(s.ps), sizeof(S2*), TARGET_PARAM | TO | FROM
//
- // map(s.ps->s.i)
- // &s, &(s.ps), sizeof(S2*), noflags
- // &(s.ps), &(s.ps->s.i), sizeof(int), ptr_flag
+ // map(from: s.ps->s.i)
+ // &s, &(s.ps), sizeof(S2*), TARGET_PARAM
+ // &s, &(s.ps), sizeof(S2*), MEMBER_OF(1)
+ // &(s.ps), &(s.ps->s.i), sizeof(int), MEMBER_OF(1) | PTR_AND_OBJ | FROM
//
- // map(s.ps->ps)
- // &s, &(s.ps), sizeof(S2*), noflags
- // &(s.ps), &(s.ps->ps), sizeof(S2*), ptr_flag
+ // map(to: s.ps->ps)
+ // &s, &(s.ps), sizeof(S2*), TARGET_PARAM
+ // &s, &(s.ps), sizeof(S2*), MEMBER_OF(1)
+ // &(s.ps), &(s.ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ | TO
//
// map(s.ps->ps->ps)
- // &s, &(s.ps), sizeof(S2*), noflags
- // &(s.ps), &(s.ps->ps), sizeof(S2*), ptr_flag
- // &(s.ps->ps), &(s.ps->ps->ps), sizeof(S2*), ptr_flag
+ // &s, &(s.ps), sizeof(S2*), TARGET_PARAM
+ // &s, &(s.ps), sizeof(S2*), MEMBER_OF(1)
+ // &(s.ps), &(s.ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ
+ // &(s.ps->ps), &(s.ps->ps->ps), sizeof(S2*), PTR_AND_OBJ | TO | FROM
//
- // map(s.ps->ps->s.f[:22])
- // &s, &(s.ps), sizeof(S2*), noflags
- // &(s.ps), &(s.ps->ps), sizeof(S2*), ptr_flag
- // &(s.ps->ps), &(s.ps->ps->s.f[0]), 22*sizeof(float), ptr_flag
+ // map(to: s.ps->ps->s.f[:22])
+ // &s, &(s.ps), sizeof(S2*), TARGET_PARAM
+ // &s, &(s.ps), sizeof(S2*), MEMBER_OF(1)
+ // &(s.ps), &(s.ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ
+ // &(s.ps->ps), &(s.ps->ps->s.f[0]), 22*sizeof(float), PTR_AND_OBJ | TO
//
// map(ps)
- // &ps, &ps, sizeof(S2*), noflags
+ // &ps, &ps, sizeof(S2*), TARGET_PARAM | TO | FROM
//
// map(ps->i)
- // ps, &(ps->i), sizeof(int), noflags
+ // ps, &(ps->i), sizeof(int), TARGET_PARAM | TO | FROM
//
// map(ps->s.f)
- // ps, &(ps->s.f[0]), 50*sizeof(float), noflags
+ // ps, &(ps->s.f[0]), 50*sizeof(float), TARGET_PARAM | TO | FROM
//
- // map(ps->p)
- // ps, &(ps->p), sizeof(double*), noflags
+ // map(from: ps->p)
+ // ps, &(ps->p), sizeof(double*), TARGET_PARAM | FROM
//
- // map(ps->p[:22])
- // ps, &(ps->p), sizeof(double*), noflags
- // &(ps->p), &(ps->p[0]), 22*sizeof(double), ptr_flag
+ // map(to: ps->p[:22])
+ // ps, &(ps->p), sizeof(double*), TARGET_PARAM
+ // ps, &(ps->p), sizeof(double*), MEMBER_OF(1)
+ // &(ps->p), &(ps->p[0]), 22*sizeof(double), MEMBER_OF(1) | PTR_AND_OBJ | TO
//
// map(ps->ps)
- // ps, &(ps->ps), sizeof(S2*), noflags
+ // ps, &(ps->ps), sizeof(S2*), TARGET_PARAM | TO | FROM
//
- // map(ps->ps->s.i)
- // ps, &(ps->ps), sizeof(S2*), noflags
- // &(ps->ps), &(ps->ps->s.i), sizeof(int), ptr_flag
+ // map(from: ps->ps->s.i)
+ // ps, &(ps->ps), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->ps), sizeof(S2*), MEMBER_OF(1)
+ // &(ps->ps), &(ps->ps->s.i), sizeof(int), MEMBER_OF(1) | PTR_AND_OBJ | FROM
//
- // map(ps->ps->ps)
- // ps, &(ps->ps), sizeof(S2*), noflags
- // &(ps->ps), &(ps->ps->ps), sizeof(S2*), ptr_flag
+ // map(from: ps->ps->ps)
+ // ps, &(ps->ps), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->ps), sizeof(S2*), MEMBER_OF(1)
+ // &(ps->ps), &(ps->ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ | FROM
//
// map(ps->ps->ps->ps)
- // ps, &(ps->ps), sizeof(S2*), noflags
- // &(ps->ps), &(ps->ps->ps), sizeof(S2*), ptr_flag
- // &(ps->ps->ps), &(ps->ps->ps->ps), sizeof(S2*), ptr_flag
+ // ps, &(ps->ps), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->ps), sizeof(S2*), MEMBER_OF(1)
+ // &(ps->ps), &(ps->ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ
+ // &(ps->ps->ps), &(ps->ps->ps->ps), sizeof(S2*), PTR_AND_OBJ | TO | FROM
//
- // map(ps->ps->ps->s.f[:22])
- // ps, &(ps->ps), sizeof(S2*), noflags
- // &(ps->ps), &(ps->ps->ps), sizeof(S2*), ptr_flag
- // &(ps->ps->ps), &(ps->ps->ps->s.f[0]), 22*sizeof(float), ptr_flag
+ // map(to: ps->ps->ps->s.f[:22])
+ // ps, &(ps->ps), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->ps), sizeof(S2*), MEMBER_OF(1)
+ // &(ps->ps), &(ps->ps->ps), sizeof(S2*), MEMBER_OF(1) | PTR_AND_OBJ
+ // &(ps->ps->ps), &(ps->ps->ps->s.f[0]), 22*sizeof(float), PTR_AND_OBJ | TO
+ //
+ // map(to: s.f[:22]) map(from: s.p[:33])
+ // &s, &(s.f[0]), 50*sizeof(float) + sizeof(struct S1) +
+ // sizeof(double*) (**), TARGET_PARAM
+ // &s, &(s.f[0]), 22*sizeof(float), MEMBER_OF(1) | TO
+ // &s, &(s.p), sizeof(double*), MEMBER_OF(1)
+ // &(s.p), &(s.p[0]), 33*sizeof(double), MEMBER_OF(1) | PTR_AND_OBJ | FROM
+ // (*) allocate contiguous space needed to fit all mapped members even if
+ // we allocate space for members not mapped (in this example,
+ // s.f[22..49] and s.s are not mapped, yet we must allocate space for
+ // them as well because they fall between &s.f[0] and &s.p)
+ //
+ // map(from: s.f[:22]) map(to: ps->p[:33])
+ // &s, &(s.f[0]), 22*sizeof(float), TARGET_PARAM | FROM
+ // ps, &(ps->p), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->p), sizeof(double*), MEMBER_OF(2) (*)
+ // &(ps->p), &(ps->p[0]), 33*sizeof(double), MEMBER_OF(2) | PTR_AND_OBJ | TO
+ // (*) the struct this entry pertains to is the 2nd element in the list of
+ // arguments, hence MEMBER_OF(2)
+ //
+ // map(from: s.f[:22], s.s) map(to: ps->p[:33])
+ // &s, &(s.f[0]), 50*sizeof(float) + sizeof(struct S1), TARGET_PARAM
+ // &s, &(s.f[0]), 22*sizeof(float), MEMBER_OF(1) | FROM
+ // &s, &(s.s), sizeof(struct S1), MEMBER_OF(1) | FROM
+ // ps, &(ps->p), sizeof(S2*), TARGET_PARAM
+ // ps, &(ps->p), sizeof(double*), MEMBER_OF(4) (*)
+ // &(ps->p), &(ps->p[0]), 33*sizeof(double), MEMBER_OF(4) | PTR_AND_OBJ | TO
+ // (*) the struct this entry pertains to is the 4th element in the list
+ // of arguments, hence MEMBER_OF(4)
// Track if the map information being generated is the first for a capture.
bool IsCaptureFirstInfo = IsFirstComponentList;
+ bool IsLink = false; // Is this variable a "declare target link"?
// Scan the components from the base to the complete expression.
auto CI = Components.rbegin();
@@ -6425,16 +6930,25 @@ private:
// Track if the map information being generated is the first for a list of
// components.
bool IsExpressionFirstInfo = true;
- llvm::Value *BP = nullptr;
+ Address BP = Address::invalid();
- if (auto *ME = dyn_cast<MemberExpr>(I->getAssociatedExpression())) {
+ if (isa<MemberExpr>(I->getAssociatedExpression())) {
// The base is the 'this' pointer. The content of the pointer is going
// to be the base of the field being mapped.
- BP = CGF.EmitScalarExpr(ME->getBase());
+ BP = CGF.LoadCXXThisAddress();
} else {
// The base is the reference to the variable.
// BP = &Var.
- BP = CGF.EmitOMPSharedLValue(I->getAssociatedExpression()).getPointer();
+ BP = CGF.EmitOMPSharedLValue(I->getAssociatedExpression()).getAddress();
+ if (const auto *VD =
+ dyn_cast_or_null<VarDecl>(I->getAssociatedDeclaration())) {
+ if (llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(VD))
+ if (*Res == OMPDeclareTargetDeclAttr::MT_Link) {
+ IsLink = true;
+ BP = CGF.CGM.getOpenMPRuntime().getAddrOfDeclareTargetLink(VD);
+ }
+ }
// If the variable is a pointer and is being dereferenced (i.e. is not
// the last component), the base has to be the pointer itself, not its
@@ -6442,10 +6956,7 @@ private:
QualType Ty =
I->getAssociatedDeclaration()->getType().getNonReferenceType();
if (Ty->isAnyPointerType() && std::next(I) != CE) {
- auto PtrAddr = CGF.MakeNaturalAlignAddrLValue(BP, Ty);
- BP = CGF.EmitLoadOfPointerLValue(PtrAddr.getAddress(),
- Ty->castAs<PointerType>())
- .getPointer();
+ BP = CGF.EmitLoadOfPointer(BP, Ty->castAs<PointerType>());
// We do not need to generate individual map information for the
// pointer, it can be associated with the combined storage.
@@ -6453,8 +6964,41 @@ private:
}
}
- uint64_t DefaultFlags = IsImplicit ? OMP_MAP_IMPLICIT : 0;
+ // Track whether a component of the list should be marked as MEMBER_OF some
+ // combined entry (for partial structs). Only the first PTR_AND_OBJ entry
+ // in a component list should be marked as MEMBER_OF, all subsequent entries
+ // do not belong to the base struct. E.g.
+ // struct S2 s;
+ // s.ps->ps->ps->f[:]
+ // (1) (2) (3) (4)
+ // ps(1) is a member pointer, ps(2) is a pointee of ps(1), so it is a
+ // PTR_AND_OBJ entry; the PTR is ps(1), so MEMBER_OF the base struct. ps(3)
+ // is the pointee of ps(2) which is not member of struct s, so it should not
+ // be marked as such (it is still PTR_AND_OBJ).
+ // The variable is initialized to false so that PTR_AND_OBJ entries which
+ // are not struct members are not considered (e.g. array of pointers to
+ // data).
+ bool ShouldBeMemberOf = false;
+
+ // Variable keeping track of whether or not we have encountered a component
+ // in the component list which is a member expression. Useful when we have a
+ // pointer or a final array section, in which case it is the previous
+ // component in the list which tells us whether we have a member expression.
+ // E.g. X.f[:]
+ // While processing the final array section "[:]" it is "f" which tells us
+ // whether we are dealing with a member of a declared struct.
+ const MemberExpr *EncounteredME = nullptr;
+
for (; I != CE; ++I) {
+ // If the current component is member of a struct (parent struct) mark it.
+ if (!EncounteredME) {
+ EncounteredME = dyn_cast<MemberExpr>(I->getAssociatedExpression());
+ // If we encounter a PTR_AND_OBJ entry from now on it should be marked
+ // as MEMBER_OF the parent struct.
+ if (EncounteredME)
+ ShouldBeMemberOf = true;
+ }
+
auto Next = std::next(I);
// We need to generate the addresses and sizes if this is the last
@@ -6472,14 +7016,12 @@ private:
const auto *OASE =
dyn_cast<OMPArraySectionExpr>(I->getAssociatedExpression());
bool IsPointer =
- (OASE &&
- OMPArraySectionExpr::getBaseOriginalType(OASE)
- .getCanonicalType()
- ->isAnyPointerType()) ||
+ (OASE && OMPArraySectionExpr::getBaseOriginalType(OASE)
+ .getCanonicalType()
+ ->isAnyPointerType()) ||
I->getAssociatedExpression()->getType()->isAnyPointerType();
if (Next == CE || IsPointer || IsFinalArraySection) {
-
// If this is not the last component, we expect the pointer to be
// associated with an array expression or member expression.
assert((Next == CE ||
@@ -6488,44 +7030,68 @@ private:
isa<OMPArraySectionExpr>(Next->getAssociatedExpression())) &&
"Unexpected expression");
- llvm::Value *LB =
- CGF.EmitOMPSharedLValue(I->getAssociatedExpression()).getPointer();
- auto *Size = getExprTypeSize(I->getAssociatedExpression());
-
- // If we have a member expression and the current component is a
- // reference, we have to map the reference too. Whenever we have a
- // reference, the section that reference refers to is going to be a
- // load instruction from the storage assigned to the reference.
- if (isa<MemberExpr>(I->getAssociatedExpression()) &&
- I->getAssociatedDeclaration()->getType()->isReferenceType()) {
- auto *LI = cast<llvm::LoadInst>(LB);
- auto *RefAddr = LI->getPointerOperand();
-
- BasePointers.push_back(BP);
- Pointers.push_back(RefAddr);
- Sizes.push_back(CGF.getTypeSize(CGF.getContext().VoidPtrTy));
- Types.push_back(DefaultFlags |
- getMapTypeBits(
- /*MapType*/ OMPC_MAP_alloc,
- /*MapTypeModifier=*/OMPC_MAP_unknown,
- !IsExpressionFirstInfo, IsCaptureFirstInfo));
- IsExpressionFirstInfo = false;
- IsCaptureFirstInfo = false;
- // The reference will be the next base address.
- BP = RefAddr;
- }
+ Address LB =
+ CGF.EmitOMPSharedLValue(I->getAssociatedExpression()).getAddress();
+ llvm::Value *Size = getExprTypeSize(I->getAssociatedExpression());
+
+ // If this component is a pointer inside the base struct then we don't
+ // need to create any entry for it - it will be combined with the object
+ // it is pointing to into a single PTR_AND_OBJ entry.
+ bool IsMemberPointer =
+ IsPointer && EncounteredME &&
+ (dyn_cast<MemberExpr>(I->getAssociatedExpression()) ==
+ EncounteredME);
+ if (!IsMemberPointer) {
+ BasePointers.push_back(BP.getPointer());
+ Pointers.push_back(LB.getPointer());
+ Sizes.push_back(Size);
+
+ // We need to add a pointer flag for each map that comes from the
+ // same expression except for the first one. We also need to signal
+ // this map is the first one that relates with the current capture
+ // (there is a set of entries for each capture).
+ OpenMPOffloadMappingFlags Flags = getMapTypeBits(
+ MapType, MapTypeModifier, IsImplicit,
+ !IsExpressionFirstInfo || IsLink, IsCaptureFirstInfo && !IsLink);
+
+ if (!IsExpressionFirstInfo) {
+ // If we have a PTR_AND_OBJ pair where the OBJ is a pointer as well,
+ // then we reset the TO/FROM/ALWAYS/DELETE flags.
+ if (IsPointer)
+ Flags &= ~(OMP_MAP_TO | OMP_MAP_FROM | OMP_MAP_ALWAYS |
+ OMP_MAP_DELETE);
+
+ if (ShouldBeMemberOf) {
+ // Set placeholder value MEMBER_OF=FFFF to indicate that the flag
+ // should be later updated with the correct value of MEMBER_OF.
+ Flags |= OMP_MAP_MEMBER_OF;
+ // From now on, all subsequent PTR_AND_OBJ entries should not be
+ // marked as MEMBER_OF.
+ ShouldBeMemberOf = false;
+ }
+ }
- BasePointers.push_back(BP);
- Pointers.push_back(LB);
- Sizes.push_back(Size);
+ Types.push_back(Flags);
+ }
- // We need to add a pointer flag for each map that comes from the
- // same expression except for the first one. We also need to signal
- // this map is the first one that relates with the current capture
- // (there is a set of entries for each capture).
- Types.push_back(DefaultFlags | getMapTypeBits(MapType, MapTypeModifier,
- !IsExpressionFirstInfo,
- IsCaptureFirstInfo));
+ // If we have encountered a member expression so far, keep track of the
+ // mapped member. If the parent is "*this", then the value declaration
+ // is nullptr.
+ if (EncounteredME) {
+ const auto *FD = dyn_cast<FieldDecl>(EncounteredME->getMemberDecl());
+ unsigned FieldIndex = FD->getFieldIndex();
+
+ // Update info about the lowest and highest elements for this struct
+ if (!PartialStruct.Base.isValid()) {
+ PartialStruct.LowestElem = {FieldIndex, LB};
+ PartialStruct.HighestElem = {FieldIndex, LB};
+ PartialStruct.Base = BP;
+ } else if (FieldIndex < PartialStruct.LowestElem.first) {
+ PartialStruct.LowestElem = {FieldIndex, LB};
+ } else if (FieldIndex > PartialStruct.HighestElem.first) {
+ PartialStruct.HighestElem = {FieldIndex, LB};
+ }
+ }
// If we have a final array section, we are done with this expression.
if (IsFinalArraySection)
@@ -6541,11 +7107,11 @@ private:
}
}
- /// \brief Return the adjusted map modifiers if the declaration a capture
- /// refers to appears in a first-private clause. This is expected to be used
- /// only with directives that start with 'target'.
- unsigned adjustMapModifiersForPrivateClauses(const CapturedStmt::Capture &Cap,
- unsigned CurrentModifiers) {
+ /// Return the adjusted map modifiers if the declaration a capture refers to
+ /// appears in a first-private clause. This is expected to be used only with
+ /// directives that start with 'target'.
+ MappableExprsHandler::OpenMPOffloadMappingFlags
+ getMapModifiersForPrivateClauses(const CapturedStmt::Capture &Cap) const {
assert(Cap.capturesVariable() && "Expected capture by reference only!");
// A first private variable captured by reference will use only the
@@ -6554,15 +7120,29 @@ private:
if (FirstPrivateDecls.count(Cap.getCapturedVar()))
return MappableExprsHandler::OMP_MAP_PRIVATE |
MappableExprsHandler::OMP_MAP_TO;
- // Reduction variable will use only the 'private ptr' and 'map to_from'
- // flag.
- if (ReductionDecls.count(Cap.getCapturedVar())) {
- return MappableExprsHandler::OMP_MAP_TO |
- MappableExprsHandler::OMP_MAP_FROM;
- }
+ return MappableExprsHandler::OMP_MAP_TO |
+ MappableExprsHandler::OMP_MAP_FROM;
+ }
+
+ static OpenMPOffloadMappingFlags getMemberOfFlag(unsigned Position) {
+ // Member of is given by the 16 MSB of the flag, so rotate by 48 bits.
+ return static_cast<OpenMPOffloadMappingFlags>(((uint64_t)Position + 1)
+ << 48);
+ }
+
+ static void setCorrectMemberOfFlag(OpenMPOffloadMappingFlags &Flags,
+ OpenMPOffloadMappingFlags MemberOfFlag) {
+ // If the entry is PTR_AND_OBJ but has not been marked with the special
+ // placeholder value 0xFFFF in the MEMBER_OF field, then it should not be
+ // marked as MEMBER_OF.
+ if ((Flags & OMP_MAP_PTR_AND_OBJ) &&
+ ((Flags & OMP_MAP_MEMBER_OF) != OMP_MAP_MEMBER_OF))
+ return;
- // We didn't modify anything.
- return CurrentModifiers;
+ // Reset the placeholder value to prepare the flag for the assignment of the
+ // proper MEMBER_OF value.
+ Flags &= ~OMP_MAP_MEMBER_OF;
+ Flags |= MemberOfFlag;
}
public:
@@ -6573,58 +7153,54 @@ public:
for (const auto *D : C->varlists())
FirstPrivateDecls.insert(
cast<VarDecl>(cast<DeclRefExpr>(D)->getDecl())->getCanonicalDecl());
- for (const auto *C : Dir.getClausesOfKind<OMPReductionClause>()) {
- for (const auto *D : C->varlists()) {
- ReductionDecls.insert(
- cast<VarDecl>(cast<DeclRefExpr>(D)->getDecl())->getCanonicalDecl());
- }
- }
// Extract device pointer clause information.
for (const auto *C : Dir.getClausesOfKind<OMPIsDevicePtrClause>())
for (auto L : C->component_lists())
DevPointersMap[L.first].push_back(L.second);
}
- /// \brief Generate all the base pointers, section pointers, sizes and map
+ /// Generate code for the combined entry if we have a partially mapped struct
+ /// and take care of the mapping flags of the arguments corresponding to
+ /// individual struct members.
+ void emitCombinedEntry(MapBaseValuesArrayTy &BasePointers,
+ MapValuesArrayTy &Pointers, MapValuesArrayTy &Sizes,
+ MapFlagsArrayTy &Types, MapFlagsArrayTy &CurTypes,
+ const StructRangeInfoTy &PartialStruct) const {
+ // Base is the base of the struct
+ BasePointers.push_back(PartialStruct.Base.getPointer());
+ // Pointer is the address of the lowest element
+ llvm::Value *LB = PartialStruct.LowestElem.second.getPointer();
+ Pointers.push_back(LB);
+ // Size is (addr of {highest+1} element) - (addr of lowest element)
+ llvm::Value *HB = PartialStruct.HighestElem.second.getPointer();
+ llvm::Value *HAddr = CGF.Builder.CreateConstGEP1_32(HB, /*Idx0=*/1);
+ llvm::Value *CLAddr = CGF.Builder.CreatePointerCast(LB, CGF.VoidPtrTy);
+ llvm::Value *CHAddr = CGF.Builder.CreatePointerCast(HAddr, CGF.VoidPtrTy);
+ llvm::Value *Diff = CGF.Builder.CreatePtrDiff(CHAddr, CLAddr);
+ llvm::Value *Size = CGF.Builder.CreateIntCast(Diff, CGF.SizeTy,
+ /*isSinged=*/false);
+ Sizes.push_back(Size);
+ // Map type is always TARGET_PARAM
+ Types.push_back(OMP_MAP_TARGET_PARAM);
+ // Remove TARGET_PARAM flag from the first element
+ (*CurTypes.begin()) &= ~OMP_MAP_TARGET_PARAM;
+
+ // All other current entries will be MEMBER_OF the combined entry
+ // (except for PTR_AND_OBJ entries which do not have a placeholder value
+ // 0xFFFF in the MEMBER_OF field).
+ OpenMPOffloadMappingFlags MemberOfFlag =
+ getMemberOfFlag(BasePointers.size() - 1);
+ for (auto &M : CurTypes)
+ setCorrectMemberOfFlag(M, MemberOfFlag);
+ }
+
+ /// Generate all the base pointers, section pointers, sizes and map
/// types for the extracted mappable expressions. Also, for each item that
/// relates with a device pointer, a pair of the relevant declaration and
/// index where it occurs is appended to the device pointers info array.
void generateAllInfo(MapBaseValuesArrayTy &BasePointers,
MapValuesArrayTy &Pointers, MapValuesArrayTy &Sizes,
MapFlagsArrayTy &Types) const {
- BasePointers.clear();
- Pointers.clear();
- Sizes.clear();
- Types.clear();
-
- struct MapInfo {
- /// Kind that defines how a device pointer has to be returned.
- enum ReturnPointerKind {
- // Don't have to return any pointer.
- RPK_None,
- // Pointer is the base of the declaration.
- RPK_Base,
- // Pointer is a member of the base declaration - 'this'
- RPK_Member,
- // Pointer is a reference and a member of the base declaration - 'this'
- RPK_MemberReference,
- };
- OMPClauseMappableExprCommon::MappableExprComponentListRef Components;
- OpenMPMapClauseKind MapType = OMPC_MAP_unknown;
- OpenMPMapClauseKind MapTypeModifier = OMPC_MAP_unknown;
- ReturnPointerKind ReturnDevicePointer = RPK_None;
- bool IsImplicit = false;
-
- MapInfo() = default;
- MapInfo(
- OMPClauseMappableExprCommon::MappableExprComponentListRef Components,
- OpenMPMapClauseKind MapType, OpenMPMapClauseKind MapTypeModifier,
- ReturnPointerKind ReturnDevicePointer, bool IsImplicit)
- : Components(Components), MapType(MapType),
- MapTypeModifier(MapTypeModifier),
- ReturnDevicePointer(ReturnDevicePointer), IsImplicit(IsImplicit) {}
- };
-
// We have to process the component lists that relate with the same
// declaration in a single chunk so that we can generate the map flags
// correctly. Therefore, we organize all lists in a map.
@@ -6636,7 +7212,7 @@ public:
const ValueDecl *D,
OMPClauseMappableExprCommon::MappableExprComponentListRef L,
OpenMPMapClauseKind MapType, OpenMPMapClauseKind MapModifier,
- MapInfo::ReturnPointerKind ReturnDevicePointer, bool IsImplicit) {
+ bool ReturnDevicePointer, bool IsImplicit) {
const ValueDecl *VD =
D ? cast<ValueDecl>(D->getCanonicalDecl()) : nullptr;
Info[VD].emplace_back(L, MapType, MapModifier, ReturnDevicePointer,
@@ -6644,33 +7220,39 @@ public:
};
// FIXME: MSVC 2013 seems to require this-> to find member CurDir.
- for (auto *C : this->CurDir.getClausesOfKind<OMPMapClause>())
- for (auto L : C->component_lists()) {
+ for (const auto *C : this->CurDir.getClausesOfKind<OMPMapClause>())
+ for (const auto &L : C->component_lists()) {
InfoGen(L.first, L.second, C->getMapType(), C->getMapTypeModifier(),
- MapInfo::RPK_None, C->isImplicit());
+ /*ReturnDevicePointer=*/false, C->isImplicit());
}
- for (auto *C : this->CurDir.getClausesOfKind<OMPToClause>())
- for (auto L : C->component_lists()) {
+ for (const auto *C : this->CurDir.getClausesOfKind<OMPToClause>())
+ for (const auto &L : C->component_lists()) {
InfoGen(L.first, L.second, OMPC_MAP_to, OMPC_MAP_unknown,
- MapInfo::RPK_None, C->isImplicit());
+ /*ReturnDevicePointer=*/false, C->isImplicit());
}
- for (auto *C : this->CurDir.getClausesOfKind<OMPFromClause>())
- for (auto L : C->component_lists()) {
+ for (const auto *C : this->CurDir.getClausesOfKind<OMPFromClause>())
+ for (const auto &L : C->component_lists()) {
InfoGen(L.first, L.second, OMPC_MAP_from, OMPC_MAP_unknown,
- MapInfo::RPK_None, C->isImplicit());
+ /*ReturnDevicePointer=*/false, C->isImplicit());
}
// Look at the use_device_ptr clause information and mark the existing map
// entries as such. If there is no map information for an entry in the
// use_device_ptr list, we create one with map type 'alloc' and zero size
- // section. It is the user fault if that was not mapped before.
+ // section. It is the user fault if that was not mapped before. If there is
+ // no map information and the pointer is a struct member, then we defer the
+ // emission of that entry until the whole struct has been processed.
+ llvm::MapVector<const ValueDecl *, SmallVector<DeferredDevicePtrEntryTy, 4>>
+ DeferredInfo;
+
// FIXME: MSVC 2013 seems to require this-> to find member CurDir.
- for (auto *C : this->CurDir.getClausesOfKind<OMPUseDevicePtrClause>())
- for (auto L : C->component_lists()) {
+ for (const auto *C :
+ this->CurDir.getClausesOfKind<OMPUseDevicePtrClause>()) {
+ for (const auto &L : C->component_lists()) {
assert(!L.second.empty() && "Not expecting empty list of components!");
const ValueDecl *VD = L.second.back().getAssociatedDeclaration();
VD = cast<ValueDecl>(VD->getCanonicalDecl());
- auto *IE = L.second.back().getAssociatedExpression();
+ const Expr *IE = L.second.back().getAssociatedExpression();
// If the first component is a member expression, we have to look into
// 'this', which maps to null in the map of map information. Otherwise
// look directly for the information.
@@ -6686,113 +7268,135 @@ public:
// If we found a map entry, signal that the pointer has to be returned
// and move on to the next declaration.
if (CI != It->second.end()) {
- CI->ReturnDevicePointer = isa<MemberExpr>(IE)
- ? (VD->getType()->isReferenceType()
- ? MapInfo::RPK_MemberReference
- : MapInfo::RPK_Member)
- : MapInfo::RPK_Base;
+ CI->ReturnDevicePointer = true;
continue;
}
}
// We didn't find any match in our map information - generate a zero
- // size array section.
+ // size array section - if the pointer is a struct member we defer this
+ // action until the whole struct has been processed.
// FIXME: MSVC 2013 seems to require this-> to find member CGF.
- llvm::Value *Ptr =
- this->CGF
- .EmitLoadOfLValue(this->CGF.EmitLValue(IE), SourceLocation())
- .getScalarVal();
- BasePointers.push_back({Ptr, VD});
- Pointers.push_back(Ptr);
- Sizes.push_back(llvm::Constant::getNullValue(this->CGF.SizeTy));
- Types.push_back(OMP_MAP_RETURN_PARAM | OMP_MAP_TARGET_PARAM);
+ if (isa<MemberExpr>(IE)) {
+ // Insert the pointer into Info to be processed by
+ // generateInfoForComponentList. Because it is a member pointer
+ // without a pointee, no entry will be generated for it, therefore
+ // we need to generate one after the whole struct has been processed.
+ // Nonetheless, generateInfoForComponentList must be called to take
+ // the pointer into account for the calculation of the range of the
+ // partial struct.
+ InfoGen(nullptr, L.second, OMPC_MAP_unknown, OMPC_MAP_unknown,
+ /*ReturnDevicePointer=*/false, C->isImplicit());
+ DeferredInfo[nullptr].emplace_back(IE, VD);
+ } else {
+ llvm::Value *Ptr = this->CGF.EmitLoadOfScalar(
+ this->CGF.EmitLValue(IE), IE->getExprLoc());
+ BasePointers.emplace_back(Ptr, VD);
+ Pointers.push_back(Ptr);
+ Sizes.push_back(llvm::Constant::getNullValue(this->CGF.SizeTy));
+ Types.push_back(OMP_MAP_RETURN_PARAM | OMP_MAP_TARGET_PARAM);
+ }
}
+ }
- for (auto &M : Info) {
+ for (const auto &M : Info) {
// We need to know when we generate information for the first component
// associated with a capture, because the mapping flags depend on it.
bool IsFirstComponentList = true;
- for (MapInfo &L : M.second) {
+
+ // Temporary versions of arrays
+ MapBaseValuesArrayTy CurBasePointers;
+ MapValuesArrayTy CurPointers;
+ MapValuesArrayTy CurSizes;
+ MapFlagsArrayTy CurTypes;
+ StructRangeInfoTy PartialStruct;
+
+ for (const MapInfo &L : M.second) {
assert(!L.Components.empty() &&
"Not expecting declaration with no component lists.");
// Remember the current base pointer index.
- unsigned CurrentBasePointersIdx = BasePointers.size();
+ unsigned CurrentBasePointersIdx = CurBasePointers.size();
// FIXME: MSVC 2013 seems to require this-> to find the member method.
this->generateInfoForComponentList(
- L.MapType, L.MapTypeModifier, L.Components, BasePointers, Pointers,
- Sizes, Types, IsFirstComponentList, L.IsImplicit);
+ L.MapType, L.MapTypeModifier, L.Components, CurBasePointers,
+ CurPointers, CurSizes, CurTypes, PartialStruct,
+ IsFirstComponentList, L.IsImplicit);
// If this entry relates with a device pointer, set the relevant
// declaration and add the 'return pointer' flag.
- if (IsFirstComponentList &&
- L.ReturnDevicePointer != MapInfo::RPK_None) {
- // If the pointer is not the base of the map, we need to skip the
- // base. If it is a reference in a member field, we also need to skip
- // the map of the reference.
- if (L.ReturnDevicePointer != MapInfo::RPK_Base) {
- ++CurrentBasePointersIdx;
- if (L.ReturnDevicePointer == MapInfo::RPK_MemberReference)
- ++CurrentBasePointersIdx;
- }
- assert(BasePointers.size() > CurrentBasePointersIdx &&
+ if (L.ReturnDevicePointer) {
+ assert(CurBasePointers.size() > CurrentBasePointersIdx &&
"Unexpected number of mapped base pointers.");
- auto *RelevantVD = L.Components.back().getAssociatedDeclaration();
+ const ValueDecl *RelevantVD =
+ L.Components.back().getAssociatedDeclaration();
assert(RelevantVD &&
"No relevant declaration related with device pointer??");
- BasePointers[CurrentBasePointersIdx].setDevicePtrDecl(RelevantVD);
- Types[CurrentBasePointersIdx] |= OMP_MAP_RETURN_PARAM;
+ CurBasePointers[CurrentBasePointersIdx].setDevicePtrDecl(RelevantVD);
+ CurTypes[CurrentBasePointersIdx] |= OMP_MAP_RETURN_PARAM;
}
IsFirstComponentList = false;
}
+
+ // Append any pending zero-length pointers which are struct members and
+ // used with use_device_ptr.
+ auto CI = DeferredInfo.find(M.first);
+ if (CI != DeferredInfo.end()) {
+ for (const DeferredDevicePtrEntryTy &L : CI->second) {
+ llvm::Value *BasePtr = this->CGF.EmitLValue(L.IE).getPointer();
+ llvm::Value *Ptr = this->CGF.EmitLoadOfScalar(
+ this->CGF.EmitLValue(L.IE), L.IE->getExprLoc());
+ CurBasePointers.emplace_back(BasePtr, L.VD);
+ CurPointers.push_back(Ptr);
+ CurSizes.push_back(llvm::Constant::getNullValue(this->CGF.SizeTy));
+ // Entry is PTR_AND_OBJ and RETURN_PARAM. Also, set the placeholder
+ // value MEMBER_OF=FFFF so that the entry is later updated with the
+ // correct value of MEMBER_OF.
+ CurTypes.push_back(OMP_MAP_PTR_AND_OBJ | OMP_MAP_RETURN_PARAM |
+ OMP_MAP_MEMBER_OF);
+ }
+ }
+
+ // If there is an entry in PartialStruct it means we have a struct with
+ // individual members mapped. Emit an extra combined entry.
+ if (PartialStruct.Base.isValid())
+ emitCombinedEntry(BasePointers, Pointers, Sizes, Types, CurTypes,
+ PartialStruct);
+
+ // We need to append the results of this capture to what we already have.
+ BasePointers.append(CurBasePointers.begin(), CurBasePointers.end());
+ Pointers.append(CurPointers.begin(), CurPointers.end());
+ Sizes.append(CurSizes.begin(), CurSizes.end());
+ Types.append(CurTypes.begin(), CurTypes.end());
}
}
- /// \brief Generate the base pointers, section pointers, sizes and map types
+ /// Generate the base pointers, section pointers, sizes and map types
/// associated to a given capture.
void generateInfoForCapture(const CapturedStmt::Capture *Cap,
llvm::Value *Arg,
MapBaseValuesArrayTy &BasePointers,
MapValuesArrayTy &Pointers,
- MapValuesArrayTy &Sizes,
- MapFlagsArrayTy &Types) const {
+ MapValuesArrayTy &Sizes, MapFlagsArrayTy &Types,
+ StructRangeInfoTy &PartialStruct) const {
assert(!Cap->capturesVariableArrayType() &&
"Not expecting to generate map info for a variable array type!");
- BasePointers.clear();
- Pointers.clear();
- Sizes.clear();
- Types.clear();
-
// We need to know when we generating information for the first component
// associated with a capture, because the mapping flags depend on it.
bool IsFirstComponentList = true;
- const ValueDecl *VD =
- Cap->capturesThis()
- ? nullptr
- : cast<ValueDecl>(Cap->getCapturedVar()->getCanonicalDecl());
+ const ValueDecl *VD = Cap->capturesThis()
+ ? nullptr
+ : Cap->getCapturedVar()->getCanonicalDecl();
// If this declaration appears in a is_device_ptr clause we just have to
// pass the pointer by value. If it is a reference to a declaration, we just
- // pass its value, otherwise, if it is a member expression, we need to map
- // 'to' the field.
- if (!VD) {
- auto It = DevPointersMap.find(VD);
- if (It != DevPointersMap.end()) {
- for (auto L : It->second) {
- generateInfoForComponentList(
- /*MapType=*/OMPC_MAP_to, /*MapTypeModifier=*/OMPC_MAP_unknown, L,
- BasePointers, Pointers, Sizes, Types, IsFirstComponentList,
- /*IsImplicit=*/false);
- IsFirstComponentList = false;
- }
- return;
- }
- } else if (DevPointersMap.count(VD)) {
- BasePointers.push_back({Arg, VD});
+ // pass its value.
+ if (DevPointersMap.count(VD)) {
+ BasePointers.emplace_back(Arg, VD);
Pointers.push_back(Arg);
Sizes.push_back(CGF.getTypeSize(CGF.getContext().VoidPtrTy));
Types.push_back(OMP_MAP_LITERAL | OMP_MAP_TARGET_PARAM);
@@ -6800,35 +7404,63 @@ public:
}
// FIXME: MSVC 2013 seems to require this-> to find member CurDir.
- for (auto *C : this->CurDir.getClausesOfKind<OMPMapClause>())
- for (auto L : C->decl_component_lists(VD)) {
+ for (const auto *C : this->CurDir.getClausesOfKind<OMPMapClause>())
+ for (const auto &L : C->decl_component_lists(VD)) {
assert(L.first == VD &&
"We got information for the wrong declaration??");
assert(!L.second.empty() &&
"Not expecting declaration with no component lists.");
- generateInfoForComponentList(
- C->getMapType(), C->getMapTypeModifier(), L.second, BasePointers,
- Pointers, Sizes, Types, IsFirstComponentList, C->isImplicit());
+ generateInfoForComponentList(C->getMapType(), C->getMapTypeModifier(),
+ L.second, BasePointers, Pointers, Sizes,
+ Types, PartialStruct, IsFirstComponentList,
+ C->isImplicit());
IsFirstComponentList = false;
}
+ }
- return;
+ /// Generate the base pointers, section pointers, sizes and map types
+ /// associated with the declare target link variables.
+ void generateInfoForDeclareTargetLink(MapBaseValuesArrayTy &BasePointers,
+ MapValuesArrayTy &Pointers,
+ MapValuesArrayTy &Sizes,
+ MapFlagsArrayTy &Types) const {
+ // Map other list items in the map clause which are not captured variables
+ // but "declare target link" global variables.,
+ for (const auto *C : this->CurDir.getClausesOfKind<OMPMapClause>()) {
+ for (const auto &L : C->component_lists()) {
+ if (!L.first)
+ continue;
+ const auto *VD = dyn_cast<VarDecl>(L.first);
+ if (!VD)
+ continue;
+ llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(VD);
+ if (!Res || *Res != OMPDeclareTargetDeclAttr::MT_Link)
+ continue;
+ StructRangeInfoTy PartialStruct;
+ generateInfoForComponentList(
+ C->getMapType(), C->getMapTypeModifier(), L.second, BasePointers,
+ Pointers, Sizes, Types, PartialStruct,
+ /*IsFirstComponentList=*/true, C->isImplicit());
+ assert(!PartialStruct.Base.isValid() &&
+ "No partial structs for declare target link expected.");
+ }
+ }
}
- /// \brief Generate the default map information for a given capture \a CI,
+ /// Generate the default map information for a given capture \a CI,
/// record field declaration \a RI and captured value \a CV.
void generateDefaultMapInfo(const CapturedStmt::Capture &CI,
const FieldDecl &RI, llvm::Value *CV,
MapBaseValuesArrayTy &CurBasePointers,
MapValuesArrayTy &CurPointers,
MapValuesArrayTy &CurSizes,
- MapFlagsArrayTy &CurMapTypes) {
-
+ MapFlagsArrayTy &CurMapTypes) const {
// Do the default mapping.
if (CI.capturesThis()) {
CurBasePointers.push_back(CV);
CurPointers.push_back(CV);
- const PointerType *PtrTy = cast<PointerType>(RI.getType().getTypePtr());
+ const auto *PtrTy = cast<PointerType>(RI.getType().getTypePtr());
CurSizes.push_back(CGF.getTypeSize(PtrTy->getPointeeType()));
// Default map type.
CurMapTypes.push_back(OMP_MAP_TO | OMP_MAP_FROM);
@@ -6843,7 +7475,7 @@ public:
} else {
// Pointers are implicitly mapped with a zero size and no flags
// (other than first map that is added for all implicit maps).
- CurMapTypes.push_back(0u);
+ CurMapTypes.push_back(OMP_MAP_NONE);
CurSizes.push_back(llvm::Constant::getNullValue(CGF.SizeTy));
}
} else {
@@ -6851,30 +7483,30 @@ public:
CurBasePointers.push_back(CV);
CurPointers.push_back(CV);
- const ReferenceType *PtrTy =
- cast<ReferenceType>(RI.getType().getTypePtr());
+ const auto *PtrTy = cast<ReferenceType>(RI.getType().getTypePtr());
QualType ElementType = PtrTy->getPointeeType();
CurSizes.push_back(CGF.getTypeSize(ElementType));
// The default map type for a scalar/complex type is 'to' because by
// default the value doesn't have to be retrieved. For an aggregate
// type, the default is 'tofrom'.
- CurMapTypes.emplace_back(adjustMapModifiersForPrivateClauses(
- CI, ElementType->isAggregateType() ? (OMP_MAP_TO | OMP_MAP_FROM)
- : OMP_MAP_TO));
+ CurMapTypes.push_back(getMapModifiersForPrivateClauses(CI));
}
// Every default map produces a single argument which is a target parameter.
CurMapTypes.back() |= OMP_MAP_TARGET_PARAM;
+
+ // Add flag stating this is an implicit map.
+ CurMapTypes.back() |= OMP_MAP_IMPLICIT;
}
};
enum OpenMPOffloadingReservedDeviceIDs {
- /// \brief Device ID if the device was not defined, runtime should get it
+ /// Device ID if the device was not defined, runtime should get it
/// from environment variables in the spec.
OMP_DEVICEID_UNDEF = -1,
};
} // anonymous namespace
-/// \brief Emit the arrays used to pass the captures and map information to the
+/// Emit the arrays used to pass the captures and map information to the
/// offloading runtime library. If there is no map or capture information,
/// return nullptr by reference.
static void
@@ -6884,8 +7516,8 @@ emitOffloadingArrays(CodeGenFunction &CGF,
MappableExprsHandler::MapValuesArrayTy &Sizes,
MappableExprsHandler::MapFlagsArrayTy &MapTypes,
CGOpenMPRuntime::TargetDataInfo &Info) {
- auto &CGM = CGF.CGM;
- auto &Ctx = CGF.getContext();
+ CodeGenModule &CGM = CGF.CGM;
+ ASTContext &Ctx = CGF.getContext();
// Reset the array information.
Info.clearArrayInfo();
@@ -6895,7 +7527,7 @@ emitOffloadingArrays(CodeGenFunction &CGF,
// Detect if we have any capture size requiring runtime evaluation of the
// size so that a constant array could be eventually used.
bool hasRuntimeEvaluationCaptureSize = false;
- for (auto *S : Sizes)
+ for (llvm::Value *S : Sizes)
if (!isa<llvm::Constant>(S)) {
hasRuntimeEvaluationCaptureSize = true;
break;
@@ -6924,48 +7556,53 @@ emitOffloadingArrays(CodeGenFunction &CGF,
// We expect all the sizes to be constant, so we collect them to create
// a constant array.
SmallVector<llvm::Constant *, 16> ConstSizes;
- for (auto S : Sizes)
+ for (llvm::Value *S : Sizes)
ConstSizes.push_back(cast<llvm::Constant>(S));
auto *SizesArrayInit = llvm::ConstantArray::get(
llvm::ArrayType::get(CGM.SizeTy, ConstSizes.size()), ConstSizes);
+ std::string Name = CGM.getOpenMPRuntime().getName({"offload_sizes"});
auto *SizesArrayGbl = new llvm::GlobalVariable(
CGM.getModule(), SizesArrayInit->getType(),
/*isConstant=*/true, llvm::GlobalValue::PrivateLinkage,
- SizesArrayInit, ".offload_sizes");
+ SizesArrayInit, Name);
SizesArrayGbl->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
Info.SizesArray = SizesArrayGbl;
}
// The map types are always constant so we don't need to generate code to
// fill arrays. Instead, we create an array constant.
+ SmallVector<uint64_t, 4> Mapping(MapTypes.size(), 0);
+ llvm::copy(MapTypes, Mapping.begin());
llvm::Constant *MapTypesArrayInit =
- llvm::ConstantDataArray::get(CGF.Builder.getContext(), MapTypes);
+ llvm::ConstantDataArray::get(CGF.Builder.getContext(), Mapping);
+ std::string MaptypesName =
+ CGM.getOpenMPRuntime().getName({"offload_maptypes"});
auto *MapTypesArrayGbl = new llvm::GlobalVariable(
CGM.getModule(), MapTypesArrayInit->getType(),
/*isConstant=*/true, llvm::GlobalValue::PrivateLinkage,
- MapTypesArrayInit, ".offload_maptypes");
+ MapTypesArrayInit, MaptypesName);
MapTypesArrayGbl->setUnnamedAddr(llvm::GlobalValue::UnnamedAddr::Global);
Info.MapTypesArray = MapTypesArrayGbl;
- for (unsigned i = 0; i < Info.NumberOfPtrs; ++i) {
- llvm::Value *BPVal = *BasePointers[i];
+ for (unsigned I = 0; I < Info.NumberOfPtrs; ++I) {
+ llvm::Value *BPVal = *BasePointers[I];
llvm::Value *BP = CGF.Builder.CreateConstInBoundsGEP2_32(
llvm::ArrayType::get(CGM.VoidPtrTy, Info.NumberOfPtrs),
- Info.BasePointersArray, 0, i);
+ Info.BasePointersArray, 0, I);
BP = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
BP, BPVal->getType()->getPointerTo(/*AddrSpace=*/0));
Address BPAddr(BP, Ctx.getTypeAlignInChars(Ctx.VoidPtrTy));
CGF.Builder.CreateStore(BPVal, BPAddr);
if (Info.requiresDevicePointerInfo())
- if (auto *DevVD = BasePointers[i].getDevicePtrDecl())
- Info.CaptureDeviceAddrMap.insert(std::make_pair(DevVD, BPAddr));
+ if (const ValueDecl *DevVD = BasePointers[I].getDevicePtrDecl())
+ Info.CaptureDeviceAddrMap.try_emplace(DevVD, BPAddr);
- llvm::Value *PVal = Pointers[i];
+ llvm::Value *PVal = Pointers[I];
llvm::Value *P = CGF.Builder.CreateConstInBoundsGEP2_32(
llvm::ArrayType::get(CGM.VoidPtrTy, Info.NumberOfPtrs),
- Info.PointersArray, 0, i);
+ Info.PointersArray, 0, I);
P = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
P, PVal->getType()->getPointerTo(/*AddrSpace=*/0));
Address PAddr(P, Ctx.getTypeAlignInChars(Ctx.VoidPtrTy));
@@ -6976,22 +7613,22 @@ emitOffloadingArrays(CodeGenFunction &CGF,
llvm::ArrayType::get(CGM.SizeTy, Info.NumberOfPtrs),
Info.SizesArray,
/*Idx0=*/0,
- /*Idx1=*/i);
+ /*Idx1=*/I);
Address SAddr(S, Ctx.getTypeAlignInChars(Ctx.getSizeType()));
CGF.Builder.CreateStore(
- CGF.Builder.CreateIntCast(Sizes[i], CGM.SizeTy, /*isSigned=*/true),
+ CGF.Builder.CreateIntCast(Sizes[I], CGM.SizeTy, /*isSigned=*/true),
SAddr);
}
}
}
}
-/// \brief Emit the arguments to be passed to the runtime library based on the
+/// Emit the arguments to be passed to the runtime library based on the
/// arrays of pointers, sizes and map types.
static void emitOffloadingArraysArgument(
CodeGenFunction &CGF, llvm::Value *&BasePointersArrayArg,
llvm::Value *&PointersArrayArg, llvm::Value *&SizesArrayArg,
llvm::Value *&MapTypesArrayArg, CGOpenMPRuntime::TargetDataInfo &Info) {
- auto &CGM = CGF.CGM;
+ CodeGenModule &CGM = CGF.CGM;
if (Info.NumberOfPtrs) {
BasePointersArrayArg = CGF.Builder.CreateConstInBoundsGEP2_32(
llvm::ArrayType::get(CGM.VoidPtrTy, Info.NumberOfPtrs),
@@ -7023,86 +7660,27 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
const OMPExecutableDirective &D,
llvm::Value *OutlinedFn,
llvm::Value *OutlinedFnID,
- const Expr *IfCond, const Expr *Device,
- ArrayRef<llvm::Value *> CapturedVars) {
+ const Expr *IfCond, const Expr *Device) {
if (!CGF.HaveInsertPoint())
return;
assert(OutlinedFn && "Invalid outlined function!");
- // Fill up the arrays with all the captured variables.
- MappableExprsHandler::MapValuesArrayTy KernelArgs;
- MappableExprsHandler::MapBaseValuesArrayTy BasePointers;
- MappableExprsHandler::MapValuesArrayTy Pointers;
- MappableExprsHandler::MapValuesArrayTy Sizes;
- MappableExprsHandler::MapFlagsArrayTy MapTypes;
-
- MappableExprsHandler::MapBaseValuesArrayTy CurBasePointers;
- MappableExprsHandler::MapValuesArrayTy CurPointers;
- MappableExprsHandler::MapValuesArrayTy CurSizes;
- MappableExprsHandler::MapFlagsArrayTy CurMapTypes;
-
- // Get mappable expression information.
- MappableExprsHandler MEHandler(D, CGF);
-
- const CapturedStmt &CS = *cast<CapturedStmt>(D.getAssociatedStmt());
- auto RI = CS.getCapturedRecordDecl()->field_begin();
- auto CV = CapturedVars.begin();
- for (CapturedStmt::const_capture_iterator CI = CS.capture_begin(),
- CE = CS.capture_end();
- CI != CE; ++CI, ++RI, ++CV) {
- CurBasePointers.clear();
- CurPointers.clear();
- CurSizes.clear();
- CurMapTypes.clear();
-
- // VLA sizes are passed to the outlined region by copy and do not have map
- // information associated.
- if (CI->capturesVariableArrayType()) {
- CurBasePointers.push_back(*CV);
- CurPointers.push_back(*CV);
- CurSizes.push_back(CGF.getTypeSize(RI->getType()));
- // Copy to the device as an argument. No need to retrieve it.
- CurMapTypes.push_back(MappableExprsHandler::OMP_MAP_LITERAL |
- MappableExprsHandler::OMP_MAP_TARGET_PARAM);
- } else {
- // If we have any information in the map clause, we use it, otherwise we
- // just do a default mapping.
- MEHandler.generateInfoForCapture(CI, *CV, CurBasePointers, CurPointers,
- CurSizes, CurMapTypes);
- if (CurBasePointers.empty())
- MEHandler.generateDefaultMapInfo(*CI, **RI, *CV, CurBasePointers,
- CurPointers, CurSizes, CurMapTypes);
- }
- // We expect to have at least an element of information for this capture.
- assert(!CurBasePointers.empty() && "Non-existing map pointer for capture!");
- assert(CurBasePointers.size() == CurPointers.size() &&
- CurBasePointers.size() == CurSizes.size() &&
- CurBasePointers.size() == CurMapTypes.size() &&
- "Inconsistent map information sizes!");
-
- // The kernel args are always the first elements of the base pointers
- // associated with a capture.
- KernelArgs.push_back(*CurBasePointers.front());
- // We need to append the results of this capture to what we already have.
- BasePointers.append(CurBasePointers.begin(), CurBasePointers.end());
- Pointers.append(CurPointers.begin(), CurPointers.end());
- Sizes.append(CurSizes.begin(), CurSizes.end());
- MapTypes.append(CurMapTypes.begin(), CurMapTypes.end());
- }
+ const bool RequiresOuterTask = D.hasClausesOfKind<OMPDependClause>();
+ llvm::SmallVector<llvm::Value *, 16> CapturedVars;
+ const CapturedStmt &CS = *D.getCapturedStmt(OMPD_target);
+ auto &&ArgsCodegen = [&CS, &CapturedVars](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ };
+ emitInlinedDirective(CGF, OMPD_unknown, ArgsCodegen);
+ CodeGenFunction::OMPTargetDataInfo InputInfo;
+ llvm::Value *MapTypesArray = nullptr;
// Fill up the pointer arrays and transfer execution to the device.
- auto &&ThenGen = [this, &BasePointers, &Pointers, &Sizes, &MapTypes, Device,
- OutlinedFn, OutlinedFnID, &D,
- &KernelArgs](CodeGenFunction &CGF, PrePostActionTy &) {
- auto &RT = CGF.CGM.getOpenMPRuntime();
- // Emit the offloading arrays.
- TargetDataInfo Info;
- emitOffloadingArrays(CGF, BasePointers, Pointers, Sizes, MapTypes, Info);
- emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, Info);
-
+ auto &&ThenGen = [this, Device, OutlinedFn, OutlinedFnID, &D, &InputInfo,
+ &MapTypesArray, &CS, RequiresOuterTask,
+ &CapturedVars](CodeGenFunction &CGF, PrePostActionTy &) {
// On top of the arrays that were filled up, the target offloading call
// takes as arguments the device id as well as the host pointer. The host
// pointer is used by the runtime library to identify the current target
@@ -7125,13 +7703,14 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
}
// Emit the number of elements in the offloading arrays.
- llvm::Value *PointerNum = CGF.Builder.getInt32(BasePointers.size());
+ llvm::Value *PointerNum =
+ CGF.Builder.getInt32(InputInfo.NumberOfTargetItems);
// Return value of the runtime offloading call.
llvm::Value *Return;
- auto *NumTeams = emitNumTeamsForTargetDirective(RT, CGF, D);
- auto *NumThreads = emitNumThreadsForTargetDirective(RT, CGF, D);
+ llvm::Value *NumTeams = emitNumTeamsForTargetDirective(*this, CGF, D);
+ llvm::Value *NumThreads = emitNumThreadsForTargetDirective(*this, CGF, D);
bool HasNowait = D.hasClausesOfKind<OMPNowaitClause>();
// The target region is an outlined function launched by the runtime
@@ -7169,25 +7748,30 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
// passed to the runtime library - a 32-bit integer with the value zero.
assert(NumThreads && "Thread limit expression should be available along "
"with number of teams.");
- llvm::Value *OffloadingArgs[] = {
- DeviceID, OutlinedFnID,
- PointerNum, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray, NumTeams,
- NumThreads};
+ llvm::Value *OffloadingArgs[] = {DeviceID,
+ OutlinedFnID,
+ PointerNum,
+ InputInfo.BasePointersArray.getPointer(),
+ InputInfo.PointersArray.getPointer(),
+ InputInfo.SizesArray.getPointer(),
+ MapTypesArray,
+ NumTeams,
+ NumThreads};
Return = CGF.EmitRuntimeCall(
- RT.createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_teams_nowait
- : OMPRTL__tgt_target_teams),
+ createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_teams_nowait
+ : OMPRTL__tgt_target_teams),
OffloadingArgs);
} else {
- llvm::Value *OffloadingArgs[] = {
- DeviceID, OutlinedFnID,
- PointerNum, Info.BasePointersArray,
- Info.PointersArray, Info.SizesArray,
- Info.MapTypesArray};
+ llvm::Value *OffloadingArgs[] = {DeviceID,
+ OutlinedFnID,
+ PointerNum,
+ InputInfo.BasePointersArray.getPointer(),
+ InputInfo.PointersArray.getPointer(),
+ InputInfo.SizesArray.getPointer(),
+ MapTypesArray};
Return = CGF.EmitRuntimeCall(
- RT.createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_nowait
- : OMPRTL__tgt_target),
+ createRuntimeFunction(HasNowait ? OMPRTL__tgt_target_nowait
+ : OMPRTL__tgt_target),
OffloadingArgs);
}
@@ -7200,17 +7784,120 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
CGF.Builder.CreateCondBr(Failed, OffloadFailedBlock, OffloadContBlock);
CGF.EmitBlock(OffloadFailedBlock);
- emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, KernelArgs);
+ if (RequiresOuterTask) {
+ CapturedVars.clear();
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ }
+ emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, CapturedVars);
CGF.EmitBranch(OffloadContBlock);
CGF.EmitBlock(OffloadContBlock, /*IsFinished=*/true);
};
// Notify that the host version must be executed.
- auto &&ElseGen = [this, &D, OutlinedFn, &KernelArgs](CodeGenFunction &CGF,
- PrePostActionTy &) {
- emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn,
- KernelArgs);
+ auto &&ElseGen = [this, &D, OutlinedFn, &CS, &CapturedVars,
+ RequiresOuterTask](CodeGenFunction &CGF,
+ PrePostActionTy &) {
+ if (RequiresOuterTask) {
+ CapturedVars.clear();
+ CGF.GenerateOpenMPCapturedVars(CS, CapturedVars);
+ }
+ emitOutlinedFunctionCall(CGF, D.getLocStart(), OutlinedFn, CapturedVars);
+ };
+
+ auto &&TargetThenGen = [this, &ThenGen, &D, &InputInfo, &MapTypesArray,
+ &CapturedVars, RequiresOuterTask,
+ &CS](CodeGenFunction &CGF, PrePostActionTy &) {
+ // Fill up the arrays with all the captured variables.
+ MappableExprsHandler::MapBaseValuesArrayTy BasePointers;
+ MappableExprsHandler::MapValuesArrayTy Pointers;
+ MappableExprsHandler::MapValuesArrayTy Sizes;
+ MappableExprsHandler::MapFlagsArrayTy MapTypes;
+
+ // Get mappable expression information.
+ MappableExprsHandler MEHandler(D, CGF);
+
+ auto RI = CS.getCapturedRecordDecl()->field_begin();
+ auto CV = CapturedVars.begin();
+ for (CapturedStmt::const_capture_iterator CI = CS.capture_begin(),
+ CE = CS.capture_end();
+ CI != CE; ++CI, ++RI, ++CV) {
+ MappableExprsHandler::MapBaseValuesArrayTy CurBasePointers;
+ MappableExprsHandler::MapValuesArrayTy CurPointers;
+ MappableExprsHandler::MapValuesArrayTy CurSizes;
+ MappableExprsHandler::MapFlagsArrayTy CurMapTypes;
+ MappableExprsHandler::StructRangeInfoTy PartialStruct;
+
+ // VLA sizes are passed to the outlined region by copy and do not have map
+ // information associated.
+ if (CI->capturesVariableArrayType()) {
+ CurBasePointers.push_back(*CV);
+ CurPointers.push_back(*CV);
+ CurSizes.push_back(CGF.getTypeSize(RI->getType()));
+ // Copy to the device as an argument. No need to retrieve it.
+ CurMapTypes.push_back(MappableExprsHandler::OMP_MAP_LITERAL |
+ MappableExprsHandler::OMP_MAP_TARGET_PARAM);
+ } else {
+ // If we have any information in the map clause, we use it, otherwise we
+ // just do a default mapping.
+ MEHandler.generateInfoForCapture(CI, *CV, CurBasePointers, CurPointers,
+ CurSizes, CurMapTypes, PartialStruct);
+ if (CurBasePointers.empty())
+ MEHandler.generateDefaultMapInfo(*CI, **RI, *CV, CurBasePointers,
+ CurPointers, CurSizes, CurMapTypes);
+ }
+ // We expect to have at least an element of information for this capture.
+ assert(!CurBasePointers.empty() &&
+ "Non-existing map pointer for capture!");
+ assert(CurBasePointers.size() == CurPointers.size() &&
+ CurBasePointers.size() == CurSizes.size() &&
+ CurBasePointers.size() == CurMapTypes.size() &&
+ "Inconsistent map information sizes!");
+
+ // If there is an entry in PartialStruct it means we have a struct with
+ // individual members mapped. Emit an extra combined entry.
+ if (PartialStruct.Base.isValid())
+ MEHandler.emitCombinedEntry(BasePointers, Pointers, Sizes, MapTypes,
+ CurMapTypes, PartialStruct);
+
+ // We need to append the results of this capture to what we already have.
+ BasePointers.append(CurBasePointers.begin(), CurBasePointers.end());
+ Pointers.append(CurPointers.begin(), CurPointers.end());
+ Sizes.append(CurSizes.begin(), CurSizes.end());
+ MapTypes.append(CurMapTypes.begin(), CurMapTypes.end());
+ }
+ // Map other list items in the map clause which are not captured variables
+ // but "declare target link" global variables.
+ MEHandler.generateInfoForDeclareTargetLink(BasePointers, Pointers, Sizes,
+ MapTypes);
+
+ TargetDataInfo Info;
+ // Fill up the arrays and create the arguments.
+ emitOffloadingArrays(CGF, BasePointers, Pointers, Sizes, MapTypes, Info);
+ emitOffloadingArraysArgument(CGF, Info.BasePointersArray,
+ Info.PointersArray, Info.SizesArray,
+ Info.MapTypesArray, Info);
+ InputInfo.NumberOfTargetItems = Info.NumberOfPtrs;
+ InputInfo.BasePointersArray =
+ Address(Info.BasePointersArray, CGM.getPointerAlign());
+ InputInfo.PointersArray =
+ Address(Info.PointersArray, CGM.getPointerAlign());
+ InputInfo.SizesArray = Address(Info.SizesArray, CGM.getPointerAlign());
+ MapTypesArray = Info.MapTypesArray;
+ if (RequiresOuterTask)
+ CGF.EmitOMPTargetTaskBasedDirective(D, ThenGen, InputInfo);
+ else
+ emitInlinedDirective(CGF, D.getDirectiveKind(), ThenGen);
+ };
+
+ auto &&TargetElseGen = [this, &ElseGen, &D, RequiresOuterTask](
+ CodeGenFunction &CGF, PrePostActionTy &) {
+ if (RequiresOuterTask) {
+ CodeGenFunction::OMPTargetDataInfo InputInfo;
+ CGF.EmitOMPTargetTaskBasedDirective(D, ElseGen, InputInfo);
+ } else {
+ emitInlinedDirective(CGF, D.getDirectiveKind(), ElseGen);
+ }
};
// If we have a target function ID it means that we need to support
@@ -7218,14 +7905,14 @@ void CGOpenMPRuntime::emitTargetCall(CodeGenFunction &CGF,
// regardless of the conditional in the if clause if, e.g., the user do not
// specify target triples.
if (OutlinedFnID) {
- if (IfCond)
- emitOMPIfClause(CGF, IfCond, ThenGen, ElseGen);
- else {
- RegionCodeGenTy ThenRCG(ThenGen);
+ if (IfCond) {
+ emitOMPIfClause(CGF, IfCond, TargetThenGen, TargetElseGen);
+ } else {
+ RegionCodeGenTy ThenRCG(TargetThenGen);
ThenRCG(CGF);
}
} else {
- RegionCodeGenTy ElseRCG(ElseGen);
+ RegionCodeGenTy ElseRCG(TargetElseGen);
ElseRCG(CGF);
}
}
@@ -7236,13 +7923,13 @@ void CGOpenMPRuntime::scanForTargetRegionsFunctions(const Stmt *S,
return;
// Codegen OMP target directives that offload compute to the device.
- bool requiresDeviceCodegen =
+ bool RequiresDeviceCodegen =
isa<OMPExecutableDirective>(S) &&
isOpenMPTargetExecutionDirective(
cast<OMPExecutableDirective>(S)->getDirectiveKind());
- if (requiresDeviceCodegen) {
- auto &E = *cast<OMPExecutableDirective>(S);
+ if (RequiresDeviceCodegen) {
+ const auto &E = *cast<OMPExecutableDirective>(S);
unsigned DeviceID;
unsigned FileID;
unsigned Line;
@@ -7255,66 +7942,118 @@ void CGOpenMPRuntime::scanForTargetRegionsFunctions(const Stmt *S,
ParentName, Line))
return;
- switch (S->getStmtClass()) {
- case Stmt::OMPTargetDirectiveClass:
- CodeGenFunction::EmitOMPTargetDeviceFunction(
- CGM, ParentName, cast<OMPTargetDirective>(*S));
+ switch (E.getDirectiveKind()) {
+ case OMPD_target:
+ CodeGenFunction::EmitOMPTargetDeviceFunction(CGM, ParentName,
+ cast<OMPTargetDirective>(E));
break;
- case Stmt::OMPTargetParallelDirectiveClass:
+ case OMPD_target_parallel:
CodeGenFunction::EmitOMPTargetParallelDeviceFunction(
- CGM, ParentName, cast<OMPTargetParallelDirective>(*S));
+ CGM, ParentName, cast<OMPTargetParallelDirective>(E));
break;
- case Stmt::OMPTargetTeamsDirectiveClass:
+ case OMPD_target_teams:
CodeGenFunction::EmitOMPTargetTeamsDeviceFunction(
- CGM, ParentName, cast<OMPTargetTeamsDirective>(*S));
+ CGM, ParentName, cast<OMPTargetTeamsDirective>(E));
break;
- case Stmt::OMPTargetTeamsDistributeDirectiveClass:
+ case OMPD_target_teams_distribute:
CodeGenFunction::EmitOMPTargetTeamsDistributeDeviceFunction(
- CGM, ParentName, cast<OMPTargetTeamsDistributeDirective>(*S));
+ CGM, ParentName, cast<OMPTargetTeamsDistributeDirective>(E));
break;
- case Stmt::OMPTargetTeamsDistributeSimdDirectiveClass:
+ case OMPD_target_teams_distribute_simd:
CodeGenFunction::EmitOMPTargetTeamsDistributeSimdDeviceFunction(
- CGM, ParentName, cast<OMPTargetTeamsDistributeSimdDirective>(*S));
+ CGM, ParentName, cast<OMPTargetTeamsDistributeSimdDirective>(E));
break;
- case Stmt::OMPTargetParallelForDirectiveClass:
+ case OMPD_target_parallel_for:
CodeGenFunction::EmitOMPTargetParallelForDeviceFunction(
- CGM, ParentName, cast<OMPTargetParallelForDirective>(*S));
+ CGM, ParentName, cast<OMPTargetParallelForDirective>(E));
break;
- case Stmt::OMPTargetParallelForSimdDirectiveClass:
+ case OMPD_target_parallel_for_simd:
CodeGenFunction::EmitOMPTargetParallelForSimdDeviceFunction(
- CGM, ParentName, cast<OMPTargetParallelForSimdDirective>(*S));
+ CGM, ParentName, cast<OMPTargetParallelForSimdDirective>(E));
break;
- case Stmt::OMPTargetSimdDirectiveClass:
+ case OMPD_target_simd:
CodeGenFunction::EmitOMPTargetSimdDeviceFunction(
- CGM, ParentName, cast<OMPTargetSimdDirective>(*S));
+ CGM, ParentName, cast<OMPTargetSimdDirective>(E));
break;
- default:
+ case OMPD_target_teams_distribute_parallel_for:
+ CodeGenFunction::EmitOMPTargetTeamsDistributeParallelForDeviceFunction(
+ CGM, ParentName,
+ cast<OMPTargetTeamsDistributeParallelForDirective>(E));
+ break;
+ case OMPD_target_teams_distribute_parallel_for_simd:
+ CodeGenFunction::
+ EmitOMPTargetTeamsDistributeParallelForSimdDeviceFunction(
+ CGM, ParentName,
+ cast<OMPTargetTeamsDistributeParallelForSimdDirective>(E));
+ break;
+ case OMPD_parallel:
+ case OMPD_for:
+ case OMPD_parallel_for:
+ case OMPD_parallel_sections:
+ case OMPD_for_simd:
+ case OMPD_parallel_for_simd:
+ case OMPD_cancel:
+ case OMPD_cancellation_point:
+ case OMPD_ordered:
+ case OMPD_threadprivate:
+ case OMPD_task:
+ case OMPD_simd:
+ case OMPD_sections:
+ case OMPD_section:
+ case OMPD_single:
+ case OMPD_master:
+ case OMPD_critical:
+ case OMPD_taskyield:
+ case OMPD_barrier:
+ case OMPD_taskwait:
+ case OMPD_taskgroup:
+ case OMPD_atomic:
+ case OMPD_flush:
+ case OMPD_teams:
+ case OMPD_target_data:
+ case OMPD_target_exit_data:
+ case OMPD_target_enter_data:
+ case OMPD_distribute:
+ case OMPD_distribute_simd:
+ case OMPD_distribute_parallel_for:
+ case OMPD_distribute_parallel_for_simd:
+ case OMPD_teams_distribute:
+ case OMPD_teams_distribute_simd:
+ case OMPD_teams_distribute_parallel_for:
+ case OMPD_teams_distribute_parallel_for_simd:
+ case OMPD_target_update:
+ case OMPD_declare_simd:
+ case OMPD_declare_target:
+ case OMPD_end_declare_target:
+ case OMPD_declare_reduction:
+ case OMPD_taskloop:
+ case OMPD_taskloop_simd:
+ case OMPD_unknown:
llvm_unreachable("Unknown target directive for OpenMP device codegen.");
}
return;
}
- if (const OMPExecutableDirective *E = dyn_cast<OMPExecutableDirective>(S)) {
- if (!E->hasAssociatedStmt())
+ if (const auto *E = dyn_cast<OMPExecutableDirective>(S)) {
+ if (!E->hasAssociatedStmt() || !E->getAssociatedStmt())
return;
scanForTargetRegionsFunctions(
- cast<CapturedStmt>(E->getAssociatedStmt())->getCapturedStmt(),
- ParentName);
+ E->getInnermostCapturedStmt()->getCapturedStmt(), ParentName);
return;
}
// If this is a lambda function, look into its body.
- if (auto *L = dyn_cast<LambdaExpr>(S))
+ if (const auto *L = dyn_cast<LambdaExpr>(S))
S = L->getBody();
// Keep looking for target regions recursively.
- for (auto *II : S->children())
+ for (const Stmt *II : S->children())
scanForTargetRegionsFunctions(II, ParentName);
}
bool CGOpenMPRuntime::emitTargetFunctions(GlobalDecl GD) {
- auto &FD = *cast<FunctionDecl>(GD.getDecl());
+ const auto *FD = cast<FunctionDecl>(GD.getDecl());
// If emitting code for the host, we do not process FD here. Instead we do
// the normal code generation.
@@ -7322,12 +8061,11 @@ bool CGOpenMPRuntime::emitTargetFunctions(GlobalDecl GD) {
return false;
// Try to detect target regions in the function.
- scanForTargetRegionsFunctions(FD.getBody(), CGM.getMangledName(GD));
+ scanForTargetRegionsFunctions(FD->getBody(), CGM.getMangledName(GD));
- // We should not emit any function other that the ones created during the
- // scanning. Therefore, we signal that this function is completely dealt
- // with.
- return true;
+ // Do not to emit function if it is not marked as declare target.
+ return !isDeclareTargetDeclaration(FD) &&
+ AlreadyEmittedTargetFunctions.count(FD->getCanonicalDecl()) == 0;
}
bool CGOpenMPRuntime::emitTargetGlobalVariable(GlobalDecl GD) {
@@ -7338,33 +8076,101 @@ bool CGOpenMPRuntime::emitTargetGlobalVariable(GlobalDecl GD) {
// regions in it. We use the complete variant to produce the kernel name
// mangling.
QualType RDTy = cast<VarDecl>(GD.getDecl())->getType();
- if (auto *RD = RDTy->getBaseElementTypeUnsafe()->getAsCXXRecordDecl()) {
- for (auto *Ctor : RD->ctors()) {
+ if (const auto *RD = RDTy->getBaseElementTypeUnsafe()->getAsCXXRecordDecl()) {
+ for (const CXXConstructorDecl *Ctor : RD->ctors()) {
StringRef ParentName =
CGM.getMangledName(GlobalDecl(Ctor, Ctor_Complete));
scanForTargetRegionsFunctions(Ctor->getBody(), ParentName);
}
- auto *Dtor = RD->getDestructor();
- if (Dtor) {
+ if (const CXXDestructorDecl *Dtor = RD->getDestructor()) {
StringRef ParentName =
CGM.getMangledName(GlobalDecl(Dtor, Dtor_Complete));
scanForTargetRegionsFunctions(Dtor->getBody(), ParentName);
}
}
- // If we are in target mode, we do not emit any global (declare target is not
- // implemented yet). Therefore we signal that GD was processed in this case.
- return true;
+ // Do not to emit variable if it is not marked as declare target.
+ llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(cast<VarDecl>(GD.getDecl()));
+ return !Res || *Res == OMPDeclareTargetDeclAttr::MT_Link;
+}
+
+void CGOpenMPRuntime::registerTargetGlobalVariable(const VarDecl *VD,
+ llvm::Constant *Addr) {
+ if (llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
+ isDeclareTargetDeclaration(VD)) {
+ OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind Flags;
+ StringRef VarName;
+ CharUnits VarSize;
+ llvm::GlobalValue::LinkageTypes Linkage;
+ switch (*Res) {
+ case OMPDeclareTargetDeclAttr::MT_To:
+ Flags = OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryTo;
+ VarName = CGM.getMangledName(VD);
+ VarSize = CGM.getContext().getTypeSizeInChars(VD->getType());
+ Linkage = CGM.getLLVMLinkageVarDefinition(VD, /*IsConstant=*/false);
+ break;
+ case OMPDeclareTargetDeclAttr::MT_Link:
+ Flags = OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryLink;
+ if (CGM.getLangOpts().OpenMPIsDevice) {
+ VarName = Addr->getName();
+ Addr = nullptr;
+ } else {
+ VarName = getAddrOfDeclareTargetLink(VD).getName();
+ Addr =
+ cast<llvm::Constant>(getAddrOfDeclareTargetLink(VD).getPointer());
+ }
+ VarSize = CGM.getPointerSize();
+ Linkage = llvm::GlobalValue::WeakAnyLinkage;
+ break;
+ }
+ OffloadEntriesInfoManager.registerDeviceGlobalVarEntryInfo(
+ VarName, Addr, VarSize, Flags, Linkage);
+ }
}
bool CGOpenMPRuntime::emitTargetGlobal(GlobalDecl GD) {
- auto *VD = GD.getDecl();
- if (isa<FunctionDecl>(VD))
+ if (isa<FunctionDecl>(GD.getDecl()))
return emitTargetFunctions(GD);
return emitTargetGlobalVariable(GD);
}
+CGOpenMPRuntime::DisableAutoDeclareTargetRAII::DisableAutoDeclareTargetRAII(
+ CodeGenModule &CGM)
+ : CGM(CGM) {
+ if (CGM.getLangOpts().OpenMPIsDevice) {
+ SavedShouldMarkAsGlobal = CGM.getOpenMPRuntime().ShouldMarkAsGlobal;
+ CGM.getOpenMPRuntime().ShouldMarkAsGlobal = false;
+ }
+}
+
+CGOpenMPRuntime::DisableAutoDeclareTargetRAII::~DisableAutoDeclareTargetRAII() {
+ if (CGM.getLangOpts().OpenMPIsDevice)
+ CGM.getOpenMPRuntime().ShouldMarkAsGlobal = SavedShouldMarkAsGlobal;
+}
+
+bool CGOpenMPRuntime::markAsGlobalTarget(GlobalDecl GD) {
+ if (!CGM.getLangOpts().OpenMPIsDevice || !ShouldMarkAsGlobal)
+ return true;
+
+ const auto *D = cast<FunctionDecl>(GD.getDecl());
+ const FunctionDecl *FD = D->getCanonicalDecl();
+ // Do not to emit function if it is marked as declare target as it was already
+ // emitted.
+ if (isDeclareTargetDeclaration(D)) {
+ if (D->hasBody() && AlreadyEmittedTargetFunctions.count(FD) == 0) {
+ if (auto *F = dyn_cast_or_null<llvm::Function>(
+ CGM.GetGlobalValue(CGM.getMangledName(GD))))
+ return !F->isDeclaration();
+ return false;
+ }
+ return true;
+ }
+
+ return !AlreadyEmittedTargetFunctions.insert(FD).second;
+}
+
llvm::Function *CGOpenMPRuntime::emitRegistrationFunction() {
// If we have offloading in the current module, we need to emit the entries
// now and register the offloading descriptor.
@@ -7384,7 +8190,7 @@ void CGOpenMPRuntime::emitTeamsCall(CodeGenFunction &CGF,
if (!CGF.HaveInsertPoint())
return;
- auto *RTLoc = emitUpdateLocation(CGF, Loc);
+ llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);
CodeGenFunction::RunCleanupsScope Scope(CGF);
// Build call __kmpc_fork_teams(loc, n, microtask, var1, .., varn);
@@ -7396,7 +8202,7 @@ void CGOpenMPRuntime::emitTeamsCall(CodeGenFunction &CGF,
RealArgs.append(std::begin(Args), std::end(Args));
RealArgs.append(CapturedVars.begin(), CapturedVars.end());
- auto RTLFn = createRuntimeFunction(OMPRTL__kmpc_fork_teams);
+ llvm::Value *RTLFn = createRuntimeFunction(OMPRTL__kmpc_fork_teams);
CGF.EmitRuntimeCall(RTLFn, RealArgs);
}
@@ -7407,16 +8213,16 @@ void CGOpenMPRuntime::emitNumTeamsClause(CodeGenFunction &CGF,
if (!CGF.HaveInsertPoint())
return;
- auto *RTLoc = emitUpdateLocation(CGF, Loc);
+ llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);
llvm::Value *NumTeamsVal =
- (NumTeams)
+ NumTeams
? CGF.Builder.CreateIntCast(CGF.EmitScalarExpr(NumTeams),
CGF.CGM.Int32Ty, /* isSigned = */ true)
: CGF.Builder.getInt32(0);
llvm::Value *ThreadLimitVal =
- (ThreadLimit)
+ ThreadLimit
? CGF.Builder.CreateIntCast(CGF.EmitScalarExpr(ThreadLimit),
CGF.CGM.Int32Ty, /* isSigned = */ true)
: CGF.Builder.getInt32(0);
@@ -7473,7 +8279,7 @@ void CGOpenMPRuntime::emitTargetDataCalls(
}
// Emit the number of elements in the offloading arrays.
- auto *PointerNum = CGF.Builder.getInt32(Info.NumberOfPtrs);
+ llvm::Value *PointerNum = CGF.Builder.getInt32(Info.NumberOfPtrs);
llvm::Value *OffloadingArgs[] = {
DeviceID, PointerNum, BasePointersArrayArg,
@@ -7509,7 +8315,7 @@ void CGOpenMPRuntime::emitTargetDataCalls(
}
// Emit the number of elements in the offloading arrays.
- auto *PointerNum = CGF.Builder.getInt32(Info.NumberOfPtrs);
+ llvm::Value *PointerNum = CGF.Builder.getInt32(Info.NumberOfPtrs);
llvm::Value *OffloadingArgs[] = {
DeviceID, PointerNum, BasePointersArrayArg,
@@ -7596,9 +8402,6 @@ void CGOpenMPRuntime::emitTargetDataStandAloneCall(
const bool HasNowait = D.hasClausesOfKind<OMPNowaitClause>();
OpenMPRTLFunction RTLFn;
switch (D.getDirectiveKind()) {
- default:
- llvm_unreachable("Unexpected standalone target data directive.");
- break;
case OMPD_target_enter_data:
RTLFn = HasNowait ? OMPRTL__tgt_target_data_begin_nowait
: OMPRTL__tgt_target_data_begin;
@@ -7611,6 +8414,58 @@ void CGOpenMPRuntime::emitTargetDataStandAloneCall(
RTLFn = HasNowait ? OMPRTL__tgt_target_data_update_nowait
: OMPRTL__tgt_target_data_update;
break;
+ case OMPD_parallel:
+ case OMPD_for:
+ case OMPD_parallel_for:
+ case OMPD_parallel_sections:
+ case OMPD_for_simd:
+ case OMPD_parallel_for_simd:
+ case OMPD_cancel:
+ case OMPD_cancellation_point:
+ case OMPD_ordered:
+ case OMPD_threadprivate:
+ case OMPD_task:
+ case OMPD_simd:
+ case OMPD_sections:
+ case OMPD_section:
+ case OMPD_single:
+ case OMPD_master:
+ case OMPD_critical:
+ case OMPD_taskyield:
+ case OMPD_barrier:
+ case OMPD_taskwait:
+ case OMPD_taskgroup:
+ case OMPD_atomic:
+ case OMPD_flush:
+ case OMPD_teams:
+ case OMPD_target_data:
+ case OMPD_distribute:
+ case OMPD_distribute_simd:
+ case OMPD_distribute_parallel_for:
+ case OMPD_distribute_parallel_for_simd:
+ case OMPD_teams_distribute:
+ case OMPD_teams_distribute_simd:
+ case OMPD_teams_distribute_parallel_for:
+ case OMPD_teams_distribute_parallel_for_simd:
+ case OMPD_declare_simd:
+ case OMPD_declare_target:
+ case OMPD_end_declare_target:
+ case OMPD_declare_reduction:
+ case OMPD_taskloop:
+ case OMPD_taskloop_simd:
+ case OMPD_target:
+ case OMPD_target_simd:
+ case OMPD_target_teams_distribute:
+ case OMPD_target_teams_distribute_simd:
+ case OMPD_target_teams_distribute_parallel_for:
+ case OMPD_target_teams_distribute_parallel_for_simd:
+ case OMPD_target_teams:
+ case OMPD_target_parallel:
+ case OMPD_target_parallel_for:
+ case OMPD_target_parallel_for_simd:
+ case OMPD_unknown:
+ llvm_unreachable("Unexpected standalone target data directive.");
+ break;
}
CGF.EmitRuntimeCall(createRuntimeFunction(RTLFn), OffloadingArgs);
};
@@ -7644,13 +8499,13 @@ void CGOpenMPRuntime::emitTargetDataStandAloneCall(
if (D.hasClausesOfKind<OMPDependClause>())
CGF.EmitOMPTargetTaskBasedDirective(D, ThenGen, InputInfo);
else
- emitInlinedDirective(CGF, OMPD_target_update, ThenGen);
+ emitInlinedDirective(CGF, D.getDirectiveKind(), ThenGen);
};
- if (IfCond)
+ if (IfCond) {
emitOMPIfClause(CGF, IfCond, TargetThenGen,
[](CodeGenFunction &CGF, PrePostActionTy &) {});
- else {
+ } else {
RegionCodeGenTy ThenRCG(TargetThenGen);
ThenRCG(CGF);
}
@@ -7693,11 +8548,11 @@ static unsigned evaluateCDTSize(const FunctionDecl *FD,
return 0;
ASTContext &C = FD->getASTContext();
QualType CDT;
- if (!RetType.isNull() && !RetType->isVoidType())
+ if (!RetType.isNull() && !RetType->isVoidType()) {
CDT = RetType;
- else {
+ } else {
unsigned Offset = 0;
- if (auto *MD = dyn_cast<CXXMethodDecl>(FD)) {
+ if (const auto *MD = dyn_cast<CXXMethodDecl>(FD)) {
if (ParamAttrs[Offset].Kind == Vector)
CDT = C.getPointerType(C.getRecordType(MD->getParent()));
++Offset;
@@ -7755,17 +8610,18 @@ emitX86DeclareSimdFunction(const FunctionDecl *FD, llvm::Function *Fn,
Masked.push_back('M');
break;
}
- for (auto Mask : Masked) {
- for (auto &Data : ISAData) {
+ for (char Mask : Masked) {
+ for (const ISADataTy &Data : ISAData) {
SmallString<256> Buffer;
llvm::raw_svector_ostream Out(Buffer);
Out << "_ZGV" << Data.ISA << Mask;
if (!VLENVal) {
Out << llvm::APSInt::getUnsigned(Data.VecRegSize /
evaluateCDTSize(FD, ParamAttrs));
- } else
+ } else {
Out << VLENVal;
- for (auto &ParamAttr : ParamAttrs) {
+ }
+ for (const ParamAttrTy &ParamAttr : ParamAttrs) {
switch (ParamAttr.Kind){
case LinearWithVarStride:
Out << 's' << ParamAttr.StrideOrArg;
@@ -7794,90 +8650,95 @@ emitX86DeclareSimdFunction(const FunctionDecl *FD, llvm::Function *Fn,
void CGOpenMPRuntime::emitDeclareSimdFunction(const FunctionDecl *FD,
llvm::Function *Fn) {
ASTContext &C = CGM.getContext();
- FD = FD->getCanonicalDecl();
+ FD = FD->getMostRecentDecl();
// Map params to their positions in function decl.
llvm::DenseMap<const Decl *, unsigned> ParamPositions;
if (isa<CXXMethodDecl>(FD))
- ParamPositions.insert({FD, 0});
+ ParamPositions.try_emplace(FD, 0);
unsigned ParamPos = ParamPositions.size();
- for (auto *P : FD->parameters()) {
- ParamPositions.insert({P->getCanonicalDecl(), ParamPos});
+ for (const ParmVarDecl *P : FD->parameters()) {
+ ParamPositions.try_emplace(P->getCanonicalDecl(), ParamPos);
++ParamPos;
}
- for (auto *Attr : FD->specific_attrs<OMPDeclareSimdDeclAttr>()) {
- llvm::SmallVector<ParamAttrTy, 8> ParamAttrs(ParamPositions.size());
- // Mark uniform parameters.
- for (auto *E : Attr->uniforms()) {
- E = E->IgnoreParenImpCasts();
- unsigned Pos;
- if (isa<CXXThisExpr>(E))
- Pos = ParamPositions[FD];
- else {
- auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
- ->getCanonicalDecl();
- Pos = ParamPositions[PVD];
- }
- ParamAttrs[Pos].Kind = Uniform;
- }
- // Get alignment info.
- auto NI = Attr->alignments_begin();
- for (auto *E : Attr->aligneds()) {
- E = E->IgnoreParenImpCasts();
- unsigned Pos;
- QualType ParmTy;
- if (isa<CXXThisExpr>(E)) {
- Pos = ParamPositions[FD];
- ParmTy = E->getType();
- } else {
- auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
- ->getCanonicalDecl();
- Pos = ParamPositions[PVD];
- ParmTy = PVD->getType();
+ while (FD) {
+ for (const auto *Attr : FD->specific_attrs<OMPDeclareSimdDeclAttr>()) {
+ llvm::SmallVector<ParamAttrTy, 8> ParamAttrs(ParamPositions.size());
+ // Mark uniform parameters.
+ for (const Expr *E : Attr->uniforms()) {
+ E = E->IgnoreParenImpCasts();
+ unsigned Pos;
+ if (isa<CXXThisExpr>(E)) {
+ Pos = ParamPositions[FD];
+ } else {
+ const auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
+ ->getCanonicalDecl();
+ Pos = ParamPositions[PVD];
+ }
+ ParamAttrs[Pos].Kind = Uniform;
}
- ParamAttrs[Pos].Alignment =
- (*NI) ? (*NI)->EvaluateKnownConstInt(C)
+ // Get alignment info.
+ auto NI = Attr->alignments_begin();
+ for (const Expr *E : Attr->aligneds()) {
+ E = E->IgnoreParenImpCasts();
+ unsigned Pos;
+ QualType ParmTy;
+ if (isa<CXXThisExpr>(E)) {
+ Pos = ParamPositions[FD];
+ ParmTy = E->getType();
+ } else {
+ const auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
+ ->getCanonicalDecl();
+ Pos = ParamPositions[PVD];
+ ParmTy = PVD->getType();
+ }
+ ParamAttrs[Pos].Alignment =
+ (*NI)
+ ? (*NI)->EvaluateKnownConstInt(C)
: llvm::APSInt::getUnsigned(
C.toCharUnitsFromBits(C.getOpenMPDefaultSimdAlign(ParmTy))
.getQuantity());
- ++NI;
- }
- // Mark linear parameters.
- auto SI = Attr->steps_begin();
- auto MI = Attr->modifiers_begin();
- for (auto *E : Attr->linears()) {
- E = E->IgnoreParenImpCasts();
- unsigned Pos;
- if (isa<CXXThisExpr>(E))
- Pos = ParamPositions[FD];
- else {
- auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
- ->getCanonicalDecl();
- Pos = ParamPositions[PVD];
+ ++NI;
}
- auto &ParamAttr = ParamAttrs[Pos];
- ParamAttr.Kind = Linear;
- if (*SI) {
- if (!(*SI)->EvaluateAsInt(ParamAttr.StrideOrArg, C,
- Expr::SE_AllowSideEffects)) {
- if (auto *DRE = cast<DeclRefExpr>((*SI)->IgnoreParenImpCasts())) {
- if (auto *StridePVD = cast<ParmVarDecl>(DRE->getDecl())) {
- ParamAttr.Kind = LinearWithVarStride;
- ParamAttr.StrideOrArg = llvm::APSInt::getUnsigned(
- ParamPositions[StridePVD->getCanonicalDecl()]);
+ // Mark linear parameters.
+ auto SI = Attr->steps_begin();
+ auto MI = Attr->modifiers_begin();
+ for (const Expr *E : Attr->linears()) {
+ E = E->IgnoreParenImpCasts();
+ unsigned Pos;
+ if (isa<CXXThisExpr>(E)) {
+ Pos = ParamPositions[FD];
+ } else {
+ const auto *PVD = cast<ParmVarDecl>(cast<DeclRefExpr>(E)->getDecl())
+ ->getCanonicalDecl();
+ Pos = ParamPositions[PVD];
+ }
+ ParamAttrTy &ParamAttr = ParamAttrs[Pos];
+ ParamAttr.Kind = Linear;
+ if (*SI) {
+ if (!(*SI)->EvaluateAsInt(ParamAttr.StrideOrArg, C,
+ Expr::SE_AllowSideEffects)) {
+ if (const auto *DRE =
+ cast<DeclRefExpr>((*SI)->IgnoreParenImpCasts())) {
+ if (const auto *StridePVD = cast<ParmVarDecl>(DRE->getDecl())) {
+ ParamAttr.Kind = LinearWithVarStride;
+ ParamAttr.StrideOrArg = llvm::APSInt::getUnsigned(
+ ParamPositions[StridePVD->getCanonicalDecl()]);
+ }
}
}
}
+ ++SI;
+ ++MI;
}
- ++SI;
- ++MI;
+ llvm::APSInt VLENVal;
+ if (const Expr *VLEN = Attr->getSimdlen())
+ VLENVal = VLEN->EvaluateKnownConstInt(C);
+ OMPDeclareSimdDeclAttr::BranchStateTy State = Attr->getBranchState();
+ if (CGM.getTriple().getArch() == llvm::Triple::x86 ||
+ CGM.getTriple().getArch() == llvm::Triple::x86_64)
+ emitX86DeclareSimdFunction(FD, Fn, VLENVal, ParamAttrs, State);
}
- llvm::APSInt VLENVal;
- if (const Expr *VLEN = Attr->getSimdlen())
- VLENVal = VLEN->EvaluateKnownConstInt(C);
- OMPDeclareSimdDeclAttr::BranchStateTy State = Attr->getBranchState();
- if (CGM.getTriple().getArch() == llvm::Triple::x86 ||
- CGM.getTriple().getArch() == llvm::Triple::x86_64)
- emitX86DeclareSimdFunction(FD, Fn, VLENVal, ParamAttrs, State);
+ FD = FD->getPreviousDecl();
}
}
@@ -7926,8 +8787,9 @@ void CGOpenMPRuntime::emitDoacrossInit(CodeGenFunction &CGF,
addFieldToRecordDecl(C, RD, Int64Ty);
RD->completeDefinition();
KmpDimTy = C.getRecordType(RD);
- } else
+ } else {
RD = cast<RecordDecl>(KmpDimTy->getAsTagDecl());
+ }
Address DimsAddr = CGF.CreateMemTemp(KmpDimTy, "dims");
CGF.EmitNullInitialization(DimsAddr, KmpDimTy);
@@ -7979,18 +8841,19 @@ void CGOpenMPRuntime::emitDoacrossOrdered(CodeGenFunction &CGF,
getThreadID(CGF, C->getLocStart()),
CntAddr.getPointer()};
llvm::Value *RTLFn;
- if (C->getDependencyKind() == OMPC_DEPEND_source)
+ if (C->getDependencyKind() == OMPC_DEPEND_source) {
RTLFn = createRuntimeFunction(OMPRTL__kmpc_doacross_post);
- else {
+ } else {
assert(C->getDependencyKind() == OMPC_DEPEND_sink);
RTLFn = createRuntimeFunction(OMPRTL__kmpc_doacross_wait);
}
CGF.EmitRuntimeCall(RTLFn, Args);
}
-void CGOpenMPRuntime::emitCall(CodeGenFunction &CGF, llvm::Value *Callee,
- ArrayRef<llvm::Value *> Args,
- SourceLocation Loc) const {
+void CGOpenMPRuntime::emitCall(CodeGenFunction &CGF, SourceLocation Loc,
+ llvm::Value *Callee,
+ ArrayRef<llvm::Value *> Args) const {
+ assert(Loc.isValid() && "Outlined function call location must be valid.");
auto DL = ApplyDebugLocation::CreateDefaultArtificial(CGF, Loc);
if (auto *Fn = dyn_cast<llvm::Function>(Callee)) {
@@ -8005,8 +8868,7 @@ void CGOpenMPRuntime::emitCall(CodeGenFunction &CGF, llvm::Value *Callee,
void CGOpenMPRuntime::emitOutlinedFunctionCall(
CodeGenFunction &CGF, SourceLocation Loc, llvm::Value *OutlinedFn,
ArrayRef<llvm::Value *> Args) const {
- assert(Loc.isValid() && "Outlined function call location must be valid.");
- emitCall(CGF, OutlinedFn, Args, Loc);
+ emitCall(CGF, Loc, OutlinedFn, Args);
}
Address CGOpenMPRuntime::getParameterAddress(CodeGenFunction &CGF,
@@ -8014,3 +8876,303 @@ Address CGOpenMPRuntime::getParameterAddress(CodeGenFunction &CGF,
const VarDecl *TargetParam) const {
return CGF.GetAddrOfLocalVar(NativeParam);
}
+
+Address CGOpenMPRuntime::getAddressOfLocalVariable(CodeGenFunction &CGF,
+ const VarDecl *VD) {
+ return Address::invalid();
+}
+
+llvm::Value *CGOpenMPSIMDRuntime::emitParallelOutlinedFunction(
+ const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
+ OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+llvm::Value *CGOpenMPSIMDRuntime::emitTeamsOutlinedFunction(
+ const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
+ OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+llvm::Value *CGOpenMPSIMDRuntime::emitTaskOutlinedFunction(
+ const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
+ const VarDecl *PartIDVar, const VarDecl *TaskTVar,
+ OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen,
+ bool Tied, unsigned &NumberOfParts) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitParallelCall(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ llvm::Value *OutlinedFn,
+ ArrayRef<llvm::Value *> CapturedVars,
+ const Expr *IfCond) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitCriticalRegion(
+ CodeGenFunction &CGF, StringRef CriticalName,
+ const RegionCodeGenTy &CriticalOpGen, SourceLocation Loc,
+ const Expr *Hint) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitMasterRegion(CodeGenFunction &CGF,
+ const RegionCodeGenTy &MasterOpGen,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskyieldCall(CodeGenFunction &CGF,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskgroupRegion(
+ CodeGenFunction &CGF, const RegionCodeGenTy &TaskgroupOpGen,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitSingleRegion(
+ CodeGenFunction &CGF, const RegionCodeGenTy &SingleOpGen,
+ SourceLocation Loc, ArrayRef<const Expr *> CopyprivateVars,
+ ArrayRef<const Expr *> DestExprs, ArrayRef<const Expr *> SrcExprs,
+ ArrayRef<const Expr *> AssignmentOps) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitOrderedRegion(CodeGenFunction &CGF,
+ const RegionCodeGenTy &OrderedOpGen,
+ SourceLocation Loc,
+ bool IsThreads) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitBarrierCall(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ OpenMPDirectiveKind Kind,
+ bool EmitChecks,
+ bool ForceSimpleCall) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitForDispatchInit(
+ CodeGenFunction &CGF, SourceLocation Loc,
+ const OpenMPScheduleTy &ScheduleKind, unsigned IVSize, bool IVSigned,
+ bool Ordered, const DispatchRTInput &DispatchValues) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitForStaticInit(
+ CodeGenFunction &CGF, SourceLocation Loc, OpenMPDirectiveKind DKind,
+ const OpenMPScheduleTy &ScheduleKind, const StaticRTInput &Values) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitDistributeStaticInit(
+ CodeGenFunction &CGF, SourceLocation Loc,
+ OpenMPDistScheduleClauseKind SchedKind, const StaticRTInput &Values) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitForOrderedIterationEnd(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ unsigned IVSize,
+ bool IVSigned) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitForStaticFinish(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ OpenMPDirectiveKind DKind) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+llvm::Value *CGOpenMPSIMDRuntime::emitForNext(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ unsigned IVSize, bool IVSigned,
+ Address IL, Address LB,
+ Address UB, Address ST) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitNumThreadsClause(CodeGenFunction &CGF,
+ llvm::Value *NumThreads,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitProcBindClause(CodeGenFunction &CGF,
+ OpenMPProcBindClauseKind ProcBind,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+Address CGOpenMPSIMDRuntime::getAddrOfThreadPrivate(CodeGenFunction &CGF,
+ const VarDecl *VD,
+ Address VDAddr,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+llvm::Function *CGOpenMPSIMDRuntime::emitThreadPrivateVarDefinition(
+ const VarDecl *VD, Address VDAddr, SourceLocation Loc, bool PerformInit,
+ CodeGenFunction *CGF) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+Address CGOpenMPSIMDRuntime::getAddrOfArtificialThreadPrivate(
+ CodeGenFunction &CGF, QualType VarType, StringRef Name) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitFlush(CodeGenFunction &CGF,
+ ArrayRef<const Expr *> Vars,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
+ const OMPExecutableDirective &D,
+ llvm::Value *TaskFunction,
+ QualType SharedsTy, Address Shareds,
+ const Expr *IfCond,
+ const OMPTaskDataTy &Data) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskLoopCall(
+ CodeGenFunction &CGF, SourceLocation Loc, const OMPLoopDirective &D,
+ llvm::Value *TaskFunction, QualType SharedsTy, Address Shareds,
+ const Expr *IfCond, const OMPTaskDataTy &Data) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitReduction(
+ CodeGenFunction &CGF, SourceLocation Loc, ArrayRef<const Expr *> Privates,
+ ArrayRef<const Expr *> LHSExprs, ArrayRef<const Expr *> RHSExprs,
+ ArrayRef<const Expr *> ReductionOps, ReductionOptionsTy Options) {
+ assert(Options.SimpleReduction && "Only simple reduction is expected.");
+ CGOpenMPRuntime::emitReduction(CGF, Loc, Privates, LHSExprs, RHSExprs,
+ ReductionOps, Options);
+}
+
+llvm::Value *CGOpenMPSIMDRuntime::emitTaskReductionInit(
+ CodeGenFunction &CGF, SourceLocation Loc, ArrayRef<const Expr *> LHSExprs,
+ ArrayRef<const Expr *> RHSExprs, const OMPTaskDataTy &Data) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskReductionFixups(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ ReductionCodeGen &RCG,
+ unsigned N) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+Address CGOpenMPSIMDRuntime::getTaskReductionItem(CodeGenFunction &CGF,
+ SourceLocation Loc,
+ llvm::Value *ReductionsPtr,
+ LValue SharedLVal) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTaskwaitCall(CodeGenFunction &CGF,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitCancellationPointCall(
+ CodeGenFunction &CGF, SourceLocation Loc,
+ OpenMPDirectiveKind CancelRegion) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitCancelCall(CodeGenFunction &CGF,
+ SourceLocation Loc, const Expr *IfCond,
+ OpenMPDirectiveKind CancelRegion) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTargetOutlinedFunction(
+ const OMPExecutableDirective &D, StringRef ParentName,
+ llvm::Function *&OutlinedFn, llvm::Constant *&OutlinedFnID,
+ bool IsOffloadEntry, const RegionCodeGenTy &CodeGen) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTargetCall(CodeGenFunction &CGF,
+ const OMPExecutableDirective &D,
+ llvm::Value *OutlinedFn,
+ llvm::Value *OutlinedFnID,
+ const Expr *IfCond, const Expr *Device) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+bool CGOpenMPSIMDRuntime::emitTargetFunctions(GlobalDecl GD) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+bool CGOpenMPSIMDRuntime::emitTargetGlobalVariable(GlobalDecl GD) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+bool CGOpenMPSIMDRuntime::emitTargetGlobal(GlobalDecl GD) {
+ return false;
+}
+
+llvm::Function *CGOpenMPSIMDRuntime::emitRegistrationFunction() {
+ return nullptr;
+}
+
+void CGOpenMPSIMDRuntime::emitTeamsCall(CodeGenFunction &CGF,
+ const OMPExecutableDirective &D,
+ SourceLocation Loc,
+ llvm::Value *OutlinedFn,
+ ArrayRef<llvm::Value *> CapturedVars) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitNumTeamsClause(CodeGenFunction &CGF,
+ const Expr *NumTeams,
+ const Expr *ThreadLimit,
+ SourceLocation Loc) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTargetDataCalls(
+ CodeGenFunction &CGF, const OMPExecutableDirective &D, const Expr *IfCond,
+ const Expr *Device, const RegionCodeGenTy &CodeGen, TargetDataInfo &Info) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitTargetDataStandAloneCall(
+ CodeGenFunction &CGF, const OMPExecutableDirective &D, const Expr *IfCond,
+ const Expr *Device) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitDoacrossInit(CodeGenFunction &CGF,
+ const OMPLoopDirective &D) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+void CGOpenMPSIMDRuntime::emitDoacrossOrdered(CodeGenFunction &CGF,
+ const OMPDependClause *C) {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+const VarDecl *
+CGOpenMPSIMDRuntime::translateParameter(const FieldDecl *FD,
+ const VarDecl *NativeParam) const {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
+Address
+CGOpenMPSIMDRuntime::getParameterAddress(CodeGenFunction &CGF,
+ const VarDecl *NativeParam,
+ const VarDecl *TargetParam) const {
+ llvm_unreachable("Not supported in SIMD-only mode");
+}
+
generated by cgit v1.2.3 (git 2.25.1) at 2025-11-02 17:09:23 +0000