/usr/src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/clang/lib/CodeGen/CGOpenMPRuntime.cpp

Bug Summary

File:	src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/clang/lib/CodeGen/CGOpenMPRuntime.cpp
Warning:	line 9199, column 30 Called C++ object pointer is null
Annotated Source Code

Press '?' to see keyboard shortcuts
Show analyzer invocation
clang -cc1 -cc1 -triple amd64-unknown-openbsd7.0 -analyze -disable-free -disable-llvm-verifier -discard-value-names -main-file-name CGOpenMPRuntime.cpp -analyzer-store=region -analyzer-opt-analyze-nested-blocks -analyzer-checker=core -analyzer-checker=apiModeling -analyzer-checker=unix -analyzer-checker=deadcode -analyzer-checker=cplusplus -analyzer-checker=security.insecureAPI.UncheckedReturn -analyzer-checker=security.insecureAPI.getpw -analyzer-checker=security.insecureAPI.gets -analyzer-checker=security.insecureAPI.mktemp -analyzer-checker=security.insecureAPI.mkstemp -analyzer-checker=security.insecureAPI.vfork -analyzer-checker=nullability.NullPassedToNonnull -analyzer-checker=nullability.NullReturnedFromNonnull -analyzer-output plist -w -setup-static-analyzer -mrelocation-model static -mframe-pointer=all -relaxed-aliasing -fno-rounding-math -mconstructor-aliases -munwind-tables -target-cpu x86-64 -tune-cpu generic -debugger-tuning=gdb -fcoverage-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangCodeGen/obj -resource-dir /usr/local/lib/clang/13.0.0 -I /usr/src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/clang/include -I /usr/src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/llvm/include -I /usr/src/gnu/usr.bin/clang/libclangCodeGen/../include -I /usr/src/gnu/usr.bin/clang/libclangCodeGen/obj -I /usr/src/gnu/usr.bin/clang/libclangCodeGen/obj/../include -D NDEBUG -D __STDC_LIMIT_MACROS -D __STDC_CONSTANT_MACROS -D __STDC_FORMAT_MACROS -D LLVM_PREFIX="/usr" -internal-isystem /usr/include/c++/v1 -internal-isystem /usr/local/lib/clang/13.0.0/include -internal-externc-isystem /usr/include -O2 -Wno-unused-parameter -Wwrite-strings -Wno-missing-field-initializers -Wno-long-long -Wno-comment -std=c++14 -fdeprecated-macro -fdebug-compilation-dir=/usr/src/gnu/usr.bin/clang/libclangCodeGen/obj -ferror-limit 19 -fvisibility-inlines-hidden -fwrapv -stack-protector 2 -fno-rtti -fgnuc-version=4.2.1 -vectorize-loops -vectorize-slp -fno-builtin-malloc -fno-builtin-calloc -fno-builtin-realloc -fno-builtin-valloc -fno-builtin-free -fno-builtin-strdup -fno-builtin-strndup -analyzer-output=html -faddrsig -D__GCC_HAVE_DWARF2_CFI_ASM=1 -o /home/ben/Projects/vmm/scan-build/2022-01-12-194120-40624-1 -x c++ /usr/src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/clang/lib/CodeGen/CGOpenMPRuntime.cpp
/usr/src/gnu/usr.bin/clang/libclangCodeGen/../../../llvm/clang/lib/CodeGen/CGOpenMPRuntime.cpp

→
1//===----- CGOpenMPRuntime.cpp - Interface to OpenMP Runtimes -------------===//
2//
3// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
4// See https://llvm.org/LICENSE.txt for license information.
5// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
6//
7//===----------------------------------------------------------------------===//
8//
9// This provides a class for OpenMP runtime code generation.
10//
11//===----------------------------------------------------------------------===//

13#include "CGOpenMPRuntime.h"
14#include "CGCXXABI.h"
15#include "CGCleanup.h"
16#include "CGRecordLayout.h"
17#include "CodeGenFunction.h"
18#include "clang/AST/APValue.h"
19#include "clang/AST/Attr.h"
20#include "clang/AST/Decl.h"
21#include "clang/AST/OpenMPClause.h"
22#include "clang/AST/StmtOpenMP.h"
23#include "clang/AST/StmtVisitor.h"
24#include "clang/Basic/BitmaskEnum.h"
25#include "clang/Basic/FileManager.h"
26#include "clang/Basic/OpenMPKinds.h"
27#include "clang/Basic/SourceManager.h"
28#include "clang/CodeGen/ConstantInitBuilder.h"
29#include "llvm/ADT/ArrayRef.h"
30#include "llvm/ADT/SetOperations.h"
31#include "llvm/ADT/StringExtras.h"
32#include "llvm/Bitcode/BitcodeReader.h"
33#include "llvm/IR/Constants.h"
34#include "llvm/IR/DerivedTypes.h"
35#include "llvm/IR/GlobalValue.h"
36#include "llvm/IR/Value.h"
37#include "llvm/Support/AtomicOrdering.h"
38#include "llvm/Support/Format.h"
39#include "llvm/Support/raw_ostream.h"
40#include <cassert>
41#include <numeric>

43using namespace clang;
44using namespace CodeGen;
45using namespace llvm::omp;

47namespace {
48/// Base class for handling code generation inside OpenMP regions.
49class CGOpenMPRegionInfo : public CodeGenFunction::CGCapturedStmtInfo {
50public:
/// Kinds of OpenMP regions used in codegen.
enum CGOpenMPRegionKind {
  /// Region with outlined function for standalone 'parallel'
  /// directive.
  ParallelOutlinedRegion,
  /// Region with outlined function for standalone 'task' directive.
  TaskOutlinedRegion,
  /// Region for constructs that do not require function outlining,
  /// like 'for', 'sections', 'atomic' etc. directives.
  InlinedRegion,
  /// Region with outlined function for standalone 'target' directive.
  TargetRegion,
};

CGOpenMPRegionInfo(const CapturedStmt &CS,
                   const CGOpenMPRegionKind RegionKind,
                   const RegionCodeGenTy &CodeGen, OpenMPDirectiveKind Kind,
                   bool HasCancel)
    : CGCapturedStmtInfo(CS, CR_OpenMP), RegionKind(RegionKind),
      CodeGen(CodeGen), Kind(Kind), HasCancel(HasCancel) {}

CGOpenMPRegionInfo(const CGOpenMPRegionKind RegionKind,
                   const RegionCodeGenTy &CodeGen, OpenMPDirectiveKind Kind,
                   bool HasCancel)
    : CGCapturedStmtInfo(CR_OpenMP), RegionKind(RegionKind), CodeGen(CodeGen),
      Kind(Kind), HasCancel(HasCancel) {}

/// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
virtual const VarDecl *getThreadIDVariable() const = 0;

/// Emit the captured statement body.
void EmitBody(CodeGenFunction &CGF, const Stmt *S) override;

/// 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);

virtual void emitUntiedSwitch(CodeGenFunction & /*CGF*/) {}

CGOpenMPRegionKind getRegionKind() const { return RegionKind; }

OpenMPDirectiveKind getDirectiveKind() const { return Kind; }

bool hasCancel() const { return HasCancel; }

static bool classof(const CGCapturedStmtInfo *Info) {
  return Info->getKind() == CR_OpenMP;
}

~CGOpenMPRegionInfo() override = default;

103protected:
CGOpenMPRegionKind RegionKind;
RegionCodeGenTy CodeGen;
OpenMPDirectiveKind Kind;
bool HasCancel;
108};

110/// API for captured statement code generation in OpenMP constructs.
111class CGOpenMPOutlinedRegionInfo final : public CGOpenMPRegionInfo {
112public:
CGOpenMPOutlinedRegionInfo(const CapturedStmt &CS, const VarDecl *ThreadIDVar,
                           const RegionCodeGenTy &CodeGen,
                           OpenMPDirectiveKind Kind, bool HasCancel,
                           StringRef HelperName)
    : CGOpenMPRegionInfo(CS, ParallelOutlinedRegion, CodeGen, Kind,
                         HasCancel),
      ThreadIDVar(ThreadIDVar), HelperName(HelperName) {
  assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.")((void)0);
}

/// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }

/// Get the name of the capture helper.
StringRef getHelperName() const override { return HelperName; }

static bool classof(const CGCapturedStmtInfo *Info) {
  return CGOpenMPRegionInfo::classof(Info) &&
         cast<CGOpenMPRegionInfo>(Info)->getRegionKind() ==
             ParallelOutlinedRegion;
}

136private:
/// A variable or parameter storing global thread id for OpenMP
/// constructs.
const VarDecl *ThreadIDVar;
StringRef HelperName;
141};

143/// API for captured statement code generation in OpenMP constructs.
144class CGOpenMPTaskOutlinedRegionInfo final : public CGOpenMPRegionInfo {
145public:
class UntiedTaskActionTy final : public PrePostActionTy {
  bool Untied;
  const VarDecl *PartIDVar;
  const RegionCodeGenTy UntiedCodeGen;
  llvm::SwitchInst *UntiedSwitch = nullptr;

public:
  UntiedTaskActionTy(bool Tied, const VarDecl *PartIDVar,
                     const RegionCodeGenTy &UntiedCodeGen)
      : Untied(!Tied), PartIDVar(PartIDVar), UntiedCodeGen(UntiedCodeGen) {}
  void Enter(CodeGenFunction &CGF) override {
    if (Untied) {
      // Emit task switching point.
      LValue PartIdLVal = CGF.EmitLoadOfPointerLValue(
          CGF.GetAddrOfLocalVar(PartIDVar),
          PartIDVar->getType()->castAs<PointerType>());
      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);
      CGF.EmitBlock(CGF.createBasicBlock(".untied.jmp."));
      UntiedSwitch->addCase(CGF.Builder.getInt32(0),
                            CGF.Builder.GetInsertBlock());
      emitUntiedSwitch(CGF);
    }
  }
  void emitUntiedSwitch(CodeGenFunction &CGF) const {
    if (Untied) {
      LValue PartIdLVal = CGF.EmitLoadOfPointerLValue(
          CGF.GetAddrOfLocalVar(PartIDVar),
          PartIDVar->getType()->castAs<PointerType>());
      CGF.EmitStoreOfScalar(CGF.Builder.getInt32(UntiedSwitch->getNumCases()),
                            PartIdLVal);
      UntiedCodeGen(CGF);
      CodeGenFunction::JumpDest CurPoint =
          CGF.getJumpDestInCurrentScope(".untied.next.");
      CGF.EmitBranch(CGF.ReturnBlock.getBlock());
      CGF.EmitBlock(CGF.createBasicBlock(".untied.jmp."));
      UntiedSwitch->addCase(CGF.Builder.getInt32(UntiedSwitch->getNumCases()),
                            CGF.Builder.GetInsertBlock());
      CGF.EmitBranchThroughCleanup(CurPoint);
      CGF.EmitBlock(CurPoint.getBlock());
    }
  }
  unsigned getNumberOfParts() const { return UntiedSwitch->getNumCases(); }
};
CGOpenMPTaskOutlinedRegionInfo(const CapturedStmt &CS,
                               const VarDecl *ThreadIDVar,
                               const RegionCodeGenTy &CodeGen,
                               OpenMPDirectiveKind Kind, bool HasCancel,
                               const UntiedTaskActionTy &Action)
    : CGOpenMPRegionInfo(CS, TaskOutlinedRegion, CodeGen, Kind, HasCancel),
      ThreadIDVar(ThreadIDVar), Action(Action) {
  assert(ThreadIDVar != nullptr && "No ThreadID in OpenMP region.")((void)0);
}

/// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override { return ThreadIDVar; }

/// Get an LValue for the current ThreadID variable.
LValue getThreadIDVariableLValue(CodeGenFunction &CGF) override;

/// Get the name of the capture helper.
StringRef getHelperName() const override { return ".omp_outlined."; }

void emitUntiedSwitch(CodeGenFunction &CGF) override {
  Action.emitUntiedSwitch(CGF);
}

static bool classof(const CGCapturedStmtInfo *Info) {
  return CGOpenMPRegionInfo::classof(Info) &&
         cast<CGOpenMPRegionInfo>(Info)->getRegionKind() ==
             TaskOutlinedRegion;
}

224private:
/// A variable or parameter storing global thread id for OpenMP
/// constructs.
const VarDecl *ThreadIDVar;
/// Action for emitting code for untied tasks.
const UntiedTaskActionTy &Action;
230};

232/// API for inlined captured statement code generation in OpenMP
233/// constructs.
234class CGOpenMPInlinedRegionInfo : public CGOpenMPRegionInfo {
235public:
CGOpenMPInlinedRegionInfo(CodeGenFunction::CGCapturedStmtInfo *OldCSI,
                          const RegionCodeGenTy &CodeGen,
                          OpenMPDirectiveKind Kind, bool HasCancel)
    : CGOpenMPRegionInfo(InlinedRegion, CodeGen, Kind, HasCancel),
      OldCSI(OldCSI),
      OuterRegionInfo(dyn_cast_or_null<CGOpenMPRegionInfo>(OldCSI)) {}

// Retrieve the value of the context parameter.
llvm::Value *getContextValue() const override {
  if (OuterRegionInfo)
    return OuterRegionInfo->getContextValue();
  llvm_unreachable("No context value for inlined OpenMP region")__builtin_unreachable();
}

void setContextValue(llvm::Value *V) override {
  if (OuterRegionInfo) {
    OuterRegionInfo->setContextValue(V);
    return;
  }
  llvm_unreachable("No context value for inlined OpenMP region")__builtin_unreachable();
}

/// Lookup the captured field decl for a variable.
const FieldDecl *lookup(const VarDecl *VD) const override {
  if (OuterRegionInfo)
    return OuterRegionInfo->lookup(VD);
  // If there is no outer outlined region,no need to lookup in a list of
  // captured variables, we can use the original one.
  return nullptr;
}

FieldDecl *getThisFieldDecl() const override {
  if (OuterRegionInfo)
    return OuterRegionInfo->getThisFieldDecl();
  return nullptr;
}

/// Get a variable or parameter for storing global thread id
/// inside OpenMP construct.
const VarDecl *getThreadIDVariable() const override {
  if (OuterRegionInfo)
    return OuterRegionInfo->getThreadIDVariable();
  return nullptr;
}

/// 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")__builtin_unreachable();
}

/// Get the name of the capture helper.
StringRef getHelperName() const override {
  if (auto *OuterRegionInfo = getOldCSI())
    return OuterRegionInfo->getHelperName();
  llvm_unreachable("No helper name for inlined OpenMP construct")__builtin_unreachable();
}

void emitUntiedSwitch(CodeGenFunction &CGF) override {
  if (OuterRegionInfo)
    OuterRegionInfo->emitUntiedSwitch(CGF);
}

CodeGenFunction::CGCapturedStmtInfo *getOldCSI() const { return OldCSI; }

static bool classof(const CGCapturedStmtInfo *Info) {
  return CGOpenMPRegionInfo::classof(Info) &&
         cast<CGOpenMPRegionInfo>(Info)->getRegionKind() == InlinedRegion;
}

~CGOpenMPInlinedRegionInfo() override = default;

309private:
/// CodeGen info about outer OpenMP region.
CodeGenFunction::CGCapturedStmtInfo *OldCSI;
CGOpenMPRegionInfo *OuterRegionInfo;
313};

315/// API for captured statement code generation in OpenMP target
316/// constructs. For this captures, implicit parameters are used instead of the
317/// captured fields. The name of the target region has to be unique in a given
318/// application so it is provided by the client, because only the client has
319/// the information to generate that.
320class CGOpenMPTargetRegionInfo final : public CGOpenMPRegionInfo {
321public:
CGOpenMPTargetRegionInfo(const CapturedStmt &CS,
                         const RegionCodeGenTy &CodeGen, StringRef HelperName)
    : CGOpenMPRegionInfo(CS, TargetRegion, CodeGen, OMPD_target,
                         /*HasCancel=*/false),
      HelperName(HelperName) {}

/// This is unused for target regions because each starts executing
/// with a single thread.
const VarDecl *getThreadIDVariable() const override { return nullptr; }

/// Get the name of the capture helper.
StringRef getHelperName() const override { return HelperName; }

static bool classof(const CGCapturedStmtInfo *Info) {
  return CGOpenMPRegionInfo::classof(Info) &&
         cast<CGOpenMPRegionInfo>(Info)->getRegionKind() == TargetRegion;
}

340private:
StringRef HelperName;
342};

344static void EmptyCodeGen(CodeGenFunction &, PrePostActionTy &) {
llvm_unreachable("No codegen for expressions")__builtin_unreachable();
346}
347/// API for generation of expressions captured in a innermost OpenMP
348/// region.
349class CGOpenMPInnerExprInfo final : public CGOpenMPInlinedRegionInfo {
350public:
CGOpenMPInnerExprInfo(CodeGenFunction &CGF, const CapturedStmt &CS)
    : CGOpenMPInlinedRegionInfo(CGF.CapturedStmtInfo, EmptyCodeGen,
                                OMPD_unknown,
                                /*HasCancel=*/false),
      PrivScope(CGF) {
  // 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 (const auto &C : CS.captures()) {
    if (!C.capturesVariable() && !C.capturesVariableByCopy())
      continue;

    const VarDecl *VD = C.getCapturedVar();
    if (VD->isLocalVarDeclOrParm())
      continue;

    DeclRefExpr DRE(CGF.getContext(), const_cast<VarDecl *>(VD),
                    /*RefersToEnclosingVariableOrCapture=*/false,
                    VD->getType().getNonReferenceType(), VK_LValue,
                    C.getLocation());
    PrivScope.addPrivate(
        VD, [&CGF, &DRE]() { return CGF.EmitLValue(&DRE).getAddress(CGF); });
  }
  (void)PrivScope.Privatize();
}

/// Lookup the captured field decl for a variable.
const FieldDecl *lookup(const VarDecl *VD) const override {
  if (const FieldDecl *FD = CGOpenMPInlinedRegionInfo::lookup(VD))
    return FD;
  return nullptr;
}

/// Emit the captured statement body.
void EmitBody(CodeGenFunction &CGF, const Stmt *S) override {
  llvm_unreachable("No body for expressions")__builtin_unreachable();
}

/// 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")__builtin_unreachable();
}

/// Get the name of the capture helper.
StringRef getHelperName() const override {
  llvm_unreachable("No helper name for expressions")__builtin_unreachable();
}

static bool classof(const CGCapturedStmtInfo *Info) { return false; }

402private:
/// Private scope to capture global variables.
CodeGenFunction::OMPPrivateScope PrivScope;
405};

407/// RAII for emitting code of OpenMP constructs.
408class InlinedOpenMPRegionRAII {
CodeGenFunction &CGF;
llvm::DenseMap<const VarDecl *, FieldDecl *> LambdaCaptureFields;
FieldDecl *LambdaThisCaptureField = nullptr;
const CodeGen::CGBlockInfo *BlockInfo = nullptr;
bool NoInheritance = false;

415public:
/// 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.
InlinedOpenMPRegionRAII(CodeGenFunction &CGF, const RegionCodeGenTy &CodeGen,
                        OpenMPDirectiveKind Kind, bool HasCancel,
                        bool NoInheritance = true)
    : CGF(CGF), NoInheritance(NoInheritance) {
  // Start emission for the construct.
  CGF.CapturedStmtInfo = new CGOpenMPInlinedRegionInfo(
      CGF.CapturedStmtInfo, CodeGen, Kind, HasCancel);
  if (NoInheritance) {
    std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
    LambdaThisCaptureField = CGF.LambdaThisCaptureField;
    CGF.LambdaThisCaptureField = nullptr;
    BlockInfo = CGF.BlockInfo;
    CGF.BlockInfo = nullptr;
  }
}

~InlinedOpenMPRegionRAII() {
  // Restore original CapturedStmtInfo only if we're done with code emission.
  auto *OldCSI =
      cast<CGOpenMPInlinedRegionInfo>(CGF.CapturedStmtInfo)->getOldCSI();
  delete CGF.CapturedStmtInfo;
  CGF.CapturedStmtInfo = OldCSI;
  if (NoInheritance) {
    std::swap(CGF.LambdaCaptureFields, LambdaCaptureFields);
    CGF.LambdaThisCaptureField = LambdaThisCaptureField;
    CGF.BlockInfo = BlockInfo;
  }
}
448};

450/// Values for bit flags used in the ident_t to describe the fields.
451/// All enumeric elements are named and described in accordance with the code
452/// from https://github.com/llvm/llvm-project/blob/main/openmp/runtime/src/kmp.h
453enum OpenMPLocationFlags : unsigned {
/// Use trampoline for internal microtask.
OMP_IDENT_IMD = 0x01,
/// Use c-style ident structure.
OMP_IDENT_KMPC = 0x02,
/// Atomic reduction option for kmpc_reduce.
OMP_ATOMIC_REDUCE = 0x10,
/// Explicit 'barrier' directive.
OMP_IDENT_BARRIER_EXPL = 0x20,
/// Implicit barrier in code.
OMP_IDENT_BARRIER_IMPL = 0x40,
/// Implicit barrier in 'for' directive.
OMP_IDENT_BARRIER_IMPL_FOR = 0x40,
/// Implicit barrier in 'sections' directive.
OMP_IDENT_BARRIER_IMPL_SECTIONS = 0xC0,
/// 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,
/// Call of __kmp_for_static_init for sections.
OMP_IDENT_WORK_SECTIONS = 0x400,
/// Call of __kmp_for_static_init for distribute.
OMP_IDENT_WORK_DISTRIBUTE = 0x800,
LLVM_MARK_AS_BITMASK_ENUM(/*LargestValue=*/OMP_IDENT_WORK_DISTRIBUTE)LLVM_BITMASK_LARGEST_ENUMERATOR = OMP_IDENT_WORK_DISTRIBUTE
477};

479namespace {
480LLVM_ENABLE_BITMASK_ENUMS_IN_NAMESPACE()using ::llvm::BitmaskEnumDetail::operator~; using ::llvm::BitmaskEnumDetail
::operator|; using ::llvm::BitmaskEnumDetail::operator&; using
 ::llvm::BitmaskEnumDetail::operator^; using ::llvm::BitmaskEnumDetail
::operator|=; using ::llvm::BitmaskEnumDetail::operator&=
; using ::llvm::BitmaskEnumDetail::operator^=;
481/// Values for bit flags for marking which requires clauses have been used.
482enum OpenMPOffloadingRequiresDirFlags : int64_t {
/// flag undefined.
OMP_REQ_UNDEFINED               = 0x000,
/// no requires clause present.
OMP_REQ_NONE                    = 0x001,
/// reverse_offload clause.
OMP_REQ_REVERSE_OFFLOAD         = 0x002,
/// unified_address clause.
OMP_REQ_UNIFIED_ADDRESS         = 0x004,
/// unified_shared_memory clause.
OMP_REQ_UNIFIED_SHARED_MEMORY   = 0x008,
/// dynamic_allocators clause.
OMP_REQ_DYNAMIC_ALLOCATORS      = 0x010,
LLVM_MARK_AS_BITMASK_ENUM(/*LargestValue=*/OMP_REQ_DYNAMIC_ALLOCATORS)LLVM_BITMASK_LARGEST_ENUMERATOR = OMP_REQ_DYNAMIC_ALLOCATORS
496};

498enum OpenMPOffloadingReservedDeviceIDs {
/// Device ID if the device was not defined, runtime should get it
/// from environment variables in the spec.
OMP_DEVICEID_UNDEF = -1,
502};
503} // anonymous namespace

505/// Describes ident structure that describes a source location.
506/// All descriptions are taken from
507/// https://github.com/llvm/llvm-project/blob/main/openmp/runtime/src/kmp.h
508/// Original structure:
509/// typedef struct ident {
510///    kmp_int32 reserved_1;   /**<  might be used in Fortran;
511///                                  see above  */
512///    kmp_int32 flags;        /**<  also f.flags; KMP_IDENT_xxx flags;
513///                                  KMP_IDENT_KMPC identifies this union
514///                                  member  */
515///    kmp_int32 reserved_2;   /**<  not really used in Fortran any more;
516///                                  see above */
517///#if USE_ITT_BUILD
518///                            /*  but currently used for storing
519///                                region-specific ITT */
520///                            /*  contextual information. */
521///#endif /* USE_ITT_BUILD */
522///    kmp_int32 reserved_3;   /**< source[4] in Fortran, do not use for
523///                                 C++  */
524///    char const *psource;    /**< String describing the source location.
525///                            The string is composed of semi-colon separated
526//                             fields which describe the source file,
527///                            the function and a pair of line numbers that
528///                            delimit the construct.
529///                             */
530/// } ident_t;
531enum IdentFieldIndex {
/// might be used in Fortran
IdentField_Reserved_1,
/// OMP_IDENT_xxx flags; OMP_IDENT_KMPC identifies this union member.
IdentField_Flags,
/// Not really used in Fortran any more
IdentField_Reserved_2,
/// Source[4] in Fortran, do not use for C++
IdentField_Reserved_3,
/// 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
544};

546/// Schedule types for 'omp for' loops (these enumerators are taken from
547/// the enum sched_type in kmp.h).
548enum OpenMPSchedType {
/// Lower bound for default (unordered) versions.
OMP_sch_lower = 32,
OMP_sch_static_chunked = 33,
OMP_sch_static = 34,
OMP_sch_dynamic_chunked = 35,
OMP_sch_guided_chunked = 36,
OMP_sch_runtime = 37,
OMP_sch_auto = 38,
/// static with chunk adjustment (e.g., simd)
OMP_sch_static_balanced_chunked = 45,
/// Lower bound for 'ordered' versions.
OMP_ord_lower = 64,
OMP_ord_static_chunked = 65,
OMP_ord_static = 66,
OMP_ord_dynamic_chunked = 67,
OMP_ord_guided_chunked = 68,
OMP_ord_runtime = 69,
OMP_ord_auto = 70,
OMP_sch_default = OMP_sch_static,
/// dist_schedule types
OMP_dist_sch_static_chunked = 91,
OMP_dist_sch_static = 92,
/// Support for OpenMP 4.5 monotonic and nonmonotonic schedule modifiers.
/// Set if the monotonic schedule modifier was present.
OMP_sch_modifier_monotonic = (1 << 29),
/// Set if the nonmonotonic schedule modifier was present.
OMP_sch_modifier_nonmonotonic = (1 << 30),
576};

578/// A basic class for pre|post-action for advanced codegen sequence for OpenMP
579/// region.
580class CleanupTy final : public EHScopeStack::Cleanup {
PrePostActionTy *Action;

583public:
explicit CleanupTy(PrePostActionTy *Action) : Action(Action) {}
void Emit(CodeGenFunction &CGF, Flags /*flags*/) override {
  if (!CGF.HaveInsertPoint())
    return;
  Action->Exit(CGF);
}
590};

592} // anonymous namespace

594void RegionCodeGenTy::operator()(CodeGenFunction &CGF) const {
CodeGenFunction::RunCleanupsScope Scope(CGF);
if (PrePostAction12.1
Field 'PrePostAction' is null
1
Field 'PrePostAction' is null
) {
13
←
Taking false branch→
  CGF.EHStack.pushCleanup<CleanupTy>(NormalAndEHCleanup, PrePostAction);
  Callback(CodeGen, CGF, *PrePostAction);
} else {
  PrePostActionTy Action;
  Callback(CodeGen, CGF, Action);
14
←
Calling 'RegionCodeGenTy::CallbackFn'→
}
603}

605/// Check if the combiner is a call to UDR combiner and if it is so return the
606/// UDR decl used for reduction.
607static const OMPDeclareReductionDecl *
608getReductionInit(const Expr *ReductionOp) {
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 (const auto *DRD = dyn_cast<OMPDeclareReductionDecl>(DRE->getDecl()))
        return DRD;
return nullptr;
616}

618static void emitInitWithReductionInitializer(CodeGenFunction &CGF,
                                           const OMPDeclareReductionDecl *DRD,
                                           const Expr *InitOp,
                                           Address Private, Address Original,
                                           QualType Ty) {
if (DRD->getInitializer()) {
  std::pair<llvm::Function *, llvm::Function *> Reduction =
      CGF.CGM.getOpenMPRuntime().getUserDefinedReduction(DRD);
  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();
  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()),
                          [=]() { return Private; });
  PrivateScope.addPrivate(cast<VarDecl>(RHSDRE->getDecl()),
                          [=]() { 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, Name);
  LValue LV = CGF.MakeNaturalAlignAddrLValue(GV, Ty);
  RValue InitRVal;
  switch (CGF.getEvaluationKind(Ty)) {
  case TEK_Scalar:
    InitRVal = CGF.EmitLoadOfLValue(LV, DRD->getLocation());
    break;
  case TEK_Complex:
    InitRVal =
        RValue::getComplex(CGF.EmitLoadOfComplex(LV, DRD->getLocation()));
    break;
  case TEK_Aggregate: {
    OpaqueValueExpr OVE(DRD->getLocation(), Ty, VK_LValue);
    CodeGenFunction::OpaqueValueMapping OpaqueMap(CGF, &OVE, LV);
    CGF.EmitAnyExprToMem(&OVE, Private, Ty.getQualifiers(),
                         /*IsInitializer=*/false);
    return;
  }
  }
  OpaqueValueExpr OVE(DRD->getLocation(), Ty, VK_PRValue);
  CodeGenFunction::OpaqueValueMapping OpaqueMap(CGF, &OVE, InitRVal);
  CGF.EmitAnyExprToMem(&OVE, Private, Ty.getQualifiers(),
                       /*IsInitializer=*/false);
}
672}

674/// Emit initialization of arrays of complex types.
675/// \param DestAddr Address of the array.
676/// \param Type Type of array.
677/// \param Init Initial expression of array.
678/// \param SrcAddr Address of the original array.
679static void EmitOMPAggregateInit(CodeGenFunction &CGF, Address DestAddr,
                               QualType Type, bool EmitDeclareReductionInit,
                               const Expr *Init,
                               const OMPDeclareReductionDecl *DRD,
                               Address SrcAddr = Address::invalid()) {
// Perform element-by-element initialization.
QualType ElementTy;

// Drill down to the base element type on both arrays.
const ArrayType *ArrayTy = Type->getAsArrayTypeUnsafe();
llvm::Value *NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, DestAddr);
DestAddr =
    CGF.Builder.CreateElementBitCast(DestAddr, DestAddr.getElementType());
if (DRD)
  SrcAddr =
      CGF.Builder.CreateElementBitCast(SrcAddr, DestAddr.getElementType());

llvm::Value *SrcBegin = nullptr;
if (DRD)
  SrcBegin = SrcAddr.getPointer();
llvm::Value *DestBegin = DestAddr.getPointer();
// Cast from pointer to array type to pointer to single element.
llvm::Value *DestEnd =
    CGF.Builder.CreateGEP(DestAddr.getElementType(), DestBegin, NumElements);
// The basic structure here is a while-do loop.
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.
llvm::BasicBlock *EntryBB = CGF.Builder.GetInsertBlock();
CGF.EmitBlock(BodyBB);

CharUnits ElementSize = CGF.getContext().getTypeSizeInChars(ElementTy);

llvm::PHINode *SrcElementPHI = nullptr;
Address SrcElementCurrent = Address::invalid();
if (DRD) {
  SrcElementPHI = CGF.Builder.CreatePHI(SrcBegin->getType(), 2,
                                        "omp.arraycpy.srcElementPast");
  SrcElementPHI->addIncoming(SrcBegin, EntryBB);
  SrcElementCurrent =
      Address(SrcElementPHI,
              SrcAddr.getAlignment().alignmentOfArrayElement(ElementSize));
}
llvm::PHINode *DestElementPHI = CGF.Builder.CreatePHI(
    DestBegin->getType(), 2, "omp.arraycpy.destElementPast");
DestElementPHI->addIncoming(DestBegin, EntryBB);
Address DestElementCurrent =
    Address(DestElementPHI,
            DestAddr.getAlignment().alignmentOfArrayElement(ElementSize));

// Emit copy.
{
  CodeGenFunction::RunCleanupsScope InitScope(CGF);
  if (EmitDeclareReductionInit) {
    emitInitWithReductionInitializer(CGF, DRD, Init, DestElementCurrent,
                                     SrcElementCurrent, ElementTy);
  } else
    CGF.EmitAnyExprToMem(Init, DestElementCurrent, ElementTy.getQualifiers(),
                         /*IsInitializer=*/false);
}

if (DRD) {
  // Shift the address forward by one element.
  llvm::Value *SrcElementNext = CGF.Builder.CreateConstGEP1_32(
      SrcAddr.getElementType(), SrcElementPHI, /*Idx0=*/1,
      "omp.arraycpy.dest.element");
  SrcElementPHI->addIncoming(SrcElementNext, CGF.Builder.GetInsertBlock());
}

// Shift the address forward by one element.
llvm::Value *DestElementNext = CGF.Builder.CreateConstGEP1_32(
    DestAddr.getElementType(), DestElementPHI, /*Idx0=*/1,
    "omp.arraycpy.dest.element");
// Check whether we've reached the end.
llvm::Value *Done =
    CGF.Builder.CreateICmpEQ(DestElementNext, DestEnd, "omp.arraycpy.done");
CGF.Builder.CreateCondBr(Done, DoneBB, BodyBB);
DestElementPHI->addIncoming(DestElementNext, CGF.Builder.GetInsertBlock());

// Done.
CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
764}

766LValue ReductionCodeGen::emitSharedLValue(CodeGenFunction &CGF, const Expr *E) {
return CGF.EmitOMPSharedLValue(E);
768}

770LValue ReductionCodeGen::emitSharedLValueUB(CodeGenFunction &CGF,
                                          const Expr *E) {
if (const auto *OASE = dyn_cast<OMPArraySectionExpr>(E))
  return CGF.EmitOMPArraySectionExpr(OASE, /*IsLowerBound=*/false);
return LValue();
775}

777void ReductionCodeGen::emitAggregateInitialization(
  CodeGenFunction &CGF, unsigned N, Address PrivateAddr, LValue SharedLVal,
  const OMPDeclareReductionDecl *DRD) {
// Emit VarDecl with copy init for arrays.
// Get the address of the original variable captured in current
// captured region.
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
bool EmitDeclareReductionInit =
    DRD && (DRD->getInitializer() || !PrivateVD->hasInit());
EmitOMPAggregateInit(CGF, PrivateAddr, PrivateVD->getType(),
                     EmitDeclareReductionInit,
                     EmitDeclareReductionInit ? ClausesData[N].ReductionOp
                                              : PrivateVD->getInit(),
                     DRD, SharedLVal.getAddress(CGF));
792}

794ReductionCodeGen::ReductionCodeGen(ArrayRef<const Expr *> Shareds,
                                 ArrayRef<const Expr *> Origs,
                                 ArrayRef<const Expr *> Privates,
                                 ArrayRef<const Expr *> ReductionOps) {
ClausesData.reserve(Shareds.size());
SharedAddresses.reserve(Shareds.size());
Sizes.reserve(Shareds.size());
BaseDecls.reserve(Shareds.size());
const auto *IOrig = Origs.begin();
const auto *IPriv = Privates.begin();
const auto *IRed = ReductionOps.begin();
for (const Expr *Ref : Shareds) {
  ClausesData.emplace_back(Ref, *IOrig, *IPriv, *IRed);
  std::advance(IOrig, 1);
  std::advance(IPriv, 1);
  std::advance(IRed, 1);
}
811}

813void ReductionCodeGen::emitSharedOrigLValue(CodeGenFunction &CGF, unsigned N) {
assert(SharedAddresses.size() == N && OrigAddresses.size() == N &&((void)0)
       "Number of generated lvalues must be exactly N.")((void)0);
LValue First = emitSharedLValue(CGF, ClausesData[N].Shared);
LValue Second = emitSharedLValueUB(CGF, ClausesData[N].Shared);
SharedAddresses.emplace_back(First, Second);
if (ClausesData[N].Shared == ClausesData[N].Ref) {
  OrigAddresses.emplace_back(First, Second);
} else {
  LValue First = emitSharedLValue(CGF, ClausesData[N].Ref);
  LValue Second = emitSharedLValueUB(CGF, ClausesData[N].Ref);
  OrigAddresses.emplace_back(First, Second);
}
826}

828void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N) {
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
bool AsArraySection = isa<OMPArraySectionExpr>(ClausesData[N].Ref);
if (!PrivateType->isVariablyModifiedType()) {
  Sizes.emplace_back(
      CGF.getTypeSize(OrigAddresses[N].first.getType().getNonReferenceType()),
      nullptr);
  return;
}
llvm::Value *Size;
llvm::Value *SizeInChars;
auto *ElemType =
    cast<llvm::PointerType>(OrigAddresses[N].first.getPointer(CGF)->getType())
        ->getElementType();
auto *ElemSizeOf = llvm::ConstantExpr::getSizeOf(ElemType);
if (AsArraySection) {
  Size = CGF.Builder.CreatePtrDiff(OrigAddresses[N].second.getPointer(CGF),
                                   OrigAddresses[N].first.getPointer(CGF));
  Size = CGF.Builder.CreateNUWAdd(
      Size, llvm::ConstantInt::get(Size->getType(), /*V=*/1));
  SizeInChars = CGF.Builder.CreateNUWMul(Size, ElemSizeOf);
} else {
  SizeInChars =
      CGF.getTypeSize(OrigAddresses[N].first.getType().getNonReferenceType());
  Size = CGF.Builder.CreateExactUDiv(SizeInChars, ElemSizeOf);
}
Sizes.emplace_back(SizeInChars, Size);
CodeGenFunction::OpaqueValueMapping OpaqueMap(
    CGF,
    cast<OpaqueValueExpr>(
        CGF.getContext().getAsVariableArrayType(PrivateType)->getSizeExpr()),
    RValue::get(Size));
CGF.EmitVariablyModifiedType(PrivateType);
863}

865void ReductionCodeGen::emitAggregateType(CodeGenFunction &CGF, unsigned N,
                                       llvm::Value *Size) {
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
if (!PrivateType->isVariablyModifiedType()) {
  assert(!Size && !Sizes[N].second &&((void)0)
         "Size should be nullptr for non-variably modified reduction "((void)0)
         "items.")((void)0);
  return;
}
CodeGenFunction::OpaqueValueMapping OpaqueMap(
    CGF,
    cast<OpaqueValueExpr>(
        CGF.getContext().getAsVariableArrayType(PrivateType)->getSizeExpr()),
    RValue::get(Size));
CGF.EmitVariablyModifiedType(PrivateType);
882}

884void ReductionCodeGen::emitInitialization(
  CodeGenFunction &CGF, unsigned N, Address PrivateAddr, LValue SharedLVal,
  llvm::function_ref<bool(CodeGenFunction &)> DefaultInit) {
assert(SharedAddresses.size() > N && "No variable was generated")((void)0);
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
const OMPDeclareReductionDecl *DRD =
    getReductionInit(ClausesData[N].ReductionOp);
QualType PrivateType = PrivateVD->getType();
PrivateAddr = CGF.Builder.CreateElementBitCast(
    PrivateAddr, CGF.ConvertTypeForMem(PrivateType));
QualType SharedType = SharedAddresses[N].first.getType();
SharedLVal = CGF.MakeAddrLValue(
    CGF.Builder.CreateElementBitCast(SharedLVal.getAddress(CGF),
                                     CGF.ConvertTypeForMem(SharedType)),
    SharedType, SharedAddresses[N].first.getBaseInfo(),
    CGF.CGM.getTBAAInfoForSubobject(SharedAddresses[N].first, SharedType));
if (CGF.getContext().getAsArrayType(PrivateVD->getType())) {
  if (DRD && DRD->getInitializer())
    (void)DefaultInit(CGF);
  emitAggregateInitialization(CGF, N, PrivateAddr, SharedLVal, DRD);
} else if (DRD && (DRD->getInitializer() || !PrivateVD->hasInit())) {
  (void)DefaultInit(CGF);
  emitInitWithReductionInitializer(CGF, DRD, ClausesData[N].ReductionOp,
                                   PrivateAddr, SharedLVal.getAddress(CGF),
                                   SharedLVal.getType());
} else if (!DefaultInit(CGF) && PrivateVD->hasInit() &&
           !CGF.isTrivialInitializer(PrivateVD->getInit())) {
  CGF.EmitAnyExprToMem(PrivateVD->getInit(), PrivateAddr,
                       PrivateVD->getType().getQualifiers(),
                       /*IsInitializer=*/false);
}
916}

918bool ReductionCodeGen::needCleanups(unsigned N) {
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
QualType::DestructionKind DTorKind = PrivateType.isDestructedType();
return DTorKind != QualType::DK_none;
924}

926void ReductionCodeGen::emitCleanups(CodeGenFunction &CGF, unsigned N,
                                  Address PrivateAddr) {
const auto *PrivateVD =
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Private)->getDecl());
QualType PrivateType = PrivateVD->getType();
QualType::DestructionKind DTorKind = PrivateType.isDestructedType();
if (needCleanups(N)) {
  PrivateAddr = CGF.Builder.CreateElementBitCast(
      PrivateAddr, CGF.ConvertTypeForMem(PrivateType));
  CGF.pushDestroy(DTorKind, PrivateAddr, PrivateType);
}
937}

939static LValue loadToBegin(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
                        LValue BaseLV) {
BaseTy = BaseTy.getNonReferenceType();
while ((BaseTy->isPointerType() || BaseTy->isReferenceType()) &&
       !CGF.getContext().hasSameType(BaseTy, ElTy)) {
  if (const auto *PtrTy = BaseTy->getAs<PointerType>()) {
    BaseLV = CGF.EmitLoadOfPointerLValue(BaseLV.getAddress(CGF), PtrTy);
  } else {
    LValue RefLVal = CGF.MakeAddrLValue(BaseLV.getAddress(CGF), BaseTy);
    BaseLV = CGF.EmitLoadOfReferenceLValue(RefLVal);
  }
  BaseTy = BaseTy->getPointeeType();
}
return CGF.MakeAddrLValue(
    CGF.Builder.CreateElementBitCast(BaseLV.getAddress(CGF),
                                     CGF.ConvertTypeForMem(ElTy)),
    BaseLV.getType(), BaseLV.getBaseInfo(),
    CGF.CGM.getTBAAInfoForSubobject(BaseLV, BaseLV.getType()));
957}

959static Address castToBase(CodeGenFunction &CGF, QualType BaseTy, QualType ElTy,
                        llvm::Type *BaseLVType, CharUnits BaseLVAlignment,
                        llvm::Value *Addr) {
Address Tmp = Address::invalid();
Address TopTmp = Address::invalid();
Address MostTopTmp = Address::invalid();
BaseTy = BaseTy.getNonReferenceType();
while ((BaseTy->isPointerType() || BaseTy->isReferenceType()) &&
       !CGF.getContext().hasSameType(BaseTy, ElTy)) {
  Tmp = CGF.CreateMemTemp(BaseTy);
  if (TopTmp.isValid())
    CGF.Builder.CreateStore(Tmp.getPointer(), TopTmp);
  else
    MostTopTmp = Tmp;
  TopTmp = Tmp;
  BaseTy = BaseTy->getPointeeType();
}
llvm::Type *Ty = BaseLVType;
if (Tmp.isValid())
  Ty = Tmp.getElementType();
Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(Addr, Ty);
if (Tmp.isValid()) {
  CGF.Builder.CreateStore(Addr, Tmp);
  return MostTopTmp;
}
return Address(Addr, BaseLVAlignment);
985}

987static const VarDecl *getBaseDecl(const Expr *Ref, const DeclRefExpr *&DE) {
const VarDecl *OrigVD = nullptr;
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 (const auto *TempASE = dyn_cast<ArraySubscriptExpr>(Base))
    Base = TempASE->getBase()->IgnoreParenImpCasts();
  DE = cast<DeclRefExpr>(Base);
  OrigVD = cast<VarDecl>(DE->getDecl());
} 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());
}
return OrigVD;
1005}

1007Address ReductionCodeGen::adjustPrivateAddress(CodeGenFunction &CGF, unsigned N,
                                             Address PrivateAddr) {
const DeclRefExpr *DE;
if (const VarDecl *OrigVD = ::getBaseDecl(ClausesData[N].Ref, DE)) {
  BaseDecls.emplace_back(OrigVD);
  LValue OriginalBaseLValue = CGF.EmitLValue(DE);
  LValue BaseLValue =
      loadToBegin(CGF, OrigVD->getType(), SharedAddresses[N].first.getType(),
                  OriginalBaseLValue);
  Address SharedAddr = SharedAddresses[N].first.getAddress(CGF);
  llvm::Value *Adjustment = CGF.Builder.CreatePtrDiff(
      BaseLValue.getPointer(CGF), SharedAddr.getPointer());
  llvm::Value *PrivatePointer =
      CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
          PrivateAddr.getPointer(), SharedAddr.getType());
  llvm::Value *Ptr = CGF.Builder.CreateGEP(
      SharedAddr.getElementType(), PrivatePointer, Adjustment);
  return castToBase(CGF, OrigVD->getType(),
                    SharedAddresses[N].first.getType(),
                    OriginalBaseLValue.getAddress(CGF).getType(),
                    OriginalBaseLValue.getAlignment(), Ptr);
}
BaseDecls.emplace_back(
    cast<VarDecl>(cast<DeclRefExpr>(ClausesData[N].Ref)->getDecl()));
return PrivateAddr;
1032}

1034bool ReductionCodeGen::usesReductionInitializer(unsigned N) const {
const OMPDeclareReductionDecl *DRD =
    getReductionInit(ClausesData[N].ReductionOp);
return DRD && DRD->getInitializer();
1038}

1040LValue CGOpenMPRegionInfo::getThreadIDVariableLValue(CodeGenFunction &CGF) {
return CGF.EmitLoadOfPointerLValue(
    CGF.GetAddrOfLocalVar(getThreadIDVariable()),
    getThreadIDVariable()->getType()->castAs<PointerType>());
1044}

1046void CGOpenMPRegionInfo::EmitBody(CodeGenFunction &CGF, const Stmt *S) {
if (!CGF.HaveInsertPoint())
  return;
// 1.2.2 OpenMP Language Terminology
// Structured block - An executable statement with a single entry at the
// top and a single exit at the bottom.
// The point of exit cannot be a branch out of the structured block.
// longjmp() and throw() must not violate the entry/exit criteria.
CGF.EHStack.pushTerminate();
if (S)
  CGF.incrementProfileCounter(S);
CodeGen(CGF);
CGF.EHStack.popTerminate();
1059}

1061LValue CGOpenMPTaskOutlinedRegionInfo::getThreadIDVariableLValue(
  CodeGenFunction &CGF) {
return CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(getThreadIDVariable()),
                          getThreadIDVariable()->getType(),
                          AlignmentSource::Decl);
1066}

1068static 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;
1077}

1079CGOpenMPRuntime::CGOpenMPRuntime(CodeGenModule &CGM, StringRef FirstSeparator,
                               StringRef Separator)
  : CGM(CGM), FirstSeparator(FirstSeparator), Separator(Separator),
    OMPBuilder(CGM.getModule()), OffloadEntriesInfoManager(CGM) {
KmpCriticalNameTy = llvm::ArrayType::get(CGM.Int32Ty, /*NumElements*/ 8);

// Initialize Types used in OpenMPIRBuilder from OMPKinds.def
OMPBuilder.initialize();
loadOffloadInfoMetadata();
1088}

1090void CGOpenMPRuntime::clear() {
InternalVars.clear();
// Clean non-target variable declarations possibly used only in debug info.
for (const auto &Data : EmittedNonTargetVariables) {
  if (!Data.getValue().pointsToAliveValue())
    continue;
  auto *GV = dyn_cast<llvm::GlobalVariable>(Data.getValue());
  if (!GV)
    continue;
  if (!GV->isDeclaration() || GV->getNumUses() > 0)
    continue;
  GV->eraseFromParent();
}
1103}

1105std::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 std::string(OS.str());
1114}

1116static llvm::Function *
1117emitCombinerOrInitializer(CodeGenModule &CGM, QualType Ty,
                        const Expr *CombinerInitializer, const VarDecl *In,
                        const VarDecl *Out, bool IsCombiner) {
// void .omp_combiner.(Ty *in, Ty *out);
ASTContext &C = CGM.getContext();
QualType PtrTy = C.getPointerType(Ty).withRestrict();
FunctionArgList Args;
ImplicitParamDecl OmpOutParm(C, /*DC=*/nullptr, Out->getLocation(),
                             /*Id=*/nullptr, PtrTy, ImplicitParamDecl::Other);
ImplicitParamDecl OmpInParm(C, /*DC=*/nullptr, In->getLocation(),
                            /*Id=*/nullptr, PtrTy, ImplicitParamDecl::Other);
Args.push_back(&OmpOutParm);
Args.push_back(&OmpInParm);
const CGFunctionInfo &FnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
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);
if (CGM.getLangOpts().Optimize) {
  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, In->getLocation(),
                  Out->getLocation());
CodeGenFunction::OMPPrivateScope Scope(CGF);
Address AddrIn = CGF.GetAddrOfLocalVar(&OmpInParm);
Scope.addPrivate(In, [&CGF, AddrIn, PtrTy]() {
  return CGF.EmitLoadOfPointerLValue(AddrIn, PtrTy->castAs<PointerType>())
      .getAddress(CGF);
});
Address AddrOut = CGF.GetAddrOfLocalVar(&OmpOutParm);
Scope.addPrivate(Out, [&CGF, AddrOut, PtrTy]() {
  return CGF.EmitLoadOfPointerLValue(AddrOut, PtrTy->castAs<PointerType>())
      .getAddress(CGF);
});
(void)Scope.Privatize();
if (!IsCombiner && Out->hasInit() &&
    !CGF.isTrivialInitializer(Out->getInit())) {
  CGF.EmitAnyExprToMem(Out->getInit(), CGF.GetAddrOfLocalVar(Out),
                       Out->getType().getQualifiers(),
                       /*IsInitializer=*/true);
}
if (CombinerInitializer)
  CGF.EmitIgnoredExpr(CombinerInitializer);
Scope.ForceCleanup();
CGF.FinishFunction();
return Fn;
1171}

1173void CGOpenMPRuntime::emitUserDefinedReduction(
  CodeGenFunction *CGF, const OMPDeclareReductionDecl *D) {
if (UDRMap.count(D) > 0)
  return;
llvm::Function *Combiner = emitCombinerOrInitializer(
    CGM, D->getType(), D->getCombiner(),
    cast<VarDecl>(cast<DeclRefExpr>(D->getCombinerIn())->getDecl()),
    cast<VarDecl>(cast<DeclRefExpr>(D->getCombinerOut())->getDecl()),
    /*IsCombiner=*/true);
llvm::Function *Initializer = nullptr;
if (const Expr *Init = D->getInitializer()) {
  Initializer = emitCombinerOrInitializer(
      CGM, D->getType(),
      D->getInitializerKind() == OMPDeclareReductionDecl::CallInit ? Init
                                                                   : nullptr,
      cast<VarDecl>(cast<DeclRefExpr>(D->getInitOrig())->getDecl()),
      cast<VarDecl>(cast<DeclRefExpr>(D->getInitPriv())->getDecl()),
      /*IsCombiner=*/false);
}
UDRMap.try_emplace(D, Combiner, Initializer);
if (CGF) {
  auto &Decls = FunctionUDRMap.FindAndConstruct(CGF->CurFn);
  Decls.second.push_back(D);
}
1197}

1199std::pair<llvm::Function *, llvm::Function *>
1200CGOpenMPRuntime::getUserDefinedReduction(const OMPDeclareReductionDecl *D) {
auto I = UDRMap.find(D);
if (I != UDRMap.end())
  return I->second;
emitUserDefinedReduction(/*CGF=*/nullptr, D);
return UDRMap.lookup(D);
1206}

1208namespace {
1209// Temporary RAII solution to perform a push/pop stack event on the OpenMP IR
1210// Builder if one is present.
1211struct PushAndPopStackRAII {
PushAndPopStackRAII(llvm::OpenMPIRBuilder *OMPBuilder, CodeGenFunction &CGF,
                    bool HasCancel, llvm::omp::Directive Kind)
    : OMPBuilder(OMPBuilder) {
  if (!OMPBuilder)
    return;

  // The following callback is the crucial part of clangs cleanup process.
  //
  // NOTE:
  // Once the OpenMPIRBuilder is used to create parallel regions (and
  // similar), the cancellation destination (Dest below) is determined via
  // IP. That means if we have variables to finalize we split the block at IP,
  // use the new block (=BB) as destination to build a JumpDest (via
  // getJumpDestInCurrentScope(BB)) which then is fed to
  // EmitBranchThroughCleanup. Furthermore, there will not be the need
  // to push & pop an FinalizationInfo object.
  // The FiniCB will still be needed but at the point where the
  // OpenMPIRBuilder is asked to construct a parallel (or similar) construct.
  auto FiniCB = [&CGF](llvm::OpenMPIRBuilder::InsertPointTy IP) {
    assert(IP.getBlock()->end() == IP.getPoint() &&((void)0)
           "Clang CG should cause non-terminated block!")((void)0);
    CGBuilderTy::InsertPointGuard IPG(CGF.Builder);
    CGF.Builder.restoreIP(IP);
    CodeGenFunction::JumpDest Dest =
        CGF.getOMPCancelDestination(OMPD_parallel);
    CGF.EmitBranchThroughCleanup(Dest);
  };

  // TODO: Remove this once we emit parallel regions through the
  //       OpenMPIRBuilder as it can do this setup internally.
  llvm::OpenMPIRBuilder::FinalizationInfo FI({FiniCB, Kind, HasCancel});
  OMPBuilder->pushFinalizationCB(std::move(FI));
}
~PushAndPopStackRAII() {
  if (OMPBuilder)
    OMPBuilder->popFinalizationCB();
}
llvm::OpenMPIRBuilder *OMPBuilder;
1250};
1251} // namespace

1253static llvm::Function *emitParallelOrTeamsOutlinedFunction(
  CodeGenModule &CGM, const OMPExecutableDirective &D, const CapturedStmt *CS,
  const VarDecl *ThreadIDVar, OpenMPDirectiveKind InnermostKind,
  const StringRef OutlinedHelperName, const RegionCodeGenTy &CodeGen) {
assert(ThreadIDVar->getType()->isPointerType() &&((void)0)
       "thread id variable must be of type kmp_int32 *")((void)0);
CodeGenFunction CGF(CGM, true);
bool HasCancel = false;
if (const auto *OPD = dyn_cast<OMPParallelDirective>(&D))
  HasCancel = OPD->hasCancel();
else if (const auto *OPD = dyn_cast<OMPTargetParallelDirective>(&D))
  HasCancel = OPD->hasCancel();
else if (const auto *OPSD = dyn_cast<OMPParallelSectionsDirective>(&D))
  HasCancel = OPSD->hasCancel();
else if (const auto *OPFD = dyn_cast<OMPParallelForDirective>(&D))
  HasCancel = OPFD->hasCancel();
else if (const auto *OPFD = dyn_cast<OMPTargetParallelForDirective>(&D))
  HasCancel = OPFD->hasCancel();
else if (const auto *OPFD = dyn_cast<OMPDistributeParallelForDirective>(&D))
  HasCancel = OPFD->hasCancel();
else if (const auto *OPFD =
             dyn_cast<OMPTeamsDistributeParallelForDirective>(&D))
  HasCancel = OPFD->hasCancel();
else if (const auto *OPFD =
             dyn_cast<OMPTargetTeamsDistributeParallelForDirective>(&D))
  HasCancel = OPFD->hasCancel();

// TODO: Temporarily inform the OpenMPIRBuilder, if any, about the new
//       parallel region to make cancellation barriers work properly.
llvm::OpenMPIRBuilder &OMPBuilder = CGM.getOpenMPRuntime().getOMPBuilder();
PushAndPopStackRAII PSR(&OMPBuilder, CGF, HasCancel, InnermostKind);
CGOpenMPOutlinedRegionInfo CGInfo(*CS, ThreadIDVar, CodeGen, InnermostKind,
                                  HasCancel, OutlinedHelperName);
CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
return CGF.GenerateOpenMPCapturedStmtFunction(*CS, D.getBeginLoc());
1288}

1290llvm::Function *CGOpenMPRuntime::emitParallelOutlinedFunction(
  const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
  OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
const CapturedStmt *CS = D.getCapturedStmt(OMPD_parallel);
return emitParallelOrTeamsOutlinedFunction(
    CGM, D, CS, ThreadIDVar, InnermostKind, getOutlinedHelperName(), CodeGen);
1296}

1298llvm::Function *CGOpenMPRuntime::emitTeamsOutlinedFunction(
  const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
  OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen) {
const CapturedStmt *CS = D.getCapturedStmt(OMPD_teams);
return emitParallelOrTeamsOutlinedFunction(
    CGM, D, CS, ThreadIDVar, InnermostKind, getOutlinedHelperName(), CodeGen);
1304}

1306llvm::Function *CGOpenMPRuntime::emitTaskOutlinedFunction(
  const OMPExecutableDirective &D, const VarDecl *ThreadIDVar,
  const VarDecl *PartIDVar, const VarDecl *TaskTVar,
  OpenMPDirectiveKind InnermostKind, const RegionCodeGenTy &CodeGen,
  bool Tied, unsigned &NumberOfParts) {
auto &&UntiedCodeGen = [this, &D, TaskTVar](CodeGenFunction &CGF,
                                            PrePostActionTy &) {
  llvm::Value *ThreadID = getThreadID(CGF, D.getBeginLoc());
  llvm::Value *UpLoc = emitUpdateLocation(CGF, D.getBeginLoc());
  llvm::Value *TaskArgs[] = {
      UpLoc, ThreadID,
      CGF.EmitLoadOfPointerLValue(CGF.GetAddrOfLocalVar(TaskTVar),
                                  TaskTVar->getType()->castAs<PointerType>())
          .getPointer(CGF)};
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_omp_task),
                      TaskArgs);
};
CGOpenMPTaskOutlinedRegionInfo::UntiedTaskActionTy Action(Tied, PartIDVar,
                                                          UntiedCodeGen);
CodeGen.setAction(Action);
assert(!ThreadIDVar->getType()->isPointerType() &&((void)0)
       "thread id variable must be of type kmp_int32 for tasks")((void)0);
const OpenMPDirectiveKind Region =
    isOpenMPTaskLoopDirective(D.getDirectiveKind()) ? OMPD_taskloop
                                                    : OMPD_task;
const CapturedStmt *CS = D.getCapturedStmt(Region);
bool HasCancel = false;
if (const auto *TD = dyn_cast<OMPTaskDirective>(&D))
  HasCancel = TD->hasCancel();
else if (const auto *TD = dyn_cast<OMPTaskLoopDirective>(&D))
  HasCancel = TD->hasCancel();
else if (const auto *TD = dyn_cast<OMPMasterTaskLoopDirective>(&D))
  HasCancel = TD->hasCancel();
else if (const auto *TD = dyn_cast<OMPParallelMasterTaskLoopDirective>(&D))
  HasCancel = TD->hasCancel();

CodeGenFunction CGF(CGM, true);
CGOpenMPTaskOutlinedRegionInfo CGInfo(*CS, ThreadIDVar, CodeGen,
                                      InnermostKind, HasCancel, Action);
CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CGInfo);
llvm::Function *Res = CGF.GenerateCapturedStmtFunction(*CS);
if (!Tied)
  NumberOfParts = Action.getNumberOfParts();
return Res;
1351}

1353static 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;
}
1369}

1371template <class... As>
1372static llvm::GlobalVariable *
1373createGlobalStruct(CodeGenModule &CGM, QualType Ty, bool IsConstant,
                 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,
    std::forward<As>(Args)...);
1384}

1386template <typename T>
1387static void
1388createConstantGlobalStructAndAddToParent(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);
1396}

1398void CGOpenMPRuntime::setLocThreadIdInsertPt(CodeGenFunction &CGF,
                                           bool AtCurrentPoint) {
auto &Elem = OpenMPLocThreadIDMap.FindAndConstruct(CGF.CurFn);
assert(!Elem.second.ServiceInsertPt && "Insert point is set already.")((void)0);

llvm::Value *Undef = llvm::UndefValue::get(CGF.Int32Ty);
if (AtCurrentPoint) {
  Elem.second.ServiceInsertPt = new llvm::BitCastInst(
      Undef, CGF.Int32Ty, "svcpt", CGF.Builder.GetInsertBlock());
} else {
  Elem.second.ServiceInsertPt =
      new llvm::BitCastInst(Undef, CGF.Int32Ty, "svcpt");
  Elem.second.ServiceInsertPt->insertAfter(CGF.AllocaInsertPt);
}
1412}

1414void CGOpenMPRuntime::clearLocThreadIdInsertPt(CodeGenFunction &CGF) {
auto &Elem = OpenMPLocThreadIDMap.FindAndConstruct(CGF.CurFn);
if (Elem.second.ServiceInsertPt) {
  llvm::Instruction *Ptr = Elem.second.ServiceInsertPt;
  Elem.second.ServiceInsertPt = nullptr;
  Ptr->eraseFromParent();
}
1421}

1423static StringRef getIdentStringFromSourceLocation(CodeGenFunction &CGF,
                                                SourceLocation Loc,
                                                SmallString<128> &Buffer) {
llvm::raw_svector_ostream OS(Buffer);
// Build debug location
PresumedLoc PLoc = CGF.getContext().getSourceManager().getPresumedLoc(Loc);
OS << ";" << PLoc.getFilename() << ";";
if (const auto *FD = dyn_cast_or_null<FunctionDecl>(CGF.CurFuncDecl))
  OS << FD->getQualifiedNameAsString();
OS << ";" << PLoc.getLine() << ";" << PLoc.getColumn() << ";;";
return OS.str();
1434}

1436llvm::Value *CGOpenMPRuntime::emitUpdateLocation(CodeGenFunction &CGF,
                                               SourceLocation Loc,
                                               unsigned Flags) {
llvm::Constant *SrcLocStr;
if (CGM.getCodeGenOpts().getDebugInfo() == codegenoptions::NoDebugInfo ||
    Loc.isInvalid()) {
  SrcLocStr = OMPBuilder.getOrCreateDefaultSrcLocStr();
} else {
  std::string FunctionName = "";
  if (const auto *FD = dyn_cast_or_null<FunctionDecl>(CGF.CurFuncDecl))
    FunctionName = FD->getQualifiedNameAsString();
  PresumedLoc PLoc = CGF.getContext().getSourceManager().getPresumedLoc(Loc);
  const char *FileName = PLoc.getFilename();
  unsigned Line = PLoc.getLine();
  unsigned Column = PLoc.getColumn();
  SrcLocStr = OMPBuilder.getOrCreateSrcLocStr(FunctionName.c_str(), FileName,
                                              Line, Column);
}
unsigned Reserved2Flags = getDefaultLocationReserved2Flags();
return OMPBuilder.getOrCreateIdent(SrcLocStr, llvm::omp::IdentFlag(Flags),
                                   Reserved2Flags);
1457}

1459llvm::Value *CGOpenMPRuntime::getThreadID(CodeGenFunction &CGF,
                                        SourceLocation Loc) {
assert(CGF.CurFn && "No function in current CodeGenFunction.")((void)0);
// If the OpenMPIRBuilder is used we need to use it for all thread id calls as
// the clang invariants used below might be broken.
if (CGM.getLangOpts().OpenMPIRBuilder) {
  SmallString<128> Buffer;
  OMPBuilder.updateToLocation(CGF.Builder.saveIP());
  auto *SrcLocStr = OMPBuilder.getOrCreateSrcLocStr(
      getIdentStringFromSourceLocation(CGF, Loc, Buffer));
  return OMPBuilder.getOrCreateThreadID(
      OMPBuilder.getOrCreateIdent(SrcLocStr));
}

llvm::Value *ThreadID = nullptr;
// Check whether we've already cached a load of the thread id in this
// function.
auto I = OpenMPLocThreadIDMap.find(CGF.CurFn);
if (I != OpenMPLocThreadIDMap.end()) {
  ThreadID = I->second.ThreadID;
  if (ThreadID != nullptr)
    return ThreadID;
}
// If exceptions are enabled, do not use parameter to avoid possible crash.
if (auto *OMPRegionInfo =
        dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo)) {
  if (OMPRegionInfo->getThreadIDVariable()) {
    // Check if this an outlined function with thread id passed as argument.
    LValue LVal = OMPRegionInfo->getThreadIDVariableLValue(CGF);
    llvm::BasicBlock *TopBlock = CGF.AllocaInsertPt->getParent();
    if (!CGF.EHStack.requiresLandingPad() || !CGF.getLangOpts().Exceptions ||
        !CGF.getLangOpts().CXXExceptions ||
        CGF.Builder.GetInsertBlock() == TopBlock ||
        !isa<llvm::Instruction>(LVal.getPointer(CGF)) ||
        cast<llvm::Instruction>(LVal.getPointer(CGF))->getParent() ==
            TopBlock ||
        cast<llvm::Instruction>(LVal.getPointer(CGF))->getParent() ==
            CGF.Builder.GetInsertBlock()) {
      ThreadID = CGF.EmitLoadOfScalar(LVal, Loc);
      // If value loaded in entry block, cache it and use it everywhere in
      // function.
      if (CGF.Builder.GetInsertBlock() == TopBlock) {
        auto &Elem = OpenMPLocThreadIDMap.FindAndConstruct(CGF.CurFn);
        Elem.second.ThreadID = ThreadID;
      }
      return ThreadID;
    }
  }
}

// This is not an outlined function region - need to call __kmpc_int32
// kmpc_global_thread_num(ident_t *loc).
// Generate thread id value and cache this value for use across the
// function.
auto &Elem = OpenMPLocThreadIDMap.FindAndConstruct(CGF.CurFn);
if (!Elem.second.ServiceInsertPt)
  setLocThreadIdInsertPt(CGF);
CGBuilderTy::InsertPointGuard IPG(CGF.Builder);
CGF.Builder.SetInsertPoint(Elem.second.ServiceInsertPt);
llvm::CallInst *Call = CGF.Builder.CreateCall(
    OMPBuilder.getOrCreateRuntimeFunction(CGM.getModule(),
                                          OMPRTL___kmpc_global_thread_num),
    emitUpdateLocation(CGF, Loc));
Call->setCallingConv(CGF.getRuntimeCC());
Elem.second.ThreadID = Call;
return Call;
1525}

1527void CGOpenMPRuntime::functionFinished(CodeGenFunction &CGF) {
assert(CGF.CurFn && "No function in current CodeGenFunction.")((void)0);
if (OpenMPLocThreadIDMap.count(CGF.CurFn)) {
  clearLocThreadIdInsertPt(CGF);
  OpenMPLocThreadIDMap.erase(CGF.CurFn);
}
if (FunctionUDRMap.count(CGF.CurFn) > 0) {
  for(const auto *D : FunctionUDRMap[CGF.CurFn])
    UDRMap.erase(D);
  FunctionUDRMap.erase(CGF.CurFn);
}
auto I = FunctionUDMMap.find(CGF.CurFn);
if (I != FunctionUDMMap.end()) {
  for(const auto *D : I->second)
    UDMMap.erase(D);
  FunctionUDMMap.erase(I);
}
LastprivateConditionalToTypes.erase(CGF.CurFn);
FunctionToUntiedTaskStackMap.erase(CGF.CurFn);
1546}

1548llvm::Type *CGOpenMPRuntime::getIdentTyPointerTy() {
return OMPBuilder.IdentPtr;
1550}

1552llvm::Type *CGOpenMPRuntime::getKmpc_MicroPointerTy() {
if (!Kmpc_MicroTy) {
  // Build void (*kmpc_micro)(kmp_int32 *global_tid, kmp_int32 *bound_tid,...)
  llvm::Type *MicroParams[] = {llvm::PointerType::getUnqual(CGM.Int32Ty),
                               llvm::PointerType::getUnqual(CGM.Int32Ty)};
  Kmpc_MicroTy = llvm::FunctionType::get(CGM.VoidTy, MicroParams, true);
}
return llvm::PointerType::getUnqual(Kmpc_MicroTy);
1560}

1562llvm::FunctionCallee
1563CGOpenMPRuntime::createForStaticInitFunction(unsigned IVSize, bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&((void)0)
       "IV size is not compatible with the omp runtime")((void)0);
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
  CGM.Int32Ty,                               // schedtype
  llvm::PointerType::getUnqual(CGM.Int32Ty), // p_lastiter
  PtrTy,                                     // p_lower
  PtrTy,                                     // p_upper
  PtrTy,                                     // p_stride
  ITy,                                       // incr
  ITy                                        // chunk
};
auto *FnTy =
    llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
1586}

1588llvm::FunctionCallee
1589CGOpenMPRuntime::createDispatchInitFunction(unsigned IVSize, bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&((void)0)
       "IV size is not compatible with the omp runtime")((void)0);
StringRef Name =
    IVSize == 32
        ? (IVSigned ? "__kmpc_dispatch_init_4" : "__kmpc_dispatch_init_4u")
        : (IVSigned ? "__kmpc_dispatch_init_8" : "__kmpc_dispatch_init_8u");
llvm::Type *ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
llvm::Type *TypeParams[] = { getIdentTyPointerTy(), // loc
                             CGM.Int32Ty,           // tid
                             CGM.Int32Ty,           // schedtype
                             ITy,                   // lower
                             ITy,                   // upper
                             ITy,                   // stride
                             ITy                    // chunk
};
auto *FnTy =
    llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
1608}

1610llvm::FunctionCallee
1611CGOpenMPRuntime::createDispatchFiniFunction(unsigned IVSize, bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&((void)0)
       "IV size is not compatible with the omp runtime")((void)0);
StringRef Name =
    IVSize == 32
        ? (IVSigned ? "__kmpc_dispatch_fini_4" : "__kmpc_dispatch_fini_4u")
        : (IVSigned ? "__kmpc_dispatch_fini_8" : "__kmpc_dispatch_fini_8u");
llvm::Type *TypeParams[] = {
    getIdentTyPointerTy(), // loc
    CGM.Int32Ty,           // tid
};
auto *FnTy =
    llvm::FunctionType::get(CGM.VoidTy, TypeParams, /*isVarArg=*/false);
return CGM.CreateRuntimeFunction(FnTy, Name);
1625}

1627llvm::FunctionCallee
1628CGOpenMPRuntime::createDispatchNextFunction(unsigned IVSize, bool IVSigned) {
assert((IVSize == 32 || IVSize == 64) &&((void)0)
       "IV size is not compatible with the omp runtime")((void)0);
StringRef Name =
    IVSize == 32
        ? (IVSigned ? "__kmpc_dispatch_next_4" : "__kmpc_dispatch_next_4u")
        : (IVSigned ? "__kmpc_dispatch_next_8" : "__kmpc_dispatch_next_8u");
llvm::Type *ITy = IVSize == 32 ? CGM.Int32Ty : CGM.Int64Ty;
auto *PtrTy = llvm::PointerType::getUnqual(ITy);
llvm::Type *TypeParams[] = {
  getIdentTyPointerTy(),                     // loc
  CGM.Int32Ty,                               // tid
  llvm::PointerType::getUnqual(CGM.Int32Ty), // p_lastiter
  PtrTy,                                     // p_lower
  PtrTy,                                     // p_upper
  PtrTy                                      // p_stride
};
auto *FnTy =
    llvm::FunctionType::get(CGM.Int32Ty, TypeParams, /*isVarArg*/ false);
return CGM.CreateRuntimeFunction(FnTy, Name);
1648}

1650/// Obtain information that uniquely identifies a target entry. This
1651/// consists of the file and device IDs as well as line number associated with
1652/// the relevant entry source location.
1653static 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.")((void)0);

PresumedLoc PLoc = SM.getPresumedLoc(Loc);
assert(PLoc.isValid() && "Source location is expected to be always valid.")((void)0);

llvm::sys::fs::UniqueID ID;
if (auto EC = llvm::sys::fs::getUniqueID(PLoc.getFilename(), ID)) {
  PLoc = SM.getPresumedLoc(Loc, /*UseLineDirectives=*/false);
  assert(PLoc.isValid() && "Source location is expected to be always valid.")((void)0);
  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();
1678}

1680Address CGOpenMPRuntime::getAddrOfDeclareTargetVar(const VarDecl *VD) {
if (CGM.getLangOpts().OpenMPSimd)
  return Address::invalid();
llvm::Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
    OMPDeclareTargetDeclAttr::isDeclareTargetDeclaration(VD);
if (Res && (*Res == OMPDeclareTargetDeclAttr::MT_Link ||
            (*Res == OMPDeclareTargetDeclAttr::MT_To &&
             HasRequiresUnifiedSharedMemory))) {
  SmallString<64> PtrName;
  {
    llvm::raw_svector_ostream OS(PtrName);
    OS << CGM.getMangledName(GlobalDecl(VD));
    if (!VD->isExternallyVisible()) {
      unsigned DeviceID, FileID, Line;
      getTargetEntryUniqueInfo(CGM.getContext(),
                               VD->getCanonicalDecl()->getBeginLoc(),
                               DeviceID, FileID, Line);
      OS << llvm::format("_%x", FileID);
    }
    OS << "_decl_tgt_ref_ptr";
  }
  llvm::Value *Ptr = CGM.getModule().getNamedValue(PtrName);
  if (!Ptr) {
    QualType PtrTy = CGM.getContext().getPointerType(VD->getType());
    Ptr = getOrCreateInternalVariable(CGM.getTypes().ConvertTypeForMem(PtrTy),
                                      PtrName);

    auto *GV = cast<llvm::GlobalVariable>(Ptr);
    GV->setLinkage(llvm::GlobalValue::WeakAnyLinkage);

    if (!CGM.getLangOpts().OpenMPIsDevice)
      GV->setInitializer(CGM.GetAddrOfGlobal(VD));
    registerTargetGlobalVariable(VD, cast<llvm::Constant>(Ptr));
  }
  return Address(Ptr, CGM.getContext().getDeclAlign(VD));
}
return Address::invalid();
1717}

1719llvm::Constant *
1720CGOpenMPRuntime::getOrCreateThreadPrivateCache(const VarDecl *VD) {
assert(!CGM.getLangOpts().OpenMPUseTLS ||((void)0)
       !CGM.getContext().getTargetInfo().isTLSSupported())((void)0);
// Lookup the entry, lazily creating it if necessary.
std::string Suffix = getName({"cache", ""});
return getOrCreateInternalVariable(
    CGM.Int8PtrPtrTy, Twine(CGM.getMangledName(VD)).concat(Suffix));
1727}

1729Address CGOpenMPRuntime::getAddrOfThreadPrivate(CodeGenFunction &CGF,
                                              const VarDecl *VD,
                                              Address VDAddr,
                                              SourceLocation Loc) {
if (CGM.getLangOpts().OpenMPUseTLS &&
    CGM.getContext().getTargetInfo().isTLSSupported())
  return VDAddr;

llvm::Type *VarTy = VDAddr.getElementType();
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
                       CGF.Builder.CreatePointerCast(VDAddr.getPointer(),
                                                     CGM.Int8PtrTy),
                       CGM.getSize(CGM.GetTargetTypeStoreSize(VarTy)),
                       getOrCreateThreadPrivateCache(VD)};
return Address(CGF.EmitRuntimeCall(
                   OMPBuilder.getOrCreateRuntimeFunction(
                       CGM.getModule(), OMPRTL___kmpc_threadprivate_cached),
                   Args),
               VDAddr.getAlignment());
1748}

1750void CGOpenMPRuntime::emitThreadPrivateVarInit(
  CodeGenFunction &CGF, Address VDAddr, llvm::Value *Ctor,
  llvm::Value *CopyCtor, llvm::Value *Dtor, SourceLocation Loc) {
// Call kmp_int32 __kmpc_global_thread_num(&loc) to init OpenMP runtime
// library.
llvm::Value *OMPLoc = emitUpdateLocation(CGF, Loc);
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), 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};
CGF.EmitRuntimeCall(
    OMPBuilder.getOrCreateRuntimeFunction(
        CGM.getModule(), OMPRTL___kmpc_threadprivate_register),
    Args);
1768}

1770llvm::Function *CGOpenMPRuntime::emitThreadPrivateVarDefinition(
  const VarDecl *VD, Address VDAddr, SourceLocation Loc,
  bool PerformInit, CodeGenFunction *CGF) {
if (CGM.getLangOpts().OpenMPUseTLS &&
    CGM.getContext().getTargetInfo().isTLSSupported())
  return nullptr;

VD = VD->getDefinition(CGM.getContext());
if (VD && ThreadPrivateWithDefinition.insert(CGM.getMangledName(VD)).second) {
  QualType ASTTy = VD->getType();

  llvm::Value *Ctor = nullptr, *CopyCtor = nullptr, *Dtor = nullptr;
  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(), /*DC=*/nullptr, Loc,
                          /*Id=*/nullptr, CGM.getContext().VoidPtrTy,
                          ImplicitParamDecl::Other);
    Args.push_back(&Dst);

    const auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
        CGM.getContext().VoidPtrTy, Args);
    llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
    std::string Name = getName({"__kmpc_global_ctor_", ""});
    llvm::Function *Fn =
        CGM.CreateGlobalInitOrCleanUpFunction(FTy, Name, FI, Loc);
    CtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidPtrTy, Fn, FI,
                          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));
    CtorCGF.EmitAnyExprToMem(Init, Arg, Init->getType().getQualifiers(),
                             /*IsInitializer=*/true);
    ArgVal = CtorCGF.EmitLoadOfScalar(
        CtorCGF.GetAddrOfLocalVar(&Dst), /*Volatile=*/false,
        CGM.getContext().VoidPtrTy, Dst.getLocation());
    CtorCGF.Builder.CreateStore(ArgVal, CtorCGF.ReturnValue);
    CtorCGF.FinishFunction();
    Ctor = Fn;
  }
  if (VD->getType().isDestructedType() != QualType::DK_none) {
    // Generate function that emits destructor call for the threadprivate copy
    // of the variable VD
    CodeGenFunction DtorCGF(CGM);
    FunctionArgList Args;
    ImplicitParamDecl Dst(CGM.getContext(), /*DC=*/nullptr, Loc,
                          /*Id=*/nullptr, CGM.getContext().VoidPtrTy,
                          ImplicitParamDecl::Other);
    Args.push_back(&Dst);

    const auto &FI = CGM.getTypes().arrangeBuiltinFunctionDeclaration(
        CGM.getContext().VoidTy, Args);
    llvm::FunctionType *FTy = CGM.getTypes().GetFunctionType(FI);
    std::string Name = getName({"__kmpc_global_dtor_", ""});
    llvm::Function *Fn =
        CGM.CreateGlobalInitOrCleanUpFunction(FTy, Name, FI, Loc);
    auto NL = ApplyDebugLocation::CreateEmpty(DtorCGF);
    DtorCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, Fn, FI, Args,
                          Loc, Loc);
    // Create a scope with an artificial location for the body of this function.
    auto AL = ApplyDebugLocation::CreateArtificial(DtorCGF);
    llvm::Value *ArgVal = DtorCGF.EmitLoadOfScalar(
        DtorCGF.GetAddrOfLocalVar(&Dst),
        /*Volatile=*/false, CGM.getContext().VoidPtrTy, Dst.getLocation());
    DtorCGF.emitDestroy(Address(ArgVal, VDAddr.getAlignment()), ASTTy,
                        DtorCGF.getDestroyer(ASTTy.isDestructedType()),
                        DtorCGF.needsEHCleanup(ASTTy.isDestructedType()));
    DtorCGF.FinishFunction();
    Dtor = Fn;
  }
  // Do not emit init function if it is not required.
  if (!Ctor && !Dtor)
    return nullptr;

  llvm::Type *CopyCtorTyArgs[] = {CGM.VoidPtrTy, CGM.VoidPtrTy};
  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();
    Ctor = llvm::Constant::getNullValue(CtorTy);
  }
  if (Dtor == nullptr) {
    auto *DtorTy = llvm::FunctionType::get(CGM.VoidTy, CGM.VoidPtrTy,
                                           /*isVarArg=*/false)
                       ->getPointerTo();
    Dtor = llvm::Constant::getNullValue(DtorTy);
  }
  if (!CGF) {
    auto *InitFunctionTy =
        llvm::FunctionType::get(CGM.VoidTy, /*isVarArg*/ false);
    std::string Name = getName({"__omp_threadprivate_init_", ""});
    llvm::Function *InitFunction = CGM.CreateGlobalInitOrCleanUpFunction(
        InitFunctionTy, Name, CGM.getTypes().arrangeNullaryFunction());
    CodeGenFunction InitCGF(CGM);
    FunctionArgList ArgList;
    InitCGF.StartFunction(GlobalDecl(), CGM.getContext().VoidTy, InitFunction,
                          CGM.getTypes().arrangeNullaryFunction(), ArgList,
                          Loc, Loc);
    emitThreadPrivateVarInit(InitCGF, VDAddr, Ctor, CopyCtor, Dtor, Loc);
    InitCGF.FinishFunction();
    return InitFunction;
  }
  emitThreadPrivateVarInit(*CGF, VDAddr, Ctor, CopyCtor, Dtor, Loc);
}
return nullptr;
1888}

1890bool CGOpenMPRuntime::emitDeclareTargetVarDefinition(const VarDecl *VD,
                                                   llvm::GlobalVariable *Addr,
                                                   bool PerformInit) {
if (CGM.getLangOpts().OMPTargetTriples.empty() &&
    !CGM.getLangOpts().OpenMPIsDevice)
  return false;
Optional<OMPDeclareTargetDeclAttr::MapTypeTy> Res =
    OMPDeclareTargetDeclAttr::isDeclareTargetDeclaration(VD);
if (!Res || *Res == OMPDeclareTargetDeclAttr::MT_Link ||
    (*Res == OMPDeclareTargetDeclAttr::MT_To &&
     HasRequiresUnifiedSharedMemory))
  return CGM.getLangOpts().OpenMPIsDevice;
VD = VD->getDefinition(CGM.getContext());
assert(VD && "Unknown VarDecl")((void)0);

if (!DeclareTargetWithDefinition.insert(CGM.getMangledName(VD)).second)
  return CGM.getLangOpts().OpenMPIsDevice;

QualType ASTTy = VD->getType();
SourceLocation Loc = VD->getCanonicalDecl()->getBeginLoc();

// 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.CreateGlobalInitOrCleanUpFunction(
        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.CreateGlobalInitOrCleanUpFunction(
        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;
2003}

2005Address CGOpenMPRuntime::getAddrOfArtificialThreadPrivate(CodeGenFunction &CGF,
                                                        QualType VarType,
                                                        StringRef Name) {
std::string Suffix = getName({"artificial", ""});
llvm::Type *VarLVType = CGF.ConvertTypeForMem(VarType);
llvm::Value *GAddr =
    getOrCreateInternalVariable(VarLVType, Twine(Name).concat(Suffix));
if (CGM.getLangOpts().OpenMP && CGM.getLangOpts().OpenMPUseTLS &&
    CGM.getTarget().isTLSSupported()) {
  cast<llvm::GlobalVariable>(GAddr)->setThreadLocal(/*Val=*/true);
  return Address(GAddr, CGM.getContext().getTypeAlignInChars(VarType));
}
std::string CacheSuffix = getName({"cache", ""});
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, Twine(Name).concat(Suffix).concat(CacheSuffix))};
return Address(
    CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        CGF.EmitRuntimeCall(
            OMPBuilder.getOrCreateRuntimeFunction(
                CGM.getModule(), OMPRTL___kmpc_threadprivate_cached),
            Args),
        VarLVType->getPointerTo(/*AddrSpace=*/0)),
    CGM.getContext().getTypeAlignInChars(VarType));
2034}

2036void CGOpenMPRuntime::emitIfClause(CodeGenFunction &CGF, const Expr *Cond,
                                 const RegionCodeGenTy &ThenGen,
                                 const RegionCodeGenTy &ElseGen) {
CodeGenFunction::LexicalScope ConditionScope(CGF, Cond->getSourceRange());

// If the condition constant folds and can be elided, try to avoid emitting
// the condition and the dead arm of the if/else.
bool CondConstant;
if (CGF.ConstantFoldsToSimpleInteger(Cond, CondConstant)) {
8
←
Assuming the condition is true→
9
←
Taking true branch→
  if (CondConstant)
10
←
Assuming 'CondConstant' is true→
11
←
Taking true branch→
    ThenGen(CGF);
12
←
Calling 'RegionCodeGenTy::operator()'→
  else
    ElseGen(CGF);
  return;
}

// Otherwise, the condition did not fold, or we couldn't elide it.  Just
// emit the conditional branch.
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.
CGF.EmitBlock(ThenBlock);
ThenGen(CGF);
CGF.EmitBranch(ContBlock);
// Emit the 'else' code if present.
// There is no need to emit line number for unconditional branch.
(void)ApplyDebugLocation::CreateEmpty(CGF);
CGF.EmitBlock(ElseBlock);
ElseGen(CGF);
// There is no need to emit line number for unconditional branch.
(void)ApplyDebugLocation::CreateEmpty(CGF);
CGF.EmitBranch(ContBlock);
// Emit the continuation block for code after the if.
CGF.EmitBlock(ContBlock, /*IsFinished=*/true);
2073}

2075void CGOpenMPRuntime::emitParallelCall(CodeGenFunction &CGF, SourceLocation Loc,
                                     llvm::Function *OutlinedFn,
                                     ArrayRef<llvm::Value *> CapturedVars,
                                     const Expr *IfCond) {
if (!CGF.HaveInsertPoint())
  return;
llvm::Value *RTLoc = emitUpdateLocation(CGF, Loc);
auto &M = CGM.getModule();
auto &&ThenGen = [&M, OutlinedFn, CapturedVars, RTLoc,
                  this](CodeGenFunction &CGF, PrePostActionTy &) {
  // Build call __kmpc_fork_call(loc, n, microtask, var1, .., varn);
  CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
  llvm::Value *Args[] = {
      RTLoc,
      CGF.Builder.getInt32(CapturedVars.size()), // Number of captured vars
      CGF.Builder.CreateBitCast(OutlinedFn, RT.getKmpc_MicroPointerTy())};
  llvm::SmallVector<llvm::Value *, 16> RealArgs;
  RealArgs.append(std::begin(Args), std::end(Args));
  RealArgs.append(CapturedVars.begin(), CapturedVars.end());

  llvm::FunctionCallee RTLFn =
      OMPBuilder.getOrCreateRuntimeFunction(M, OMPRTL___kmpc_fork_call);
  CGF.EmitRuntimeCall(RTLFn, RealArgs);
};
auto &&ElseGen = [&M, OutlinedFn, CapturedVars, RTLoc, Loc,
                  this](CodeGenFunction &CGF, PrePostActionTy &) {
  CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
  llvm::Value *ThreadID = RT.getThreadID(CGF, Loc);
  // Build calls:
  // __kmpc_serialized_parallel(&Loc, GTid);
  llvm::Value *Args[] = {RTLoc, ThreadID};
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          M, OMPRTL___kmpc_serialized_parallel),
                      Args);

  // OutlinedFn(&GTid, &zero_bound, CapturedStruct);
  Address ThreadIDAddr = RT.emitThreadIDAddress(CGF, Loc);
  Address ZeroAddrBound =
      CGF.CreateDefaultAlignTempAlloca(CGF.Int32Ty,
                                       /*Name=*/".bound.zero.addr");
  CGF.InitTempAlloca(ZeroAddrBound, CGF.Builder.getInt32(/*C*/ 0));
  llvm::SmallVector<llvm::Value *, 16> OutlinedFnArgs;
  // ThreadId for serialized parallels is 0.
  OutlinedFnArgs.push_back(ThreadIDAddr.getPointer());
  OutlinedFnArgs.push_back(ZeroAddrBound.getPointer());
  OutlinedFnArgs.append(CapturedVars.begin(), CapturedVars.end());

  // Ensure we do not inline the function. This is trivially true for the ones
  // passed to __kmpc_fork_call but the ones called in serialized regions
  // could be inlined. This is not a perfect but it is closer to the invariant
  // we want, namely, every data environment starts with a new function.
  // TODO: We should pass the if condition to the runtime function and do the
  //       handling there. Much cleaner code.
  OutlinedFn->removeFnAttr(llvm::Attribute::AlwaysInline);
  OutlinedFn->addFnAttr(llvm::Attribute::NoInline);
  RT.emitOutlinedFunctionCall(CGF, Loc, OutlinedFn, OutlinedFnArgs);

  // __kmpc_end_serialized_parallel(&Loc, GTid);
  llvm::Value *EndArgs[] = {RT.emitUpdateLocation(CGF, Loc), ThreadID};
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          M, OMPRTL___kmpc_end_serialized_parallel),
                      EndArgs);
};
if (IfCond) {
  emitIfClause(CGF, IfCond, ThenGen, ElseGen);
} else {
  RegionCodeGenTy ThenRCG(ThenGen);
  ThenRCG(CGF);
}
2144}

2146// If we're inside an (outlined) parallel region, use the region info's
2147// thread-ID variable (it is passed in a first argument of the outlined function
2148// as "kmp_int32 *gtid"). Otherwise, if we're not inside parallel region, but in
2149// regular serial code region, get thread ID by calling kmp_int32
2150// kmpc_global_thread_num(ident_t *loc), stash this thread ID in a temporary and
2151// return the address of that temp.
2152Address CGOpenMPRuntime::emitThreadIDAddress(CodeGenFunction &CGF,
                                           SourceLocation Loc) {
if (auto *OMPRegionInfo =
        dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo))
  if (OMPRegionInfo->getThreadIDVariable())
    return OMPRegionInfo->getThreadIDVariableLValue(CGF).getAddress(CGF);

llvm::Value *ThreadID = getThreadID(CGF, Loc);
QualType Int32Ty =
    CGF.getContext().getIntTypeForBitwidth(/*DestWidth*/ 32, /*Signed*/ true);
Address ThreadIDTemp = CGF.CreateMemTemp(Int32Ty, /*Name*/ ".threadid_temp.");
CGF.EmitStoreOfScalar(ThreadID,
                      CGF.MakeAddrLValue(ThreadIDTemp, Int32Ty));

return ThreadIDTemp;
2167}

2169llvm::Constant *CGOpenMPRuntime::getOrCreateInternalVariable(
  llvm::Type *Ty, const llvm::Twine &Name, unsigned AddressSpace) {
SmallString<256> Buffer;
llvm::raw_svector_ostream Out(Buffer);
Out << Name;
StringRef RuntimeName = Out.str();
auto &Elem = *InternalVars.try_emplace(RuntimeName, nullptr).first;
if (Elem.second) {
  assert(Elem.second->getType()->getPointerElementType() == Ty &&((void)0)
         "OMP internal variable has different type than requested")((void)0);
  return &*Elem.second;
}

return Elem.second = new llvm::GlobalVariable(
           CGM.getModule(), Ty, /*IsConstant*/ false,
           llvm::GlobalValue::CommonLinkage, llvm::Constant::getNullValue(Ty),
           Elem.first(), /*InsertBefore=*/nullptr,
           llvm::GlobalValue::NotThreadLocal, AddressSpace);
2187}

2189llvm::Value *CGOpenMPRuntime::getCriticalRegionLock(StringRef CriticalName) {
std::string Prefix = Twine("gomp_critical_user_", CriticalName).str();
std::string Name = getName({Prefix, "var"});
return getOrCreateInternalVariable(KmpCriticalNameTy, Name);
2193}

2195namespace {
2196/// Common pre(post)-action for different OpenMP constructs.
2197class CommonActionTy final : public PrePostActionTy {
llvm::FunctionCallee EnterCallee;
ArrayRef<llvm::Value *> EnterArgs;
llvm::FunctionCallee ExitCallee;
ArrayRef<llvm::Value *> ExitArgs;
bool Conditional;
llvm::BasicBlock *ContBlock = nullptr;

2205public:
CommonActionTy(llvm::FunctionCallee EnterCallee,
               ArrayRef<llvm::Value *> EnterArgs,
               llvm::FunctionCallee ExitCallee,
               ArrayRef<llvm::Value *> ExitArgs, bool Conditional = false)
    : EnterCallee(EnterCallee), EnterArgs(EnterArgs), ExitCallee(ExitCallee),
      ExitArgs(ExitArgs), Conditional(Conditional) {}
void Enter(CodeGenFunction &CGF) override {
  llvm::Value *EnterRes = CGF.EmitRuntimeCall(EnterCallee, EnterArgs);
  if (Conditional) {
    llvm::Value *CallBool = CGF.Builder.CreateIsNotNull(EnterRes);
    auto *ThenBlock = CGF.createBasicBlock("omp_if.then");
    ContBlock = CGF.createBasicBlock("omp_if.end");
    // Generate the branch (If-stmt)
    CGF.Builder.CreateCondBr(CallBool, ThenBlock, ContBlock);
    CGF.EmitBlock(ThenBlock);
  }
}
void Done(CodeGenFunction &CGF) {
  // Emit the rest of blocks/branches
  CGF.EmitBranch(ContBlock);
  CGF.EmitBlock(ContBlock, true);
}
void Exit(CodeGenFunction &CGF) override {
  CGF.EmitRuntimeCall(ExitCallee, ExitArgs);
}
2231};
2232} // anonymous namespace

2234void CGOpenMPRuntime::emitCriticalRegion(CodeGenFunction &CGF,
                                       StringRef CriticalName,
                                       const RegionCodeGenTy &CriticalOpGen,
                                       SourceLocation Loc, const Expr *Hint) {
// __kmpc_critical[_with_hint](ident_t *, gtid, Lock[, hint]);
// CriticalOpGen();
// __kmpc_end_critical(ident_t *, gtid, Lock);
// Prepare arguments and build a call to __kmpc_critical
if (!CGF.HaveInsertPoint())
  return;
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
                       getCriticalRegionLock(CriticalName)};
llvm::SmallVector<llvm::Value *, 4> EnterArgs(std::begin(Args),
                                              std::end(Args));
if (Hint) {
  EnterArgs.push_back(CGF.Builder.CreateIntCast(
      CGF.EmitScalarExpr(Hint), CGM.Int32Ty, /*isSigned=*/false));
}
CommonActionTy Action(
    OMPBuilder.getOrCreateRuntimeFunction(
        CGM.getModule(),
        Hint ? OMPRTL___kmpc_critical_with_hint : OMPRTL___kmpc_critical),
    EnterArgs,
    OMPBuilder.getOrCreateRuntimeFunction(CGM.getModule(),
                                          OMPRTL___kmpc_end_critical),
    Args);
CriticalOpGen.setAction(Action);
emitInlinedDirective(CGF, OMPD_critical, CriticalOpGen);
2262}

2264void CGOpenMPRuntime::emitMasterRegion(CodeGenFunction &CGF,
                                     const RegionCodeGenTy &MasterOpGen,
                                     SourceLocation Loc) {
if (!CGF.HaveInsertPoint())
  return;
// if(__kmpc_master(ident_t *, gtid)) {
//   MasterOpGen();
//   __kmpc_end_master(ident_t *, gtid);
// }
// Prepare arguments and build a call to __kmpc_master
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_master),
                      Args,
                      OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_end_master),
                      Args,
                      /*Conditional=*/true);
MasterOpGen.setAction(Action);
emitInlinedDirective(CGF, OMPD_master, MasterOpGen);
Action.Done(CGF);
2285}

2287void CGOpenMPRuntime::emitMaskedRegion(CodeGenFunction &CGF,
                                     const RegionCodeGenTy &MaskedOpGen,
                                     SourceLocation Loc, const Expr *Filter) {
if (!CGF.HaveInsertPoint())
  return;
// if(__kmpc_masked(ident_t *, gtid, filter)) {
//   MaskedOpGen();
//   __kmpc_end_masked(iden_t *, gtid);
// }
// Prepare arguments and build a call to __kmpc_masked
llvm::Value *FilterVal = Filter
                             ? CGF.EmitScalarExpr(Filter, CGF.Int32Ty)
                             : llvm::ConstantInt::get(CGM.Int32Ty, /*V=*/0);
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
                       FilterVal};
llvm::Value *ArgsEnd[] = {emitUpdateLocation(CGF, Loc),
                          getThreadID(CGF, Loc)};
CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_masked),
                      Args,
                      OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_end_masked),
                      ArgsEnd,
                      /*Conditional=*/true);
MaskedOpGen.setAction(Action);
emitInlinedDirective(CGF, OMPD_masked, MaskedOpGen);
Action.Done(CGF);
2314}

2316void CGOpenMPRuntime::emitTaskyieldCall(CodeGenFunction &CGF,
                                      SourceLocation Loc) {
if (!CGF.HaveInsertPoint())
  return;
if (CGF.CGM.getLangOpts().OpenMPIRBuilder) {
  OMPBuilder.createTaskyield(CGF.Builder);
} else {
  // Build call __kmpc_omp_taskyield(loc, thread_id, 0);
  llvm::Value *Args[] = {
      emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
      llvm::ConstantInt::get(CGM.IntTy, /*V=*/0, /*isSigned=*/true)};
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_omp_taskyield),
                      Args);
}

if (auto *Region = dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo))
  Region->emitUntiedSwitch(CGF);
2334}

2336void CGOpenMPRuntime::emitTaskgroupRegion(CodeGenFunction &CGF,
                                        const RegionCodeGenTy &TaskgroupOpGen,
                                        SourceLocation Loc) {
if (!CGF.HaveInsertPoint())
  return;
// __kmpc_taskgroup(ident_t *, gtid);
// TaskgroupOpGen();
// __kmpc_end_taskgroup(ident_t *, gtid);
// Prepare arguments and build a call to __kmpc_taskgroup
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_taskgroup),
                      Args,
                      OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_end_taskgroup),
                      Args);
TaskgroupOpGen.setAction(Action);
emitInlinedDirective(CGF, OMPD_taskgroup, TaskgroupOpGen);
2354}

2356/// Given an array of pointers to variables, project the address of a
2357/// given variable.
2358static Address emitAddrOfVarFromArray(CodeGenFunction &CGF, Address Array,
                                    unsigned Index, const VarDecl *Var) {
// Pull out the pointer to the variable.
Address PtrAddr = CGF.Builder.CreateConstArrayGEP(Array, Index);
llvm::Value *Ptr = CGF.Builder.CreateLoad(PtrAddr);

Address Addr = Address(Ptr, CGF.getContext().getDeclAlign(Var));
Addr = CGF.Builder.CreateElementBitCast(
    Addr, CGF.ConvertTypeForMem(Var->getType()));
return Addr;
2368}

2370static llvm::Value *emitCopyprivateCopyFunction(
  CodeGenModule &CGM, llvm::Type *ArgsType,
  ArrayRef<const Expr *> CopyprivateVars, ArrayRef<const Expr *> DestExprs,
  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, /*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);
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, Loc, Loc);
// Dest = (void*[n])(LHSArg);
// Src = (void*[n])(RHSArg);
Address LHS(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    CGF.Builder.CreateLoad(CGF.GetAddrOfLocalVar(&LHSArg)),
    ArgsType), CGF.getPointerAlign());
Address RHS(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    CGF.Builder.CreateLoad(CGF.GetAddrOfLocalVar(&RHSArg)),
    ArgsType), CGF.getPointerAlign());
// *(Type0*)Dst[0] = *(Type0*)Src[0];
// *(Type1*)Dst[1] = *(Type1*)Src[1];
// ...
// *(Typen*)Dst[n] = *(Typen*)Src[n];
for (unsigned I = 0, E = AssignmentOps.size(); I < E; ++I) {
  const auto *DestVar =
      cast<VarDecl>(cast<DeclRefExpr>(DestExprs[I])->getDecl());
  Address DestAddr = emitAddrOfVarFromArray(CGF, LHS, I, DestVar);

  const auto *SrcVar =
      cast<VarDecl>(cast<DeclRefExpr>(SrcExprs[I])->getDecl());
  Address SrcAddr = emitAddrOfVarFromArray(CGF, RHS, I, SrcVar);

  const auto *VD = cast<DeclRefExpr>(CopyprivateVars[I])->getDecl();
  QualType Type = VD->getType();
  CGF.EmitOMPCopy(Type, DestAddr, SrcAddr, DestVar, SrcVar, AssignmentOps[I]);
}
CGF.FinishFunction();
return Fn;
2422}

2424void CGOpenMPRuntime::emitSingleRegion(CodeGenFunction &CGF,
                                     const RegionCodeGenTy &SingleOpGen,
                                     SourceLocation Loc,
                                     ArrayRef<const Expr *> CopyprivateVars,
                                     ArrayRef<const Expr *> SrcExprs,
                                     ArrayRef<const Expr *> DstExprs,
                                     ArrayRef<const Expr *> AssignmentOps) {
if (!CGF.HaveInsertPoint())
  return;
assert(CopyprivateVars.size() == SrcExprs.size() &&((void)0)
       CopyprivateVars.size() == DstExprs.size() &&((void)0)
       CopyprivateVars.size() == AssignmentOps.size())((void)0);
ASTContext &C = CGM.getContext();
// int32 did_it = 0;
// if(__kmpc_single(ident_t *, gtid)) {
//   SingleOpGen();
//   __kmpc_end_single(ident_t *, gtid);
//   did_it = 1;
// }
// call __kmpc_copyprivate(ident_t *, gtid, <buf_size>, <copyprivate list>,
// <copy_func>, did_it);

Address DidIt = Address::invalid();
if (!CopyprivateVars.empty()) {
  // int32 did_it = 0;
  QualType KmpInt32Ty =
      C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/1);
  DidIt = CGF.CreateMemTemp(KmpInt32Ty, ".omp.copyprivate.did_it");
  CGF.Builder.CreateStore(CGF.Builder.getInt32(0), DidIt);
}
// Prepare arguments and build a call to __kmpc_single
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_single),
                      Args,
                      OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_end_single),
                      Args,
                      /*Conditional=*/true);
SingleOpGen.setAction(Action);
emitInlinedDirective(CGF, OMPD_single, SingleOpGen);
if (DidIt.isValid()) {
  // did_it = 1;
  CGF.Builder.CreateStore(CGF.Builder.getInt32(1), DidIt);
}
Action.Done(CGF);
// call __kmpc_copyprivate(ident_t *, gtid, <buf_size>, <copyprivate list>,
// <copy_func>, did_it);
if (DidIt.isValid()) {
  llvm::APInt ArraySize(/*unsigned int numBits=*/32, CopyprivateVars.size());
  QualType CopyprivateArrayTy = C.getConstantArrayType(
      C.VoidPtrTy, ArraySize, nullptr, ArrayType::Normal,
      /*IndexTypeQuals=*/0);
  // Create a list of all private variables for copyprivate.
  Address CopyprivateList =
      CGF.CreateMemTemp(CopyprivateArrayTy, ".omp.copyprivate.cpr_list");
  for (unsigned I = 0, E = CopyprivateVars.size(); I < E; ++I) {
    Address Elem = CGF.Builder.CreateConstArrayGEP(CopyprivateList, I);
    CGF.Builder.CreateStore(
        CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
            CGF.EmitLValue(CopyprivateVars[I]).getPointer(CGF),
            CGF.VoidPtrTy),
        Elem);
  }
  // Build function that copies private values from single region to all other
  // threads in the corresponding parallel region.
  llvm::Value *CpyFn = emitCopyprivateCopyFunction(
      CGM, CGF.ConvertTypeForMem(CopyprivateArrayTy)->getPointerTo(),
      CopyprivateVars, SrcExprs, DstExprs, AssignmentOps, Loc);
  llvm::Value *BufSize = CGF.getTypeSize(CopyprivateArrayTy);
  Address CL =
    CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(CopyprivateList,
                                                    CGF.VoidPtrTy);
  llvm::Value *DidItVal = CGF.Builder.CreateLoad(DidIt);
  llvm::Value *Args[] = {
      emitUpdateLocation(CGF, Loc), // ident_t *<loc>
      getThreadID(CGF, Loc),        // i32 <gtid>
      BufSize,                      // size_t <buf_size>
      CL.getPointer(),              // void *<copyprivate list>
      CpyFn,                        // void (*) (void *, void *) <copy_func>
      DidItVal                      // i32 did_it
  };
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_copyprivate),
                      Args);
}
2510}

2512void CGOpenMPRuntime::emitOrderedRegion(CodeGenFunction &CGF,
                                      const RegionCodeGenTy &OrderedOpGen,
                                      SourceLocation Loc, bool IsThreads) {
if (!CGF.HaveInsertPoint())
  return;
// __kmpc_ordered(ident_t *, gtid);
// OrderedOpGen();
// __kmpc_end_ordered(ident_t *, gtid);
// Prepare arguments and build a call to __kmpc_ordered
if (IsThreads) {
  llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
  CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                            CGM.getModule(), OMPRTL___kmpc_ordered),
                        Args,
                        OMPBuilder.getOrCreateRuntimeFunction(
                            CGM.getModule(), OMPRTL___kmpc_end_ordered),
                        Args);
  OrderedOpGen.setAction(Action);
  emitInlinedDirective(CGF, OMPD_ordered, OrderedOpGen);
  return;
}
emitInlinedDirective(CGF, OMPD_ordered, OrderedOpGen);
2534}

2536unsigned CGOpenMPRuntime::getDefaultFlagsForBarriers(OpenMPDirectiveKind Kind) {
unsigned Flags;
if (Kind == OMPD_for)
  Flags = OMP_IDENT_BARRIER_IMPL_FOR;
else if (Kind == OMPD_sections)
  Flags = OMP_IDENT_BARRIER_IMPL_SECTIONS;
else if (Kind == OMPD_single)
  Flags = OMP_IDENT_BARRIER_IMPL_SINGLE;
else if (Kind == OMPD_barrier)
  Flags = OMP_IDENT_BARRIER_EXPL;
else
  Flags = OMP_IDENT_BARRIER_IMPL;
return Flags;
2549}

2551void CGOpenMPRuntime::getDefaultScheduleAndChunk(
  CodeGenFunction &CGF, const OMPLoopDirective &S,
  OpenMPScheduleClauseKind &ScheduleKind, const Expr *&ChunkExpr) const {
// Check if the loop directive is actually a doacross loop directive. In this
// case choose static, 1 schedule.
if (llvm::any_of(
        S.getClausesOfKind<OMPOrderedClause>(),
        [](const OMPOrderedClause *C) { return C->getNumForLoops(); })) {
  ScheduleKind = OMPC_SCHEDULE_static;
  // Chunk size is 1 in this case.
  llvm::APInt ChunkSize(32, 1);
  ChunkExpr = IntegerLiteral::Create(
      CGF.getContext(), ChunkSize,
      CGF.getContext().getIntTypeForBitwidth(32, /*Signed=*/0),
      SourceLocation());
}
2567}

2569void CGOpenMPRuntime::emitBarrierCall(CodeGenFunction &CGF, SourceLocation Loc,
                                    OpenMPDirectiveKind Kind, bool EmitChecks,
                                    bool ForceSimpleCall) {
// Check if we should use the OMPBuilder
auto *OMPRegionInfo =
    dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo);
if (CGF.CGM.getLangOpts().OpenMPIRBuilder) {
  CGF.Builder.restoreIP(OMPBuilder.createBarrier(
      CGF.Builder, Kind, ForceSimpleCall, EmitChecks));
  return;
}

if (!CGF.HaveInsertPoint())
  return;
// Build call __kmpc_cancel_barrier(loc, thread_id);
// Build call __kmpc_barrier(loc, thread_id);
unsigned Flags = getDefaultFlagsForBarriers(Kind);
// Build call __kmpc_cancel_barrier(loc, thread_id) or __kmpc_barrier(loc,
// thread_id);
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc, Flags),
                       getThreadID(CGF, Loc)};
if (OMPRegionInfo) {
  if (!ForceSimpleCall && OMPRegionInfo->hasCancel()) {
    llvm::Value *Result = CGF.EmitRuntimeCall(
        OMPBuilder.getOrCreateRuntimeFunction(CGM.getModule(),
                                              OMPRTL___kmpc_cancel_barrier),
        Args);
    if (EmitChecks) {
      // if (__kmpc_cancel_barrier()) {
      //   exit from construct;
      // }
      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;
      CodeGenFunction::JumpDest CancelDestination =
          CGF.getOMPCancelDestination(OMPRegionInfo->getDirectiveKind());
      CGF.EmitBranchThroughCleanup(CancelDestination);
      CGF.EmitBlock(ContBB, /*IsFinished=*/true);
    }
    return;
  }
}
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), OMPRTL___kmpc_barrier),
                    Args);
2617}

2619/// Map the OpenMP loop schedule to the runtime enumeration.
2620static OpenMPSchedType getRuntimeSchedule(OpenMPScheduleClauseKind ScheduleKind,
                                        bool Chunked, bool Ordered) {
switch (ScheduleKind) {
case OMPC_SCHEDULE_static:
  return Chunked ? (Ordered ? OMP_ord_static_chunked : OMP_sch_static_chunked)
                 : (Ordered ? OMP_ord_static : OMP_sch_static);
case OMPC_SCHEDULE_dynamic:
  return Ordered ? OMP_ord_dynamic_chunked : OMP_sch_dynamic_chunked;
case OMPC_SCHEDULE_guided:
  return Ordered ? OMP_ord_guided_chunked : OMP_sch_guided_chunked;
case OMPC_SCHEDULE_runtime:
  return Ordered ? OMP_ord_runtime : OMP_sch_runtime;
case OMPC_SCHEDULE_auto:
  return Ordered ? OMP_ord_auto : OMP_sch_auto;
case OMPC_SCHEDULE_unknown:
  assert(!Chunked && "chunk was specified but schedule kind not known")((void)0);
  return Ordered ? OMP_ord_static : OMP_sch_static;
}
llvm_unreachable("Unexpected runtime schedule")__builtin_unreachable();
2639}

2641/// Map the OpenMP distribute schedule to the runtime enumeration.
2642static OpenMPSchedType
2643getRuntimeSchedule(OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) {
// only static is allowed for dist_schedule
return Chunked ? OMP_dist_sch_static_chunked : OMP_dist_sch_static;
2646}

2648bool CGOpenMPRuntime::isStaticNonchunked(OpenMPScheduleClauseKind ScheduleKind,
                                       bool Chunked) const {
OpenMPSchedType Schedule =
    getRuntimeSchedule(ScheduleKind, Chunked, /*Ordered=*/false);
return Schedule == OMP_sch_static;
2653}

2655bool CGOpenMPRuntime::isStaticNonchunked(
  OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) const {
OpenMPSchedType Schedule = getRuntimeSchedule(ScheduleKind, Chunked);
return Schedule == OMP_dist_sch_static;
2659}

2661bool CGOpenMPRuntime::isStaticChunked(OpenMPScheduleClauseKind ScheduleKind,
                                    bool Chunked) const {
OpenMPSchedType Schedule =
    getRuntimeSchedule(ScheduleKind, Chunked, /*Ordered=*/false);
return Schedule == OMP_sch_static_chunked;
2666}

2668bool CGOpenMPRuntime::isStaticChunked(
  OpenMPDistScheduleClauseKind ScheduleKind, bool Chunked) const {
OpenMPSchedType Schedule = getRuntimeSchedule(ScheduleKind, Chunked);
return Schedule == OMP_dist_sch_static_chunked;
2672}

2674bool CGOpenMPRuntime::isDynamic(OpenMPScheduleClauseKind ScheduleKind) const {
OpenMPSchedType Schedule =
    getRuntimeSchedule(ScheduleKind, /*Chunked=*/false, /*Ordered=*/false);
assert(Schedule != OMP_sch_static_chunked && "cannot be chunked here")((void)0);
return Schedule != OMP_sch_static;
2679}

2681static int addMonoNonMonoModifier(CodeGenModule &CGM, OpenMPSchedType Schedule,
                                OpenMPScheduleClauseModifier M1,
                                OpenMPScheduleClauseModifier M2) {
int Modifier = 0;
switch (M1) {
case OMPC_SCHEDULE_MODIFIER_monotonic:
  Modifier = OMP_sch_modifier_monotonic;
  break;
case OMPC_SCHEDULE_MODIFIER_nonmonotonic:
  Modifier = OMP_sch_modifier_nonmonotonic;
  break;
case OMPC_SCHEDULE_MODIFIER_simd:
  if (Schedule == OMP_sch_static_chunked)
    Schedule = OMP_sch_static_balanced_chunked;
  break;
case OMPC_SCHEDULE_MODIFIER_last:
case OMPC_SCHEDULE_MODIFIER_unknown:
  break;
}
switch (M2) {
case OMPC_SCHEDULE_MODIFIER_monotonic:
  Modifier = OMP_sch_modifier_monotonic;
  break;
case OMPC_SCHEDULE_MODIFIER_nonmonotonic:
  Modifier = OMP_sch_modifier_nonmonotonic;
  break;
case OMPC_SCHEDULE_MODIFIER_simd:
  if (Schedule == OMP_sch_static_chunked)
    Schedule = OMP_sch_static_balanced_chunked;
  break;
case OMPC_SCHEDULE_MODIFIER_last:
case OMPC_SCHEDULE_MODIFIER_unknown:
  break;
}
// OpenMP 5.0, 2.9.2 Worksharing-Loop Construct, Desription.
// If the static schedule kind is specified or if the ordered clause is
// specified, and if the nonmonotonic modifier is not specified, the effect is
// as if the monotonic modifier is specified. Otherwise, unless the monotonic
// modifier is specified, the effect is as if the nonmonotonic modifier is
// specified.
if (CGM.getLangOpts().OpenMP >= 50 && Modifier == 0) {
  if (!(Schedule == OMP_sch_static_chunked || Schedule == OMP_sch_static ||
        Schedule == OMP_sch_static_balanced_chunked ||
        Schedule == OMP_ord_static_chunked || Schedule == OMP_ord_static ||
        Schedule == OMP_dist_sch_static_chunked ||
        Schedule == OMP_dist_sch_static))
    Modifier = OMP_sch_modifier_nonmonotonic;
}
return Schedule | Modifier;
2730}

2732void CGOpenMPRuntime::emitForDispatchInit(
  CodeGenFunction &CGF, SourceLocation Loc,
  const OpenMPScheduleTy &ScheduleKind, unsigned IVSize, bool IVSigned,
  bool Ordered, const DispatchRTInput &DispatchValues) {
if (!CGF.HaveInsertPoint())
  return;
OpenMPSchedType Schedule = getRuntimeSchedule(
    ScheduleKind.Schedule, DispatchValues.Chunk != nullptr, Ordered);
assert(Ordered ||((void)0)
       (Schedule != OMP_sch_static && Schedule != OMP_sch_static_chunked &&((void)0)
        Schedule != OMP_ord_static && Schedule != OMP_ord_static_chunked &&((void)0)
        Schedule != OMP_sch_static_balanced_chunked))((void)0);
// Call __kmpc_dispatch_init(
//          ident_t *loc, kmp_int32 tid, kmp_int32 schedule,
//          kmp_int[32|64] lower, kmp_int[32|64] upper,
//          kmp_int[32|64] stride, kmp_int[32|64] chunk);

// If the Chunk was not specified in the clause - use default value 1.
llvm::Value *Chunk = DispatchValues.Chunk ? DispatchValues.Chunk
                                          : CGF.Builder.getIntN(IVSize, 1);
llvm::Value *Args[] = {
    emitUpdateLocation(CGF, Loc),
    getThreadID(CGF, Loc),
    CGF.Builder.getInt32(addMonoNonMonoModifier(
        CGM, Schedule, ScheduleKind.M1, ScheduleKind.M2)), // Schedule type
    DispatchValues.LB,                                     // Lower
    DispatchValues.UB,                                     // Upper
    CGF.Builder.getIntN(IVSize, 1),                        // Stride
    Chunk                                                  // Chunk
};
CGF.EmitRuntimeCall(createDispatchInitFunction(IVSize, IVSigned), Args);
2763}

2765static void emitForStaticInitCall(
  CodeGenFunction &CGF, llvm::Value *UpdateLocation, llvm::Value *ThreadId,
  llvm::FunctionCallee ForStaticInitFunction, OpenMPSchedType Schedule,
  OpenMPScheduleClauseModifier M1, OpenMPScheduleClauseModifier M2,
  const CGOpenMPRuntime::StaticRTInput &Values) {
if (!CGF.HaveInsertPoint())
  return;

assert(!Values.Ordered)((void)0);
assert(Schedule == OMP_sch_static || Schedule == OMP_sch_static_chunked ||((void)0)
       Schedule == OMP_sch_static_balanced_chunked ||((void)0)
       Schedule == OMP_ord_static || Schedule == OMP_ord_static_chunked ||((void)0)
       Schedule == OMP_dist_sch_static ||((void)0)
       Schedule == OMP_dist_sch_static_chunked)((void)0);

// Call __kmpc_for_static_init(
//          ident_t *loc, kmp_int32 tid, kmp_int32 schedtype,
//          kmp_int32 *p_lastiter, kmp_int[32|64] *p_lower,
//          kmp_int[32|64] *p_upper, kmp_int[32|64] *p_stride,
//          kmp_int[32|64] incr, kmp_int[32|64] chunk);
llvm::Value *Chunk = Values.Chunk;
if (Chunk == nullptr) {
  assert((Schedule == OMP_sch_static || Schedule == OMP_ord_static ||((void)0)
          Schedule == OMP_dist_sch_static) &&((void)0)
         "expected static non-chunked schedule")((void)0);
  // If the Chunk was not specified in the clause - use default value 1.
  Chunk = CGF.Builder.getIntN(Values.IVSize, 1);
} else {
  assert((Schedule == OMP_sch_static_chunked ||((void)0)
          Schedule == OMP_sch_static_balanced_chunked ||((void)0)
          Schedule == OMP_ord_static_chunked ||((void)0)
          Schedule == OMP_dist_sch_static_chunked) &&((void)0)
         "expected static chunked schedule")((void)0);
}
llvm::Value *Args[] = {
    UpdateLocation,
    ThreadId,
    CGF.Builder.getInt32(addMonoNonMonoModifier(CGF.CGM, Schedule, M1,
                                                M2)), // Schedule type
    Values.IL.getPointer(),                           // &isLastIter
    Values.LB.getPointer(),                           // &LB
    Values.UB.getPointer(),                           // &UB
    Values.ST.getPointer(),                           // &Stride
    CGF.Builder.getIntN(Values.IVSize, 1),            // Incr
    Chunk                                             // Chunk
};
CGF.EmitRuntimeCall(ForStaticInitFunction, Args);
2812}

2814void CGOpenMPRuntime::emitForStaticInit(CodeGenFunction &CGF,
                                      SourceLocation Loc,
                                      OpenMPDirectiveKind DKind,
                                      const OpenMPScheduleTy &ScheduleKind,
                                      const StaticRTInput &Values) {
OpenMPSchedType ScheduleNum = getRuntimeSchedule(
    ScheduleKind.Schedule, Values.Chunk != nullptr, Values.Ordered);
assert(isOpenMPWorksharingDirective(DKind) &&((void)0)
       "Expected loop-based or sections-based directive.")((void)0);
llvm::Value *UpdatedLocation = emitUpdateLocation(CGF, Loc,
                                           isOpenMPLoopDirective(DKind)
                                               ? OMP_IDENT_WORK_LOOP
                                               : OMP_IDENT_WORK_SECTIONS);
llvm::Value *ThreadId = getThreadID(CGF, Loc);
llvm::FunctionCallee StaticInitFunction =
    createForStaticInitFunction(Values.IVSize, Values.IVSigned);
auto DL = ApplyDebugLocation::CreateDefaultArtificial(CGF, Loc);
emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
                      ScheduleNum, ScheduleKind.M1, ScheduleKind.M2, Values);
2833}

2835void CGOpenMPRuntime::emitDistributeStaticInit(
  CodeGenFunction &CGF, SourceLocation Loc,
  OpenMPDistScheduleClauseKind SchedKind,
  const CGOpenMPRuntime::StaticRTInput &Values) {
OpenMPSchedType ScheduleNum =
    getRuntimeSchedule(SchedKind, Values.Chunk != nullptr);
llvm::Value *UpdatedLocation =
    emitUpdateLocation(CGF, Loc, OMP_IDENT_WORK_DISTRIBUTE);
llvm::Value *ThreadId = getThreadID(CGF, Loc);
llvm::FunctionCallee StaticInitFunction =
    createForStaticInitFunction(Values.IVSize, Values.IVSigned);
emitForStaticInitCall(CGF, UpdatedLocation, ThreadId, StaticInitFunction,
                      ScheduleNum, OMPC_SCHEDULE_MODIFIER_unknown,
                      OMPC_SCHEDULE_MODIFIER_unknown, Values);
2849}

2851void CGOpenMPRuntime::emitForStaticFinish(CodeGenFunction &CGF,
                                        SourceLocation Loc,
                                        OpenMPDirectiveKind DKind) {
if (!CGF.HaveInsertPoint())
  return;
// Call __kmpc_for_static_fini(ident_t *loc, kmp_int32 tid);
llvm::Value *Args[] = {
    emitUpdateLocation(CGF, Loc,
                       isOpenMPDistributeDirective(DKind)
                           ? OMP_IDENT_WORK_DISTRIBUTE
                           : isOpenMPLoopDirective(DKind)
                                 ? OMP_IDENT_WORK_LOOP
                                 : OMP_IDENT_WORK_SECTIONS),
    getThreadID(CGF, Loc)};
auto DL = ApplyDebugLocation::CreateDefaultArtificial(CGF, Loc);
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), OMPRTL___kmpc_for_static_fini),
                    Args);
2869}

2871void CGOpenMPRuntime::emitForOrderedIterationEnd(CodeGenFunction &CGF,
                                               SourceLocation Loc,
                                               unsigned IVSize,
                                               bool IVSigned) {
if (!CGF.HaveInsertPoint())
  return;
// Call __kmpc_for_dynamic_fini_(4|8)[u](ident_t *loc, kmp_int32 tid);
llvm::Value *Args[] = {emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc)};
CGF.EmitRuntimeCall(createDispatchFiniFunction(IVSize, IVSigned), Args);
2880}

2882llvm::Value *CGOpenMPRuntime::emitForNext(CodeGenFunction &CGF,
                                        SourceLocation Loc, unsigned IVSize,
                                        bool IVSigned, Address IL,
                                        Address LB, Address UB,
                                        Address ST) {
// Call __kmpc_dispatch_next(
//          ident_t *loc, kmp_int32 tid, kmp_int32 *p_lastiter,
//          kmp_int[32|64] *p_lower, kmp_int[32|64] *p_upper,
//          kmp_int[32|64] *p_stride);
llvm::Value *Args[] = {
    emitUpdateLocation(CGF, Loc),
    getThreadID(CGF, Loc),
    IL.getPointer(), // &isLastIter
    LB.getPointer(), // &Lower
    UB.getPointer(), // &Upper
    ST.getPointer()  // &Stride
};
llvm::Value *Call =
    CGF.EmitRuntimeCall(createDispatchNextFunction(IVSize, IVSigned), Args);
return CGF.EmitScalarConversion(
    Call, CGF.getContext().getIntTypeForBitwidth(32, /*Signed=*/1),
    CGF.getContext().BoolTy, Loc);
2904}

2906void CGOpenMPRuntime::emitNumThreadsClause(CodeGenFunction &CGF,
                                         llvm::Value *NumThreads,
                                         SourceLocation Loc) {
if (!CGF.HaveInsertPoint())
  return;
// Build call __kmpc_push_num_threads(&loc, global_tid, num_threads)
llvm::Value *Args[] = {
    emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
    CGF.Builder.CreateIntCast(NumThreads, CGF.Int32Ty, /*isSigned*/ true)};
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), OMPRTL___kmpc_push_num_threads),
                    Args);
2918}

2920void CGOpenMPRuntime::emitProcBindClause(CodeGenFunction &CGF,
                                       ProcBindKind ProcBind,
                                       SourceLocation Loc) {
if (!CGF.HaveInsertPoint())
  return;
assert(ProcBind != OMP_PROC_BIND_unknown && "Unsupported proc_bind value.")((void)0);
// Build call __kmpc_push_proc_bind(&loc, global_tid, proc_bind)
llvm::Value *Args[] = {
    emitUpdateLocation(CGF, Loc), getThreadID(CGF, Loc),
    llvm::ConstantInt::get(CGM.IntTy, unsigned(ProcBind), /*isSigned=*/true)};
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), OMPRTL___kmpc_push_proc_bind),
                    Args);
2933}

2935void CGOpenMPRuntime::emitFlush(CodeGenFunction &CGF, ArrayRef<const Expr *>,
                              SourceLocation Loc, llvm::AtomicOrdering AO) {
if (CGF.CGM.getLangOpts().OpenMPIRBuilder) {
  OMPBuilder.createFlush(CGF.Builder);
} else {
  if (!CGF.HaveInsertPoint())
    return;
  // Build call void __kmpc_flush(ident_t *loc)
  CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                          CGM.getModule(), OMPRTL___kmpc_flush),
                      emitUpdateLocation(CGF, Loc));
}
2947}

2949namespace {
2950/// Indexes of fields for type kmp_task_t.
2951enum KmpTaskTFields {
/// List of shared variables.
KmpTaskTShareds,
/// Task routine.
KmpTaskTRoutine,
/// Partition id for the untied tasks.
KmpTaskTPartId,
/// Function with call of destructors for private variables.
Data1,
/// Task priority.
Data2,
/// (Taskloops only) Lower bound.
KmpTaskTLowerBound,
/// (Taskloops only) Upper bound.
KmpTaskTUpperBound,
/// (Taskloops only) Stride.
KmpTaskTStride,
/// (Taskloops only) Is last iteration flag.
KmpTaskTLastIter,
/// (Taskloops only) Reduction data.
KmpTaskTReductions,
2972};
2973} // anonymous namespace

2975bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::empty() const {
return OffloadEntriesTargetRegion.empty() &&
       OffloadEntriesDeviceGlobalVar.empty();
2978}

2980/// Initialize target region entry.
2981void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
  initializeTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
                                  StringRef ParentName, unsigned LineNum,
                                  unsigned Order) {
assert(CGM.getLangOpts().OpenMPIsDevice && "Initialization of entries is "((void)0)
                                           "only required for the device "((void)0)
                                           "code generation.")((void)0);
OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] =
    OffloadEntryInfoTargetRegion(Order, /*Addr=*/nullptr, /*ID=*/nullptr,
                                 OMPTargetRegionEntryTargetRegion);
++OffloadingEntriesNum;
2992}

2994void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
  registerTargetRegionEntryInfo(unsigned DeviceID, unsigned FileID,
                                StringRef ParentName, unsigned LineNum,
                                llvm::Constant *Addr, llvm::Constant *ID,
                                OMPTargetRegionEntryKind Flags) {
// If we are emitting code for a target, the entry is already initialized,
// only has to be registered.
if (CGM.getLangOpts().OpenMPIsDevice) {
  // This could happen if the device compilation is invoked standalone.
  if (!hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum))
    return;
  auto &Entry =
      OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum];
  Entry.setAddress(Addr);
  Entry.setID(ID);
  Entry.setFlags(Flags);
} else {
  if (Flags ==
          OffloadEntriesInfoManagerTy::OMPTargetRegionEntryTargetRegion &&
      hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum,
                               /*IgnoreAddressId*/ true))
    return;
  assert(!hasTargetRegionEntryInfo(DeviceID, FileID, ParentName, LineNum) &&((void)0)
         "Target region entry already registered!")((void)0);
  OffloadEntryInfoTargetRegion Entry(OffloadingEntriesNum, Addr, ID, Flags);
  OffloadEntriesTargetRegion[DeviceID][FileID][ParentName][LineNum] = Entry;
  ++OffloadingEntriesNum;
}
3022}

3024bool CGOpenMPRuntime::OffloadEntriesInfoManagerTy::hasTargetRegionEntryInfo(
  unsigned DeviceID, unsigned FileID, StringRef ParentName, unsigned LineNum,
  bool IgnoreAddressId) const {
auto PerDevice = OffloadEntriesTargetRegion.find(DeviceID);
if (PerDevice == OffloadEntriesTargetRegion.end())
  return false;
auto PerFile = PerDevice->second.find(FileID);
if (PerFile == PerDevice->second.end())
  return false;
auto PerParentName = PerFile->second.find(ParentName);
if (PerParentName == PerFile->second.end())
  return false;
auto PerLine = PerParentName->second.find(LineNum);
if (PerLine == PerParentName->second.end())
  return false;
// Fail if this entry is already registered.
if (!IgnoreAddressId &&
    (PerLine->second.getAddress() || PerLine->second.getID()))
  return false;
return true;
3044}

3046void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::actOnTargetRegionEntriesInfo(
  const OffloadTargetRegionEntryInfoActTy &Action) {
// Scan all target region entries and perform the provided action.
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);
3054}

3056void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
  initializeDeviceGlobalVarEntryInfo(StringRef Name,
                                     OMPTargetGlobalVarEntryKind Flags,
                                     unsigned Order) {
assert(CGM.getLangOpts().OpenMPIsDevice && "Initialization of entries is "((void)0)
                                           "only required for the device "((void)0)
                                           "code generation.")((void)0);
OffloadEntriesDeviceGlobalVar.try_emplace(Name, Order, Flags);
++OffloadingEntriesNum;
3065}

3067void CGOpenMPRuntime::OffloadEntriesInfoManagerTy::
  registerDeviceGlobalVarEntryInfo(StringRef VarName, llvm::Constant *Addr,
                                   CharUnits VarSize,
                                   OMPTargetGlobalVarEntryKind Flags,
                                   llvm::GlobalValue::LinkageTypes Linkage) {
if (CGM.getLangOpts().OpenMPIsDevice) {
  // This could happen if the device compilation is invoked standalone.
  if (!hasDeviceGlobalVarEntryInfo(VarName))
    return;
  auto &Entry = OffloadEntriesDeviceGlobalVar[VarName];
  if (Entry.getAddress() && hasDeviceGlobalVarEntryInfo(VarName)) {
    if (Entry.getVarSize().isZero()) {
      Entry.setVarSize(VarSize);
      Entry.setLinkage(Linkage);
    }
    return;
  }
  Entry.setVarSize(VarSize);
  Entry.setLinkage(Linkage);
  Entry.setAddress(Addr);
} else {
  if (hasDeviceGlobalVarEntryInfo(VarName)) {
    auto &Entry = OffloadEntriesDeviceGlobalVar[VarName];
    assert(Entry.isValid() && Entry.getFlags() == Flags &&((void)0)
           "Entry not initialized!")((void)0);
    if (Entry.getVarSize().isZero()) {
      Entry.setVarSize(VarSize);
      Entry.setLinkage(Linkage);
    }
    return;
  }
  OffloadEntriesDeviceGlobalVar.try_emplace(
      VarName, OffloadingEntriesNum, Addr, VarSize, Flags, Linkage);
  ++OffloadingEntriesNum;
}
3102}

3104void 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());
3110}

3112void CGOpenMPRuntime::createOffloadEntry(
  llvm::Constant *ID, llvm::Constant *Addr, uint64_t Size, int32_t Flags,
  llvm::GlobalValue::LinkageTypes Linkage) {
StringRef Name = Addr->getName();
llvm::Module &M = CGM.getModule();
llvm::LLVMContext &C = M.getContext();

// Create constant string with the name.
llvm::Constant *StrPtrInit = llvm::ConstantDataArray::getString(C, 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 *Data[] = {
    llvm::ConstantExpr::getPointerBitCastOrAddrSpaceCast(ID, CGM.VoidPtrTy),
    llvm::ConstantExpr::getPointerBitCastOrAddrSpaceCast(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 = createGlobalStruct(
    CGM, getTgtOffloadEntryQTy(), /*IsConstant=*/true, 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");
3141}

3143void CGOpenMPRuntime::createOffloadEntriesAndInfoMetadata() {
// Emit the offloading entries and metadata so that the device codegen side
// can easily figure out what to emit. The produced metadata looks like
// this:
//
// !omp_offload.info = !{!1, ...}
//
// Right now we only generate metadata for function that contain target
// regions.

// If we are in simd mode or there are no entries, we don't need to do
// anything.
if (CGM.getLangOpts().OpenMPSimd || OffloadEntriesInfoManager.empty())
  return;

llvm::Module &M = CGM.getModule();
llvm::LLVMContext &C = M.getContext();
SmallVector<std::tuple<const OffloadEntriesInfoManagerTy::OffloadEntryInfo *,
                       SourceLocation, StringRef>,
            16>
    OrderedEntries(OffloadEntriesInfoManager.size());
llvm::SmallVector<StringRef, 16> ParentFunctions(
    OffloadEntriesInfoManager.size());

// Auxiliary methods to create metadata values and strings.
auto &&GetMDInt = [this](unsigned V) {
  return llvm::ConstantAsMetadata::get(
      llvm::ConstantInt::get(CGM.Int32Ty, 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 =
    [this, &C, MD, &OrderedEntries, &ParentFunctions, &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())};

      SourceLocation Loc;
      for (auto I = CGM.getContext().getSourceManager().fileinfo_begin(),
                E = CGM.getContext().getSourceManager().fileinfo_end();
           I != E; ++I) {
        if (I->getFirst()->getUniqueID().getDevice() == DeviceID &&
            I->getFirst()->getUniqueID().getFile() == FileID) {
          Loc = CGM.getContext().getSourceManager().translateFileLineCol(
              I->getFirst(), Line, 1);
          break;
        }
      }
      // Save this entry in the right position of the ordered entries array.
      OrderedEntries[E.getOrder()] = std::make_tuple(&E, Loc, ParentName);
      ParentFunctions[E.getOrder()] = ParentName;

      // Add metadata to the named metadata node.
      MD->addOperand(llvm::MDNode::get(C, Ops));
    };

OffloadEntriesInfoManager.actOnTargetRegionEntriesInfo(
    TargetRegionMetadataEmitter);

// 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()] =
          std::make_tuple(&E, SourceLocation(), MangledName);

      // Add metadata to the named metadata node.
      MD->addOperand(llvm::MDNode::get(C, Ops));
    };

OffloadEntriesInfoManager.actOnDeviceGlobalVarEntriesInfo(
    DeviceGlobalVarMetadataEmitter);

for (const auto &E : OrderedEntries) {
  assert(std::get<0>(E) && "All ordered entries must exist!")((void)0);
  if (const auto *CE =
          dyn_cast<OffloadEntriesInfoManagerTy::OffloadEntryInfoTargetRegion>(
              std::get<0>(E))) {
    if (!CE->getID() || !CE->getAddress()) {
      // Do not blame the entry if the parent funtion is not emitted.
      StringRef FnName = ParentFunctions[CE->getOrder()];
      if (!CGM.GetGlobalValue(FnName))
        continue;
      unsigned DiagID = CGM.getDiags().getCustomDiagID(
          DiagnosticsEngine::Error,
          "Offloading entry for target region in %0 is incorrect: either the "
          "address or the ID is invalid.");
      CGM.getDiags().Report(std::get<1>(E), DiagID) << FnName;
      continue;
    }
    createOffloadEntry(CE->getID(), CE->getAddress(), /*Size=*/0,
                       CE->getFlags(), llvm::GlobalValue::WeakAnyLinkage);
  } else if (const auto *CE = dyn_cast<OffloadEntriesInfoManagerTy::
                                           OffloadEntryInfoDeviceGlobalVar>(
                 std::get<0>(E))) {
    OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind Flags =
        static_cast<OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryKind>(
            CE->getFlags());
    switch (Flags) {
    case OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryTo: {
      if (CGM.getLangOpts().OpenMPIsDevice &&
          CGM.getOpenMPRuntime().hasRequiresUnifiedSharedMemory())
        continue;
      if (!CE->getAddress()) {
        unsigned DiagID = CGM.getDiags().getCustomDiagID(
            DiagnosticsEngine::Error, "Offloading entry for declare target "
                                      "variable %0 is incorrect: the "
                                      "address is invalid.");
        CGM.getDiags().Report(std::get<1>(E), DiagID) << std::get<2>(E);
        continue;
      }
      // The vaiable has no definition - no need to add the entry.
      if (CE->getVarSize().isZero())
        continue;
      break;
    }
    case OffloadEntriesInfoManagerTy::OMPTargetGlobalVarEntryLink:
      assert(((CGM.getLangOpts().OpenMPIsDevice && !CE->getAddress()) ||((void)0)
              (!CGM.getLangOpts().OpenMPIsDevice && CE->getAddress())) &&((void)0)
             "Declaret target link address is set.")((void)0);
      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.")__builtin_unreachable();
  }
}
3315}

3317/// Loads all the offload entries information from the host IR
3318/// metadata.
3319void CGOpenMPRuntime::loadOffloadInfoMetadata() {
// If we are in target mode, load the metadata from the host IR. This code has
// to match the metadaata creation in createOffloadEntriesAndInfoMetadata().

if (!CGM.getLangOpts().OpenMPIsDevice)
  return;

if (CGM.getLangOpts().OMPHostIRFile.empty())
  return;

auto Buf = llvm::MemoryBuffer::getFile(CGM.getLangOpts().OMPHostIRFile);
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 (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 (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 = [MN](unsigned Idx) {
    auto *V = cast<llvm::MDString>(MN->getOperand(Idx));
    return V->getString();
  };

  switch (GetMDInt(0)) {
  default:
    llvm_unreachable("Unexpected metadata!")__builtin_unreachable();
    break;
  case OffloadEntriesInfoManagerTy::OffloadEntryInfo::
      OffloadingEntryInfoTargetRegion:
    OffloadEntriesInfoManager.initializeTargetRegionEntryInfo(
        /*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;
  }
}
3384}

3386void CGOpenMPRuntime::emitKmpRoutineEntryT(QualType KmpInt32Ty) {
if (!KmpRoutineEntryPtrTy) {
  // Build typedef kmp_int32 (* kmp_routine_entry_t)(kmp_int32, void *); type.
  ASTContext &C = CGM.getContext();
  QualType KmpRoutineEntryTyArgs[] = {KmpInt32Ty, C.VoidPtrTy};
  FunctionProtoType::ExtProtoInfo EPI;
  KmpRoutineEntryPtrQTy = C.getPointerType(
      C.getFunctionType(KmpInt32Ty, KmpRoutineEntryTyArgs, EPI));
  KmpRoutineEntryPtrTy = CGM.getTypes().ConvertType(KmpRoutineEntryPtrQTy);
}
3396}

3398QualType CGOpenMPRuntime::getTgtOffloadEntryQTy() {
// Make sure the type of the entry is already created. This is the type we
// have to create:
// struct __tgt_offload_entry{
//   void      *addr;       // Pointer to the offload entry info.
//                          // (function or global)
//   char      *name;       // Name of the function or global.
//   size_t     size;       // Size of the entry info (0 if it a function).
//   int32_t    flags;      // Flags associated with the entry, e.g. 'link'.
//   int32_t    reserved;   // Reserved, to use by the runtime library.
// };
if (TgtOffloadEntryQTy.isNull()) {
  ASTContext &C = CGM.getContext();
  RecordDecl *RD = C.buildImplicitRecord("__tgt_offload_entry");
  RD->startDefinition();
  addFieldToRecordDecl(C, RD, C.VoidPtrTy);
  addFieldToRecordDecl(C, RD, C.getPointerType(C.CharTy));
  addFieldToRecordDecl(C, RD, C.getSizeType());
  addFieldToRecordDecl(
      C, RD, C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true));
  addFieldToRecordDecl(
      C, RD, C.getIntTypeForBitwidth(/*DestWidth=*/32, /*Signed=*/true));
  RD->completeDefinition();
  RD->addAttr(PackedAttr::CreateImplicit(C));
  TgtOffloadEntryQTy = C.getRecordType(RD);
}
return TgtOffloadEntryQTy;
3425}

3427namespace {
3428struct PrivateHelpersTy {
PrivateHelpersTy(const Expr *OriginalRef, const VarDecl *Original,
                 const VarDecl *PrivateCopy, const VarDecl *PrivateElemInit)
    : OriginalRef(OriginalRef), Original(Original), PrivateCopy(PrivateCopy),
      PrivateElemInit(PrivateElemInit) {}
PrivateHelpersTy(const VarDecl *Original) : Original(Original) {}
const Expr *OriginalRef = nullptr;
const VarDecl *Original = nullptr;
const VarDecl *PrivateCopy = nullptr;
const VarDecl *PrivateElemInit = nullptr;
bool isLocalPrivate() const {
  return !OriginalRef && !PrivateCopy && !PrivateElemInit;
}
3441};
3442typedef std::pair<CharUnits /*Align*/, PrivateHelpersTy> PrivateDataTy;
3443} // anonymous namespace

3445static bool isAllocatableDecl(const VarDecl *VD) {
const VarDecl *CVD = VD->getCanonicalDecl();
if (!CVD->hasAttr<OMPAllocateDeclAttr>())
  return false;
const auto *AA = CVD->getAttr<OMPAllocateDeclAttr>();
// Use the default allocation.
return !((AA->getAllocatorType() == OMPAllocateDeclAttr::OMPDefaultMemAlloc ||
          AA->getAllocatorType() == OMPAllocateDeclAttr::OMPNullMemAlloc) &&
         !AA->getAllocator());
3454}

3456static RecordDecl *
3457createPrivatesRecordDecl(CodeGenModule &CGM, ArrayRef<PrivateDataTy> Privates) {
if (!Privates.empty()) {
  ASTContext &C = CGM.getContext();
  // Build struct .kmp_privates_t. {
  //         /*  private vars  */
  //       };
  RecordDecl *RD = C.buildImplicitRecord(".kmp_privates.t");
  RD->startDefinition();
  for (const auto &Pair : Privates) {
    const VarDecl *VD = Pair.second.Original;
    QualType Type = VD->getType().getNonReferenceType();
    // If the private variable is a local variable with lvalue ref type,
    // allocate the pointer instead of the pointee type.
    if (Pair.second.isLocalPrivate()) {
      if (VD->getType()->isLValueReferenceType())
        Type = C.getPointerType(Type);
      if (isAllocatableDecl(VD))
        Type = C.getPointerType(Type);
    }
    FieldDecl *FD = addFieldToRecordDecl(C, RD, Type);
    if (VD->hasAttrs()) {
      for (specific_attr_iterator<AlignedAttr> I(VD->getAttrs().begin()),
           E(VD->getAttrs().end());
           I != E; ++I)
        FD->addAttr(*I);
    }
  }
  RD->completeDefinition();
  return RD;
}
return nullptr;
3488}

3490static RecordDecl *
3491createKmpTaskTRecordDecl(CodeGenModule &CGM, OpenMPDirectiveKind Kind,
                       QualType KmpInt32Ty,
                       QualType KmpRoutineEntryPointerQTy) {
ASTContext &C = CGM.getContext();
// Build struct kmp_task_t {
//         void *              shareds;
//         kmp_routine_entry_t routine;
//         kmp_int32           part_id;
//         kmp_cmplrdata_t data1;
//         kmp_cmplrdata_t data2;
// For taskloops additional fields:
//         kmp_uint64          lb;
//         kmp_uint64          ub;
//         kmp_int64           st;
//         kmp_int32           liter;
//         void *              reductions;
//       };
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);
RecordDecl *RD = C.buildImplicitRecord("kmp_task_t");
RD->startDefinition();
addFieldToRecordDecl(C, RD, C.VoidPtrTy);
addFieldToRecordDecl(C, RD, KmpRoutineEntryPointerQTy);
addFieldToRecordDecl(C, RD, KmpInt32Ty);
addFieldToRecordDecl(C, RD, KmpCmplrdataTy);
addFieldToRecordDecl(C, RD, KmpCmplrdataTy);
if (isOpenMPTaskLoopDirective(Kind)) {
  QualType KmpUInt64Ty =
      CGM.getContext().getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/0);
  QualType KmpInt64Ty =
      CGM.getContext().getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/1);
  addFieldToRecordDecl(C, RD, KmpUInt64Ty);
  addFieldToRecordDecl(C, RD, KmpUInt64Ty);
  addFieldToRecordDecl(C, RD, KmpInt64Ty);
  addFieldToRecordDecl(C, RD, KmpInt32Ty);
  addFieldToRecordDecl(C, RD, C.VoidPtrTy);
}
RD->completeDefinition();
return RD;
3534}

3536static RecordDecl *
3537createKmpTaskTWithPrivatesRecordDecl(CodeGenModule &CGM, QualType KmpTaskTQTy,
                                   ArrayRef<PrivateDataTy> Privates) {
ASTContext &C = CGM.getContext();
// Build struct kmp_task_t_with_privates {
//         kmp_task_t task_data;
//         .kmp_privates_t. privates;
//       };
RecordDecl *RD = C.buildImplicitRecord("kmp_task_t_with_privates");
RD->startDefinition();
addFieldToRecordDecl(C, RD, KmpTaskTQTy);
if (const RecordDecl *PrivateRD = createPrivatesRecordDecl(CGM, Privates))
  addFieldToRecordDecl(C, RD, C.getRecordType(PrivateRD));
RD->completeDefinition();
return RD;
3551}

3553/// Emit a proxy function which accepts kmp_task_t as the second
3554/// argument.
3555/// \code
3556/// kmp_int32 .omp_task_entry.(kmp_int32 gtid, kmp_task_t *tt) {
3557///   TaskFunction(gtid, tt->part_id, &tt->privates, task_privates_map, tt,
3558///   For taskloops:
3559///   tt->task_data.lb, tt->task_data.ub, tt->task_data.st, tt->task_data.liter,
3560///   tt->reductions, tt->shareds);
3561///   return 0;
3562/// }
3563/// \endcode
3564static llvm::Function *
3565emitProxyTaskFunction(CodeGenModule &CGM, SourceLocation Loc,
                    OpenMPDirectiveKind Kind, QualType KmpInt32Ty,
                    QualType KmpTaskTWithPrivatesPtrQTy,
                    QualType KmpTaskTWithPrivatesQTy, QualType KmpTaskTQTy,
                    QualType SharedsPtrTy, llvm::Function *TaskFunction,
                    llvm::Value *TaskPrivatesMap) {
ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl GtidArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, KmpInt32Ty,
                          ImplicitParamDecl::Other);
ImplicitParamDecl TaskTypeArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
                              KmpTaskTWithPrivatesPtrQTy.withRestrict(),
                              ImplicitParamDecl::Other);
Args.push_back(&GtidArg);
Args.push_back(&TaskTypeArg);
const auto &TaskEntryFnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(KmpInt32Ty, Args);
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,
                  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);
llvm::Value *GtidParam = CGF.EmitLoadOfScalar(
    CGF.GetAddrOfLocalVar(&GtidArg), /*Volatile=*/false, KmpInt32Ty, Loc);
LValue TDBase = CGF.EmitLoadOfPointerLValue(
    CGF.GetAddrOfLocalVar(&TaskTypeArg),
    KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
const auto *KmpTaskTWithPrivatesQTyRD =
    cast<RecordDecl>(KmpTaskTWithPrivatesQTy->getAsTagDecl());
LValue Base =
    CGF.EmitLValueForField(TDBase, *KmpTaskTWithPrivatesQTyRD->field_begin());
const auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
auto PartIdFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTPartId);
LValue PartIdLVal = CGF.EmitLValueForField(Base, *PartIdFI);
llvm::Value *PartidParam = PartIdLVal.getPointer(CGF);

auto SharedsFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTShareds);
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()) {
  LValue PrivatesLVal = CGF.EmitLValueForField(TDBase, *PrivatesFI);
  PrivatesParam = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
      PrivatesLVal.getPointer(CGF), CGF.VoidPtrTy);
} else {
  PrivatesParam = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
}

llvm::Value *CommonArgs[] = {GtidParam, PartidParam, PrivatesParam,
                             TaskPrivatesMap,
                             CGF.Builder
                                 .CreatePointerBitCastOrAddrSpaceCast(
                                     TDBase.getAddress(CGF), CGF.VoidPtrTy)
                                 .getPointer()};
SmallVector<llvm::Value *, 16> CallArgs(std::begin(CommonArgs),
                                        std::end(CommonArgs));
if (isOpenMPTaskLoopDirective(Kind)) {
  auto LBFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTLowerBound);
  LValue LBLVal = CGF.EmitLValueForField(Base, *LBFI);
  llvm::Value *LBParam = CGF.EmitLoadOfScalar(LBLVal, Loc);
  auto UBFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTUpperBound);
  LValue UBLVal = CGF.EmitLValueForField(Base, *UBFI);
  llvm::Value *UBParam = CGF.EmitLoadOfScalar(UBLVal, Loc);
  auto StFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTStride);
  LValue StLVal = CGF.EmitLValueForField(Base, *StFI);
  llvm::Value *StParam = CGF.EmitLoadOfScalar(StLVal, Loc);
  auto LIFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTLastIter);
  LValue LILVal = CGF.EmitLValueForField(Base, *LIFI);
  llvm::Value *LIParam = CGF.EmitLoadOfScalar(LILVal, Loc);
  auto RFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTReductions);
  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);
  CallArgs.push_back(LIParam);
  CallArgs.push_back(RParam);
}
CallArgs.push_back(SharedsParam);

CGM.getOpenMPRuntime().emitOutlinedFunctionCall(CGF, Loc, TaskFunction,
                                                CallArgs);
CGF.EmitStoreThroughLValue(RValue::get(CGF.Builder.getInt32(/*C=*/0)),
                           CGF.MakeAddrLValue(CGF.ReturnValue, KmpInt32Ty));
CGF.FinishFunction();
return TaskEntry;
3666}

3668static llvm::Value *emitDestructorsFunction(CodeGenModule &CGM,
                                          SourceLocation Loc,
                                          QualType KmpInt32Ty,
                                          QualType KmpTaskTWithPrivatesPtrQTy,
                                          QualType KmpTaskTWithPrivatesQTy) {
ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl GtidArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, KmpInt32Ty,
                          ImplicitParamDecl::Other);
ImplicitParamDecl TaskTypeArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
                              KmpTaskTWithPrivatesPtrQTy.withRestrict(),
                              ImplicitParamDecl::Other);
Args.push_back(&GtidArg);
Args.push_back(&TaskTypeArg);
const auto &DestructorFnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(KmpInt32Ty, Args);
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,
                           Name, &CGM.getModule());
CGM.SetInternalFunctionAttributes(GlobalDecl(), DestructorFn,
                                  DestructorFnInfo);
DestructorFn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
CGF.StartFunction(GlobalDecl(), KmpInt32Ty, DestructorFn, DestructorFnInfo,
                  Args, Loc, Loc);

LValue Base = CGF.EmitLoadOfPointerLValue(
    CGF.GetAddrOfLocalVar(&TaskTypeArg),
    KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
const auto *KmpTaskTWithPrivatesQTyRD =
    cast<RecordDecl>(KmpTaskTWithPrivatesQTy->getAsTagDecl());
auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
Base = CGF.EmitLValueForField(Base, *FI);
for (const auto *Field :
     cast<RecordDecl>(FI->getType()->getAsTagDecl())->fields()) {
  if (QualType::DestructionKind DtorKind =
          Field->getType().isDestructedType()) {
    LValue FieldLValue = CGF.EmitLValueForField(Base, Field);
    CGF.pushDestroy(DtorKind, FieldLValue.getAddress(CGF), Field->getType());
  }
}
CGF.FinishFunction();
return DestructorFn;
3715}

3717/// Emit a privates mapping function for correct handling of private and
3718/// firstprivate variables.
3719/// \code
3720/// void .omp_task_privates_map.(const .privates. *noalias privs, <ty1>
3721/// **noalias priv1,...,  <tyn> **noalias privn) {
3722///   *priv1 = &.privates.priv1;
3723///   ...;
3724///   *privn = &.privates.privn;
3725/// }
3726/// \endcode
3727static llvm::Value *
3728emitTaskPrivateMappingFunction(CodeGenModule &CGM, SourceLocation Loc,
                             const OMPTaskDataTy &Data, QualType PrivatesQTy,
                             ArrayRef<PrivateDataTy> Privates) {
ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl TaskPrivatesArg(
    C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
    C.getPointerType(PrivatesQTy).withConst().withRestrict(),
    ImplicitParamDecl::Other);
Args.push_back(&TaskPrivatesArg);
llvm::DenseMap<CanonicalDeclPtr<const VarDecl>, unsigned> PrivateVarsPos;
unsigned Counter = 1;
for (const Expr *E : Data.PrivateVars) {
  Args.push_back(ImplicitParamDecl::Create(
      C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
      C.getPointerType(C.getPointerType(E->getType()))
          .withConst()
          .withRestrict(),
      ImplicitParamDecl::Other));
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  PrivateVarsPos[VD] = Counter;
  ++Counter;
}
for (const Expr *E : Data.FirstprivateVars) {
  Args.push_back(ImplicitParamDecl::Create(
      C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
      C.getPointerType(C.getPointerType(E->getType()))
          .withConst()
          .withRestrict(),
      ImplicitParamDecl::Other));
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  PrivateVarsPos[VD] = Counter;
  ++Counter;
}
for (const Expr *E : Data.LastprivateVars) {
  Args.push_back(ImplicitParamDecl::Create(
      C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
      C.getPointerType(C.getPointerType(E->getType()))
          .withConst()
          .withRestrict(),
      ImplicitParamDecl::Other));
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  PrivateVarsPos[VD] = Counter;
  ++Counter;
}
for (const VarDecl *VD : Data.PrivateLocals) {
  QualType Ty = VD->getType().getNonReferenceType();
  if (VD->getType()->isLValueReferenceType())
    Ty = C.getPointerType(Ty);
  if (isAllocatableDecl(VD))
    Ty = C.getPointerType(Ty);
  Args.push_back(ImplicitParamDecl::Create(
      C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
      C.getPointerType(C.getPointerType(Ty)).withConst().withRestrict(),
      ImplicitParamDecl::Other));
  PrivateVarsPos[VD] = Counter;
  ++Counter;
}
const auto &TaskPrivatesMapFnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
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, Name,
    &CGM.getModule());
CGM.SetInternalFunctionAttributes(GlobalDecl(), TaskPrivatesMap,
                                  TaskPrivatesMapFnInfo);
if (CGM.getLangOpts().Optimize) {
  TaskPrivatesMap->removeFnAttr(llvm::Attribute::NoInline);
  TaskPrivatesMap->removeFnAttr(llvm::Attribute::OptimizeNone);
  TaskPrivatesMap->addFnAttr(llvm::Attribute::AlwaysInline);
}
CodeGenFunction CGF(CGM);
CGF.StartFunction(GlobalDecl(), C.VoidTy, TaskPrivatesMap,
                  TaskPrivatesMapFnInfo, Args, Loc, Loc);

// *privi = &.privates.privi;
LValue Base = CGF.EmitLoadOfPointerLValue(
    CGF.GetAddrOfLocalVar(&TaskPrivatesArg),
    TaskPrivatesArg.getType()->castAs<PointerType>());
const auto *PrivatesQTyRD = cast<RecordDecl>(PrivatesQTy->getAsTagDecl());
Counter = 0;
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(CGF), RefLVal.getType()->castAs<PointerType>());
  CGF.EmitStoreOfScalar(FieldLVal.getPointer(CGF), RefLoadLVal);
  ++Counter;
}
CGF.FinishFunction();
return TaskPrivatesMap;
3824}

3826/// Emit initialization for private variables in task-based directives.
3827static void emitPrivatesInit(CodeGenFunction &CGF,
                           const OMPExecutableDirective &D,
                           Address KmpTaskSharedsPtr, LValue TDBase,
                           const RecordDecl *KmpTaskTWithPrivatesQTyRD,
                           QualType SharedsTy, QualType SharedsPtrTy,
                           const OMPTaskDataTy &Data,
                           ArrayRef<PrivateDataTy> Privates, bool ForDup) {
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()) ||
    isOpenMPTargetExecutionDirective(D.getDirectiveKind());
// For target-based directives skip 4 firstprivate arrays BasePointersArray,
// PointersArray, SizesArray, and MappersArray. The original variables for
// these arrays are not captured and we get their addresses explicitly.
if ((!IsTargetTask && !Data.FirstprivateVars.empty() && ForDup) ||
    (IsTargetTask && KmpTaskSharedsPtr.isValid())) {
  SrcBase = CGF.MakeAddrLValue(
      CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
          KmpTaskSharedsPtr, CGF.ConvertTypeForMem(SharedsPtrTy)),
      SharedsTy);
}
FI = cast<RecordDecl>(FI->getType()->getAsTagDecl())->field_begin();
for (const PrivateDataTy &Pair : Privates) {
  // Do not initialize private locals.
  if (Pair.second.isLocalPrivate()) {
    ++FI;
    continue;
  }
  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 (const VarDecl *Elem = Pair.second.PrivateElemInit) {
      const VarDecl *OriginalVD = Pair.second.Original;
      // Check if the variable is the target-based BasePointersArray,
      // PointersArray, SizesArray, or MappersArray.
      LValue SharedRefLValue;
      QualType Type = PrivateLValue.getType();
      const FieldDecl *SharedField = CapturesInfo.lookup(OriginalVD);
      if (IsTargetTask && !SharedField) {
        assert(isa<ImplicitParamDecl>(OriginalVD) &&((void)0)
               isa<CapturedDecl>(OriginalVD->getDeclContext()) &&((void)0)
               cast<CapturedDecl>(OriginalVD->getDeclContext())((void)0)
                       ->getNumParams() == 0 &&((void)0)
               isa<TranslationUnitDecl>(((void)0)
                   cast<CapturedDecl>(OriginalVD->getDeclContext())((void)0)
                       ->getDeclContext()) &&((void)0)
               "Expected artificial target data variable.")((void)0);
        SharedRefLValue =
            CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(OriginalVD), Type);
      } else if (ForDup) {
        SharedRefLValue = CGF.EmitLValueForField(SrcBase, SharedField);
        SharedRefLValue = CGF.MakeAddrLValue(
            Address(SharedRefLValue.getPointer(CGF),
                    C.getDeclAlign(OriginalVD)),
            SharedRefLValue.getType(), LValueBaseInfo(AlignmentSource::Decl),
            SharedRefLValue.getTBAAInfo());
      } else if (CGF.LambdaCaptureFields.count(
                     Pair.second.Original->getCanonicalDecl()) > 0 ||
                 dyn_cast_or_null<BlockDecl>(CGF.CurCodeDecl)) {
        SharedRefLValue = CGF.EmitLValue(Pair.second.OriginalRef);
      } else {
        // Processing for implicitly captured variables.
        InlinedOpenMPRegionRAII Region(
            CGF, [](CodeGenFunction &, PrePostActionTy &) {}, OMPD_unknown,
            /*HasCancel=*/false, /*NoInheritance=*/true);
        SharedRefLValue = CGF.EmitLValue(Pair.second.OriginalRef);
      }
      if (Type->isArrayType()) {
        // Initialize firstprivate array.
        if (!isa<CXXConstructExpr>(Init) || CGF.isTrivialInitializer(Init)) {
          // Perform simple memcpy.
          CGF.EmitAggregateAssign(PrivateLValue, SharedRefLValue, Type);
        } else {
          // Initialize firstprivate array using element-by-element
          // initialization.
          CGF.EmitOMPAggregateAssign(
              PrivateLValue.getAddress(CGF), SharedRefLValue.getAddress(CGF),
              Type,
              [&CGF, Elem, Init, &CapturesInfo](Address DestElement,
                                                Address SrcElement) {
                // Clean up any temporaries needed by the initialization.
                CodeGenFunction::OMPPrivateScope InitScope(CGF);
                InitScope.addPrivate(
                    Elem, [SrcElement]() -> Address { return SrcElement; });
                (void)InitScope.Privatize();
                // Emit initialization for single element.
                CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(
                    CGF, &CapturesInfo);
                CGF.EmitAnyExprToMem(Init, DestElement,
                                     Init->getType().getQualifiers(),
                                     /*IsInitializer=*/false);
              });
        }
      } else {
        CodeGenFunction::OMPPrivateScope InitScope(CGF);
        InitScope.addPrivate(Elem, [SharedRefLValue, &CGF]() -> Address {
          return SharedRefLValue.getAddress(CGF);
        });
        (void)InitScope.Privatize();
        CodeGenFunction::CGCapturedStmtRAII CapInfoRAII(CGF, &CapturesInfo);
        CGF.EmitExprAsInit(Init, VD, PrivateLValue,
                           /*capturedByInit=*/false);
      }
    } else {
      CGF.EmitExprAsInit(Init, VD, PrivateLValue, /*capturedByInit=*/false);
    }
  }
  ++FI;
}
3946}

3948/// Check if duplication function is required for taskloops.
3949static bool checkInitIsRequired(CodeGenFunction &CGF,
                              ArrayRef<PrivateDataTy> Privates) {
bool InitRequired = false;
for (const PrivateDataTy &Pair : Privates) {
  if (Pair.second.isLocalPrivate())
    continue;
  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;
3963}


3966/// Emit task_dup function (for initialization of
3967/// private/firstprivate/lastprivate vars and last_iter flag)
3968/// \code
3969/// void __task_dup_entry(kmp_task_t *task_dst, const kmp_task_t *task_src, int
3970/// lastpriv) {
3971/// // setup lastprivate flag
3972///    task_dst->last = lastpriv;
3973/// // could be constructor calls here...
3974/// }
3975/// \endcode
3976static llvm::Value *
3977emitTaskDupFunction(CodeGenModule &CGM, SourceLocation Loc,
                  const OMPExecutableDirective &D,
                  QualType KmpTaskTWithPrivatesPtrQTy,
                  const RecordDecl *KmpTaskTWithPrivatesQTyRD,
                  const RecordDecl *KmpTaskTQTyRD, QualType SharedsTy,
                  QualType SharedsPtrTy, const OMPTaskDataTy &Data,
                  ArrayRef<PrivateDataTy> Privates, bool WithLastIter) {
ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl DstArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
                         KmpTaskTWithPrivatesPtrQTy,
                         ImplicitParamDecl::Other);
ImplicitParamDecl SrcArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr,
                         KmpTaskTWithPrivatesPtrQTy,
                         ImplicitParamDecl::Other);
ImplicitParamDecl LastprivArg(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.IntTy,
                              ImplicitParamDecl::Other);
Args.push_back(&DstArg);
Args.push_back(&SrcArg);
Args.push_back(&LastprivArg);
const auto &TaskDupFnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
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.StartFunction(GlobalDecl(), C.VoidTy, TaskDup, TaskDupFnInfo, Args, Loc,
                  Loc);

LValue TDBase = CGF.EmitLoadOfPointerLValue(
    CGF.GetAddrOfLocalVar(&DstArg),
    KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
// task_dst->liter = lastpriv;
if (WithLastIter) {
  auto LIFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTLastIter);
  LValue Base = CGF.EmitLValueForField(
      TDBase, *KmpTaskTWithPrivatesQTyRD->field_begin());
  LValue LILVal = CGF.EmitLValueForField(Base, *LIFI);
  llvm::Value *Lastpriv = CGF.EmitLoadOfScalar(
      CGF.GetAddrOfLocalVar(&LastprivArg), /*Volatile=*/false, C.IntTy, Loc);
  CGF.EmitStoreOfScalar(Lastpriv, LILVal);
}

// Emit initial values for private copies (if any).
assert(!Privates.empty())((void)0);
Address KmpTaskSharedsPtr = Address::invalid();
if (!Data.FirstprivateVars.empty()) {
  LValue TDBase = CGF.EmitLoadOfPointerLValue(
      CGF.GetAddrOfLocalVar(&SrcArg),
      KmpTaskTWithPrivatesPtrQTy->castAs<PointerType>());
  LValue Base = CGF.EmitLValueForField(
      TDBase, *KmpTaskTWithPrivatesQTyRD->field_begin());
  KmpTaskSharedsPtr = Address(
      CGF.EmitLoadOfScalar(CGF.EmitLValueForField(
                               Base, *std::next(KmpTaskTQTyRD->field_begin(),
                                                KmpTaskTShareds)),
                           Loc),
      CGM.getNaturalTypeAlignment(SharedsTy));
}
emitPrivatesInit(CGF, D, KmpTaskSharedsPtr, TDBase, KmpTaskTWithPrivatesQTyRD,
                 SharedsTy, SharedsPtrTy, Data, Privates, /*ForDup=*/true);
CGF.FinishFunction();
return TaskDup;
4043}

4045/// Checks if destructor function is required to be generated.
4046/// \return true if cleanups are required, false otherwise.
4047static bool
4048checkDestructorsRequired(const RecordDecl *KmpTaskTWithPrivatesQTyRD,
                       ArrayRef<PrivateDataTy> Privates) {
for (const PrivateDataTy &P : Privates) {
  if (P.second.isLocalPrivate())
    continue;
  QualType Ty = P.second.Original->getType().getNonReferenceType();
  if (Ty.isDestructedType())
    return true;
}
return false;
4058}

4060namespace {
4061/// Loop generator for OpenMP iterator expression.
4062class OMPIteratorGeneratorScope final
  : public CodeGenFunction::OMPPrivateScope {
CodeGenFunction &CGF;
const OMPIteratorExpr *E = nullptr;
SmallVector<CodeGenFunction::JumpDest, 4> ContDests;
SmallVector<CodeGenFunction::JumpDest, 4> ExitDests;
OMPIteratorGeneratorScope() = delete;
OMPIteratorGeneratorScope(OMPIteratorGeneratorScope &) = delete;

4071public:
OMPIteratorGeneratorScope(CodeGenFunction &CGF, const OMPIteratorExpr *E)
    : CodeGenFunction::OMPPrivateScope(CGF), CGF(CGF), E(E) {
  if (!E)
    return;
  SmallVector<llvm::Value *, 4> Uppers;
  for (unsigned I = 0, End = E->numOfIterators(); I < End; ++I) {
    Uppers.push_back(CGF.EmitScalarExpr(E->getHelper(I).Upper));
    const auto *VD = cast<VarDecl>(E->getIteratorDecl(I));
    addPrivate(VD, [&CGF, VD]() {
      return CGF.CreateMemTemp(VD->getType(), VD->getName());
    });
    const OMPIteratorHelperData &HelperData = E->getHelper(I);
    addPrivate(HelperData.CounterVD, [&CGF, &HelperData]() {
      return CGF.CreateMemTemp(HelperData.CounterVD->getType(),
                               "counter.addr");
    });
  }
  Privatize();

  for (unsigned I = 0, End = E->numOfIterators(); I < End; ++I) {
    const OMPIteratorHelperData &HelperData = E->getHelper(I);
    LValue CLVal =
        CGF.MakeAddrLValue(CGF.GetAddrOfLocalVar(HelperData.CounterVD),
                           HelperData.CounterVD->getType());
    // Counter = 0;
    CGF.EmitStoreOfScalar(
        llvm::ConstantInt::get(CLVal.getAddress(CGF).getElementType(), 0),
        CLVal);
    CodeGenFunction::JumpDest &ContDest =
        ContDests.emplace_back(CGF.getJumpDestInCurrentScope("iter.cont"));
    CodeGenFunction::JumpDest &ExitDest =
        ExitDests.emplace_back(CGF.getJumpDestInCurrentScope("iter.exit"));
    // N = <number-of_iterations>;
    llvm::Value *N = Uppers[I];
    // cont:
    // if (Counter < N) goto body; else goto exit;
    CGF.EmitBlock(ContDest.getBlock());
    auto *CVal =
        CGF.EmitLoadOfScalar(CLVal, HelperData.CounterVD->getLocation());
    llvm::Value *Cmp =
        HelperData.CounterVD->getType()->isSignedIntegerOrEnumerationType()
            ? CGF.Builder.CreateICmpSLT(CVal, N)
            : CGF.Builder.CreateICmpULT(CVal, N);
    llvm::BasicBlock *BodyBB = CGF.createBasicBlock("iter.body");
    CGF.Builder.CreateCondBr(Cmp, BodyBB, ExitDest.getBlock());
    // body:
    CGF.EmitBlock(BodyBB);
    // Iteri = Begini + Counter * Stepi;
    CGF.EmitIgnoredExpr(HelperData.Update);
  }
}
~OMPIteratorGeneratorScope() {
  if (!E)
    return;
  for (unsigned I = E->numOfIterators(); I > 0; --I) {
    // Counter = Counter + 1;
    const OMPIteratorHelperData &HelperData = E->getHelper(I - 1);
    CGF.EmitIgnoredExpr(HelperData.CounterUpdate);
    // goto cont;
    CGF.EmitBranchThroughCleanup(ContDests[I - 1]);
    // exit:
    CGF.EmitBlock(ExitDests[I - 1].getBlock(), /*IsFinished=*/I == 1);
  }
}
4136};
4137} // namespace

4139static std::pair<llvm::Value *, llvm::Value *>
4140getPointerAndSize(CodeGenFunction &CGF, const Expr *E) {
const auto *OASE = dyn_cast<OMPArrayShapingExpr>(E);
llvm::Value *Addr;
if (OASE) {
  const Expr *Base = OASE->getBase();
  Addr = CGF.EmitScalarExpr(Base);
} else {
  Addr = CGF.EmitLValue(E).getPointer(CGF);
}
llvm::Value *SizeVal;
QualType Ty = E->getType();
if (OASE) {
  SizeVal = CGF.getTypeSize(OASE->getBase()->getType()->getPointeeType());
  for (const Expr *SE : OASE->getDimensions()) {
    llvm::Value *Sz = CGF.EmitScalarExpr(SE);
    Sz = CGF.EmitScalarConversion(
        Sz, SE->getType(), CGF.getContext().getSizeType(), SE->getExprLoc());
    SizeVal = CGF.Builder.CreateNUWMul(SizeVal, Sz);
  }
} else if (const auto *ASE =
               dyn_cast<OMPArraySectionExpr>(E->IgnoreParenImpCasts())) {
  LValue UpAddrLVal =
      CGF.EmitOMPArraySectionExpr(ASE, /*IsLowerBound=*/false);
  Address UpAddrAddress = UpAddrLVal.getAddress(CGF);
  llvm::Value *UpAddr = CGF.Builder.CreateConstGEP1_32(
      UpAddrAddress.getElementType(), UpAddrAddress.getPointer(), /*Idx0=*/1);
  llvm::Value *LowIntPtr = CGF.Builder.CreatePtrToInt(Addr, CGF.SizeTy);
  llvm::Value *UpIntPtr = CGF.Builder.CreatePtrToInt(UpAddr, CGF.SizeTy);
  SizeVal = CGF.Builder.CreateNUWSub(UpIntPtr, LowIntPtr);
} else {
  SizeVal = CGF.getTypeSize(Ty);
}
return std::make_pair(Addr, SizeVal);
4173}

4175/// Builds kmp_depend_info, if it is not built yet, and builds flags type.
4176static void getKmpAffinityType(ASTContext &C, QualType &KmpTaskAffinityInfoTy) {
QualType FlagsTy = C.getIntTypeForBitwidth(32, /*Signed=*/false);
if (KmpTaskAffinityInfoTy.isNull()) {
  RecordDecl *KmpAffinityInfoRD =
      C.buildImplicitRecord("kmp_task_affinity_info_t");
  KmpAffinityInfoRD->startDefinition();
  addFieldToRecordDecl(C, KmpAffinityInfoRD, C.getIntPtrType());
  addFieldToRecordDecl(C, KmpAffinityInfoRD, C.getSizeType());
  addFieldToRecordDecl(C, KmpAffinityInfoRD, FlagsTy);
  KmpAffinityInfoRD->completeDefinition();
  KmpTaskAffinityInfoTy = C.getRecordType(KmpAffinityInfoRD);
}
4188}

4190CGOpenMPRuntime::TaskResultTy
4191CGOpenMPRuntime::emitTaskInit(CodeGenFunction &CGF, SourceLocation Loc,
                            const OMPExecutableDirective &D,
                            llvm::Function *TaskFunction, QualType SharedsTy,
                            Address Shareds, const OMPTaskDataTy &Data) {
ASTContext &C = CGM.getContext();
llvm::SmallVector<PrivateDataTy, 4> Privates;
// Aggregate privates and sort them by the alignment.
const auto *I = Data.PrivateCopies.begin();
for (const Expr *E : Data.PrivateVars) {
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  Privates.emplace_back(
      C.getDeclAlign(VD),
      PrivateHelpersTy(E, VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
                       /*PrivateElemInit=*/nullptr));
  ++I;
}
I = Data.FirstprivateCopies.begin();
const auto *IElemInitRef = Data.FirstprivateInits.begin();
for (const Expr *E : Data.FirstprivateVars) {
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  Privates.emplace_back(
      C.getDeclAlign(VD),
      PrivateHelpersTy(
          E, VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
          cast<VarDecl>(cast<DeclRefExpr>(*IElemInitRef)->getDecl())));
  ++I;
  ++IElemInitRef;
}
I = Data.LastprivateCopies.begin();
for (const Expr *E : Data.LastprivateVars) {
  const auto *VD = cast<VarDecl>(cast<DeclRefExpr>(E)->getDecl());
  Privates.emplace_back(
      C.getDeclAlign(VD),
      PrivateHelpersTy(E, VD, cast<VarDecl>(cast<DeclRefExpr>(*I)->getDecl()),
                       /*PrivateElemInit=*/nullptr));
  ++I;
}
for (const VarDecl *VD : Data.PrivateLocals) {
  if (isAllocatableDecl(VD))
    Privates.emplace_back(CGM.getPointerAlign(), PrivateHelpersTy(VD));
  else
    Privates.emplace_back(C.getDeclAlign(VD), PrivateHelpersTy(VD));
}
llvm::stable_sort(Privates,
                  [](const PrivateDataTy &L, const PrivateDataTy &R) {
                    return L.first > R.first;
                  });
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).
if (isOpenMPTaskLoopDirective(D.getDirectiveKind())) {
  if (SavedKmpTaskloopTQTy.isNull()) {
    SavedKmpTaskloopTQTy = C.getRecordType(createKmpTaskTRecordDecl(
        CGM, D.getDirectiveKind(), KmpInt32Ty, KmpRoutineEntryPtrQTy));
  }
  KmpTaskTQTy = SavedKmpTaskloopTQTy;
} else {
  assert((D.getDirectiveKind() == OMPD_task ||((void)0)
          isOpenMPTargetExecutionDirective(D.getDirectiveKind()) ||((void)0)
          isOpenMPTargetDataManagementDirective(D.getDirectiveKind())) &&((void)0)
         "Expected taskloop, task or target directive")((void)0);
  if (SavedKmpTaskTQTy.isNull()) {
    SavedKmpTaskTQTy = C.getRecordType(createKmpTaskTRecordDecl(
        CGM, D.getDirectiveKind(), KmpInt32Ty, KmpRoutineEntryPtrQTy));
  }
  KmpTaskTQTy = SavedKmpTaskTQTy;
}
const auto *KmpTaskTQTyRD = cast<RecordDecl>(KmpTaskTQTy->getAsTagDecl());
// Build particular struct kmp_task_t for the given task.
const RecordDecl *KmpTaskTWithPrivatesQTyRD =
    createKmpTaskTWithPrivatesRecordDecl(CGM, KmpTaskTQTy, Privates);
QualType KmpTaskTWithPrivatesQTy = C.getRecordType(KmpTaskTWithPrivatesQTyRD);
QualType KmpTaskTWithPrivatesPtrQTy =
    C.getPointerType(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;
llvm::Type *TaskPrivatesMapTy =
    std::next(TaskFunction->arg_begin(), 3)->getType();
if (!Privates.empty()) {
  auto FI = std::next(KmpTaskTWithPrivatesQTyRD->field_begin());
  TaskPrivatesMap =
      emitTaskPrivateMappingFunction(CGM, Loc, Data, FI->getType(), Privates);
  TaskPrivatesMap = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
      TaskPrivatesMap, TaskPrivatesMapTy);
} else {
  TaskPrivatesMap = llvm::ConstantPointerNull::get(
      cast<llvm::PointerType>(TaskPrivatesMapTy));
}
// Build a proxy function kmp_int32 .omp_task_entry.(kmp_int32 gtid,
// kmp_task_t *tt);
llvm::Function *TaskEntry = emitProxyTaskFunction(
    CGM, Loc, D.getDirectiveKind(), KmpInt32Ty, KmpTaskTWithPrivatesPtrQTy,
    KmpTaskTWithPrivatesQTy, KmpTaskTQTy, SharedsPtrTy, TaskFunction,
    TaskPrivatesMap);

// Build call kmp_task_t * __kmpc_omp_task_alloc(ident_t *, kmp_int32 gtid,
// kmp_int32 flags, size_t sizeof_kmp_task_t, size_t sizeof_shareds,
// kmp_routine_entry_t *task_entry);
// Task flags. Format is taken from
// https://github.com/llvm/llvm-project/blob/main/openmp/runtime/src/kmp.h,
// description of kmp_tasking_flags struct.
enum {
  TiedFlag = 0x1,
  FinalFlag = 0x2,
  DestructorsFlag = 0x8,
  PriorityFlag = 0x20,
  DetachableFlag = 0x40,
};
unsigned Flags = Data.Tied ? TiedFlag : 0;
bool NeedsCleanup = false;
if (!Privates.empty()) {
  NeedsCleanup =
      checkDestructorsRequired(KmpTaskTWithPrivatesQTyRD, Privates);
  if (NeedsCleanup)
    Flags = Flags | DestructorsFlag;
}
if (Data.Priority.getInt())
  Flags = Flags | PriorityFlag;
if (D.hasClausesOfKind<OMPDetachClause>())
  Flags = Flags | DetachableFlag;
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));
llvm::Value *SharedsSize = CGM.getSize(C.getTypeSizeInChars(SharedsTy));
SmallVector<llvm::Value *, 8> AllocArgs = {emitUpdateLocation(CGF, Loc),
    getThreadID(CGF, Loc), TaskFlags, KmpTaskTWithPrivatesTySize,
    SharedsSize, CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        TaskEntry, KmpRoutineEntryPtrTy)};
llvm::Value *NewTask;
if (D.hasClausesOfKind<OMPNowaitClause>()) {
  // Check if we have any device clause associated with the directive.
  const Expr *Device = nullptr;
  if (auto *C = D.getSingleClause<OMPDeviceClause>())
    Device = C->getDevice();
  // Emit device ID if any otherwise use default value.
  llvm::Value *DeviceID;
  if (Device)
    DeviceID = CGF.Builder.CreateIntCast(CGF.EmitScalarExpr(Device),
                                         CGF.Int64Ty, /*isSigned=*/true);
  else
    DeviceID = CGF.Builder.getInt64(OMP_DEVICEID_UNDEF);
  AllocArgs.push_back(DeviceID);
  NewTask = CGF.EmitRuntimeCall(
      OMPBuilder.getOrCreateRuntimeFunction(
          CGM.getModule(), OMPRTL___kmpc_omp_target_task_alloc),
      AllocArgs);
} else {
  NewTask =
      CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                              CGM.getModule(), OMPRTL___kmpc_omp_task_alloc),
                          AllocArgs);
}
// Emit detach clause initialization.
// evt = (typeof(evt))__kmpc_task_allow_completion_event(loc, tid,
// task_descriptor);
if (const auto *DC = D.getSingleClause<OMPDetachClause>()) {
  const Expr *Evt = DC->getEventHandler()->IgnoreParenImpCasts();
  LValue EvtLVal = CGF.EmitLValue(Evt);

  // Build kmp_event_t *__kmpc_task_allow_completion_event(ident_t *loc_ref,
  // int gtid, kmp_task_t *task);
  llvm::Value *Loc = emitUpdateLocation(CGF, DC->getBeginLoc());
  llvm::Value *Tid = getThreadID(CGF, DC->getBeginLoc());
  Tid = CGF.Builder.CreateIntCast(Tid, CGF.IntTy, /*isSigned=*/false);
  llvm::Value *EvtVal = CGF.EmitRuntimeCall(
      OMPBuilder.getOrCreateRuntimeFunction(
          CGM.getModule(), OMPRTL___kmpc_task_allow_completion_event),
      {Loc, Tid, NewTask});
  EvtVal = CGF.EmitScalarConversion(EvtVal, C.VoidPtrTy, Evt->getType(),
                                    Evt->getExprLoc());
  CGF.EmitStoreOfScalar(EvtVal, EvtLVal);
}
// Process affinity clauses.
if (D.hasClausesOfKind<OMPAffinityClause>()) {
  // Process list of affinity data.
  ASTContext &C = CGM.getContext();
  Address AffinitiesArray = Address::invalid();
  // Calculate number of elements to form the array of affinity data.
  llvm::Value *NumOfElements = nullptr;
  unsigned NumAffinities = 0;
  for (const auto *C : D.getClausesOfKind<OMPAffinityClause>()) {
    if (const Expr *Modifier = C->getModifier()) {
      const auto *IE = cast<OMPIteratorExpr>(Modifier->IgnoreParenImpCasts());
      for (unsigned I = 0, E = IE->numOfIterators(); I < E; ++I) {
        llvm::Value *Sz = CGF.EmitScalarExpr(IE->getHelper(I).Upper);
        Sz = CGF.Builder.CreateIntCast(Sz, CGF.SizeTy, /*isSigned=*/false);
        NumOfElements =
            NumOfElements ? CGF.Builder.CreateNUWMul(NumOfElements, Sz) : Sz;
      }
    } else {
      NumAffinities += C->varlist_size();
    }
  }
  getKmpAffinityType(CGM.getContext(), KmpTaskAffinityInfoTy);
  // Fields ids in kmp_task_affinity_info record.
  enum RTLAffinityInfoFieldsTy { BaseAddr, Len, Flags };

  QualType KmpTaskAffinityInfoArrayTy;
  if (NumOfElements) {
    NumOfElements = CGF.Builder.CreateNUWAdd(
        llvm::ConstantInt::get(CGF.SizeTy, NumAffinities), NumOfElements);
    OpaqueValueExpr OVE(
        Loc,
        C.getIntTypeForBitwidth(C.getTypeSize(C.getSizeType()), /*Signed=*/0),
        VK_PRValue);
    CodeGenFunction::OpaqueValueMapping OpaqueMap(CGF, &OVE,
                                                  RValue::get(NumOfElements));
    KmpTaskAffinityInfoArrayTy =
        C.getVariableArrayType(KmpTaskAffinityInfoTy, &OVE, ArrayType::Normal,
                               /*IndexTypeQuals=*/0, SourceRange(Loc, Loc));
    // Properly emit variable-sized array.
    auto *PD = ImplicitParamDecl::Create(C, KmpTaskAffinityInfoArrayTy,
                                         ImplicitParamDecl::Other);
    CGF.EmitVarDecl(*PD);
    AffinitiesArray = CGF.GetAddrOfLocalVar(PD);
    NumOfElements = CGF.Builder.CreateIntCast(NumOfElements, CGF.Int32Ty,
                                              /*isSigned=*/false);
  } else {
    KmpTaskAffinityInfoArrayTy = C.getConstantArrayType(
        KmpTaskAffinityInfoTy,
        llvm::APInt(C.getTypeSize(C.getSizeType()), NumAffinities), nullptr,
        ArrayType::Normal, /*IndexTypeQuals=*/0);
    AffinitiesArray =
        CGF.CreateMemTemp(KmpTaskAffinityInfoArrayTy, ".affs.arr.addr");
    AffinitiesArray = CGF.Builder.CreateConstArrayGEP(AffinitiesArray, 0);
    NumOfElements = llvm::ConstantInt::get(CGM.Int32Ty, NumAffinities,
                                           /*isSigned=*/false);
  }

  const auto *KmpAffinityInfoRD = KmpTaskAffinityInfoTy->getAsRecordDecl();
  // Fill array by elements without iterators.
  unsigned Pos = 0;
  bool HasIterator = false;
  for (const auto *C : D.getClausesOfKind<OMPAffinityClause>()) {
    if (C->getModifier()) {
      HasIterator = true;
      continue;
    }
    for (const Expr *E : C->varlists()) {
      llvm::Value *Addr;
      llvm::Value *Size;
      std::tie(Addr, Size) = getPointerAndSize(CGF, E);
      LValue Base =
          CGF.MakeAddrLValue(CGF.Builder.CreateConstGEP(AffinitiesArray, Pos),
                             KmpTaskAffinityInfoTy);
      // affs[i].base_addr = &<Affinities[i].second>;
      LValue BaseAddrLVal = CGF.EmitLValueForField(
          Base, *std::next(KmpAffinityInfoRD->field_begin(), BaseAddr));
      CGF.EmitStoreOfScalar(CGF.Builder.CreatePtrToInt(Addr, CGF.IntPtrTy),
                            BaseAddrLVal);
      // affs[i].len = sizeof(<Affinities[i].second>);
      LValue LenLVal = CGF.EmitLValueForField(
          Base, *std::next(KmpAffinityInfoRD->field_begin(), Len));
      CGF.EmitStoreOfScalar(Size, LenLVal);
      ++Pos;
    }
  }
  LValue PosLVal;
  if (HasIterator) {
    PosLVal = CGF.MakeAddrLValue(
        CGF.CreateMemTemp(C.getSizeType(), "affs.counter.addr"),
        C.getSizeType());
    CGF.EmitStoreOfScalar(llvm::ConstantInt::get(CGF.SizeTy, Pos), PosLVal);
  }
  // Process elements with iterators.
  for (const auto *C : D.getClausesOfKind<OMPAffinityClause>()) {
    const Expr *Modifier = C->getModifier();
    if (!Modifier)
      continue;
    OMPIteratorGeneratorScope IteratorScope(
        CGF, cast_or_null<OMPIteratorExpr>(Modifier->IgnoreParenImpCasts()));
    for (const Expr *E : C->varlists()) {
      llvm::Value *Addr;
      llvm::Value *Size;
      std::tie(Addr, Size) = getPointerAndSize(CGF, E);
      llvm::Value *Idx = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
      LValue Base = CGF.MakeAddrLValue(
          Address(CGF.Builder.CreateGEP(AffinitiesArray.getElementType(),
                                        AffinitiesArray.getPointer(), Idx),
                  AffinitiesArray.getAlignment()),
          KmpTaskAffinityInfoTy);
      // affs[i].base_addr = &<Affinities[i].second>;
      LValue BaseAddrLVal = CGF.EmitLValueForField(
          Base, *std::next(KmpAffinityInfoRD->field_begin(), BaseAddr));
      CGF.EmitStoreOfScalar(CGF.Builder.CreatePtrToInt(Addr, CGF.IntPtrTy),
                            BaseAddrLVal);
      // affs[i].len = sizeof(<Affinities[i].second>);
      LValue LenLVal = CGF.EmitLValueForField(
          Base, *std::next(KmpAffinityInfoRD->field_begin(), Len));
      CGF.EmitStoreOfScalar(Size, LenLVal);
      Idx = CGF.Builder.CreateNUWAdd(
          Idx, llvm::ConstantInt::get(Idx->getType(), 1));
      CGF.EmitStoreOfScalar(Idx, PosLVal);
    }
  }
  // Call to kmp_int32 __kmpc_omp_reg_task_with_affinity(ident_t *loc_ref,
  // kmp_int32 gtid, kmp_task_t *new_task, kmp_int32
  // naffins, kmp_task_affinity_info_t *affin_list);
  llvm::Value *LocRef = emitUpdateLocation(CGF, Loc);
  llvm::Value *GTid = getThreadID(CGF, Loc);
  llvm::Value *AffinListPtr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
      AffinitiesArray.getPointer(), CGM.VoidPtrTy);
  // FIXME: Emit the function and ignore its result for now unless the
  // runtime function is properly implemented.
  (void)CGF.EmitRuntimeCall(
      OMPBuilder.getOrCreateRuntimeFunction(
          CGM.getModule(), OMPRTL___kmpc_omp_reg_task_with_affinity),
      {LocRef, GTid, NewTask, NumOfElements, AffinListPtr});
}
llvm::Value *NewTaskNewTaskTTy =
    CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        NewTask, KmpTaskTWithPrivatesPtrTy);
LValue Base = CGF.MakeNaturalAlignAddrLValue(NewTaskNewTaskTTy,
                                             KmpTaskTWithPrivatesQTy);
LValue TDBase =
    CGF.EmitLValueForField(Base, *KmpTaskTWithPrivatesQTyRD->field_begin());
// Fill the data in the resulting kmp_task_t record.
// Copy shareds if there are any.
Address KmpTaskSharedsPtr = Address::invalid();
if (!SharedsTy->getAsStructureType()->getDecl()->field_empty()) {
  KmpTaskSharedsPtr =
      Address(CGF.EmitLoadOfScalar(
                  CGF.EmitLValueForField(
                      TDBase, *std::next(KmpTaskTQTyRD->field_begin(),
                                         KmpTaskTShareds)),
                  Loc),
              CGM.getNaturalTypeAlignment(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;
if (!Privates.empty()) {
  emitPrivatesInit(CGF, D, KmpTaskSharedsPtr, Base, KmpTaskTWithPrivatesQTyRD,
                   SharedsTy, SharedsPtrTy, Data, Privates,
                   /*ForDup=*/false);
  if (isOpenMPTaskLoopDirective(D.getDirectiveKind()) &&
      (!Data.LastprivateVars.empty() || checkInitIsRequired(CGF, Privates))) {
    Result.TaskDupFn = emitTaskDupFunction(
        CGM, Loc, D, KmpTaskTWithPrivatesPtrQTy, KmpTaskTWithPrivatesQTyRD,
        KmpTaskTQTyRD, SharedsTy, SharedsPtrTy, Data, Privates,
        /*WithLastIter=*/!Data.LastprivateVars.empty());
  }
}
// Fields of union "kmp_cmplrdata_t" for destructors and priority.
enum { Priority = 0, Destructors = 1 };
// Provide pointer to function with destructors for privates.
auto FI = std::next(KmpTaskTQTyRD->field_begin(), Data1);
const RecordDecl *KmpCmplrdataUD =
    (*FI)->getType()->getAsUnionType()->getDecl();
if (NeedsCleanup) {
  llvm::Value *DestructorFn = emitDestructorsFunction(
      CGM, Loc, KmpInt32Ty, KmpTaskTWithPrivatesPtrQTy,
      KmpTaskTWithPrivatesQTy);
  LValue Data1LV = CGF.EmitLValueForField(TDBase, *FI);
  LValue DestructorsLV = CGF.EmitLValueForField(
      Data1LV, *std::next(KmpCmplrdataUD->field_begin(), Destructors));
  CGF.EmitStoreOfScalar(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
                            DestructorFn, KmpRoutineEntryPtrTy),
                        DestructorsLV);
}
// Set priority.
if (Data.Priority.getInt()) {
  LValue Data2LV = CGF.EmitLValueForField(
      TDBase, *std::next(KmpTaskTQTyRD->field_begin(), Data2));
  LValue PriorityLV = CGF.EmitLValueForField(
      Data2LV, *std::next(KmpCmplrdataUD->field_begin(), Priority));
  CGF.EmitStoreOfScalar(Data.Priority.getPointer(), PriorityLV);
}
Result.NewTask = NewTask;
Result.TaskEntry = TaskEntry;
Result.NewTaskNewTaskTTy = NewTaskNewTaskTTy;
Result.TDBase = TDBase;
Result.KmpTaskTQTyRD = KmpTaskTQTyRD;
return Result;
4579}

4581namespace {
4582/// Dependence kind for RTL.
4583enum RTLDependenceKindTy {
DepIn = 0x01,
DepInOut = 0x3,
DepMutexInOutSet = 0x4
4587};
4588/// Fields ids in kmp_depend_info record.
4589enum RTLDependInfoFieldsTy { BaseAddr, Len, Flags };
4590} // namespace

4592/// Translates internal dependency kind into the runtime kind.
4593static RTLDependenceKindTy translateDependencyKind(OpenMPDependClauseKind K) {
RTLDependenceKindTy DepKind;
switch (K) {
case OMPC_DEPEND_in:
  DepKind = DepIn;
  break;
// Out and InOut dependencies must use the same code.
case OMPC_DEPEND_out:
case OMPC_DEPEND_inout:
  DepKind = DepInOut;
  break;
case OMPC_DEPEND_mutexinoutset:
  DepKind = DepMutexInOutSet;
  break;
case OMPC_DEPEND_source:
case OMPC_DEPEND_sink:
case OMPC_DEPEND_depobj:
case OMPC_DEPEND_unknown:
  llvm_unreachable("Unknown task dependence type")__builtin_unreachable();
}
return DepKind;
4614}

4616/// Builds kmp_depend_info, if it is not built yet, and builds flags type.
4617static void getDependTypes(ASTContext &C, QualType &KmpDependInfoTy,
                         QualType &FlagsTy) {
FlagsTy = C.getIntTypeForBitwidth(C.getTypeSize(C.BoolTy), /*Signed=*/false);
if (KmpDependInfoTy.isNull()) {
  RecordDecl *KmpDependInfoRD = C.buildImplicitRecord("kmp_depend_info");
  KmpDependInfoRD->startDefinition();
  addFieldToRecordDecl(C, KmpDependInfoRD, C.getIntPtrType());
  addFieldToRecordDecl(C, KmpDependInfoRD, C.getSizeType());
  addFieldToRecordDecl(C, KmpDependInfoRD, FlagsTy);
  KmpDependInfoRD->completeDefinition();
  KmpDependInfoTy = C.getRecordType(KmpDependInfoRD);
}
4629}

4631std::pair<llvm::Value *, LValue>
4632CGOpenMPRuntime::getDepobjElements(CodeGenFunction &CGF, LValue DepobjLVal,
                                 SourceLocation Loc) {
ASTContext &C = CGM.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
LValue Base = CGF.EmitLoadOfPointerLValue(
    DepobjLVal.getAddress(CGF),
    C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
QualType KmpDependInfoPtrTy = C.getPointerType(KmpDependInfoTy);
Address Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        Base.getAddress(CGF), CGF.ConvertTypeForMem(KmpDependInfoPtrTy));
Base = CGF.MakeAddrLValue(Addr, KmpDependInfoTy, Base.getBaseInfo(),
                          Base.getTBAAInfo());
llvm::Value *DepObjAddr = CGF.Builder.CreateGEP(
    Addr.getElementType(), Addr.getPointer(),
    llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
LValue NumDepsBase = CGF.MakeAddrLValue(
    Address(DepObjAddr, Addr.getAlignment()), KmpDependInfoTy,
    Base.getBaseInfo(), Base.getTBAAInfo());
// NumDeps = deps[i].base_addr;
LValue BaseAddrLVal = CGF.EmitLValueForField(
    NumDepsBase, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
llvm::Value *NumDeps = CGF.EmitLoadOfScalar(BaseAddrLVal, Loc);
return std::make_pair(NumDeps, Base);
4658}

4660static void emitDependData(CodeGenFunction &CGF, QualType &KmpDependInfoTy,
                         llvm::PointerUnion<unsigned *, LValue *> Pos,
                         const OMPTaskDataTy::DependData &Data,
                         Address DependenciesArray) {
CodeGenModule &CGM = CGF.CGM;
ASTContext &C = CGM.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
llvm::Type *LLVMFlagsTy = CGF.ConvertTypeForMem(FlagsTy);

OMPIteratorGeneratorScope IteratorScope(
    CGF, cast_or_null<OMPIteratorExpr>(
             Data.IteratorExpr ? Data.IteratorExpr->IgnoreParenImpCasts()
                               : nullptr));
for (const Expr *E : Data.DepExprs) {
  llvm::Value *Addr;
  llvm::Value *Size;
  std::tie(Addr, Size) = getPointerAndSize(CGF, E);
  LValue Base;
  if (unsigned *P = Pos.dyn_cast<unsigned *>()) {
    Base = CGF.MakeAddrLValue(
        CGF.Builder.CreateConstGEP(DependenciesArray, *P), KmpDependInfoTy);
  } else {
    LValue &PosLVal = *Pos.get<LValue *>();
    llvm::Value *Idx = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
    Base = CGF.MakeAddrLValue(
        Address(CGF.Builder.CreateGEP(DependenciesArray.getElementType(),
                                      DependenciesArray.getPointer(), Idx),
                DependenciesArray.getAlignment()),
        KmpDependInfoTy);
  }
  // deps[i].base_addr = &<Dependencies[i].second>;
  LValue BaseAddrLVal = CGF.EmitLValueForField(
      Base, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
  CGF.EmitStoreOfScalar(CGF.Builder.CreatePtrToInt(Addr, CGF.IntPtrTy),
                        BaseAddrLVal);
  // deps[i].len = sizeof(<Dependencies[i].second>);
  LValue LenLVal = CGF.EmitLValueForField(
      Base, *std::next(KmpDependInfoRD->field_begin(), Len));
  CGF.EmitStoreOfScalar(Size, LenLVal);
  // deps[i].flags = <Dependencies[i].first>;
  RTLDependenceKindTy DepKind = translateDependencyKind(Data.DepKind);
  LValue FlagsLVal = CGF.EmitLValueForField(
      Base, *std::next(KmpDependInfoRD->field_begin(), Flags));
  CGF.EmitStoreOfScalar(llvm::ConstantInt::get(LLVMFlagsTy, DepKind),
                        FlagsLVal);
  if (unsigned *P = Pos.dyn_cast<unsigned *>()) {
    ++(*P);
  } else {
    LValue &PosLVal = *Pos.get<LValue *>();
    llvm::Value *Idx = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
    Idx = CGF.Builder.CreateNUWAdd(Idx,
                                   llvm::ConstantInt::get(Idx->getType(), 1));
    CGF.EmitStoreOfScalar(Idx, PosLVal);
  }
}
4718}

4720static SmallVector<llvm::Value *, 4>
4721emitDepobjElementsSizes(CodeGenFunction &CGF, QualType &KmpDependInfoTy,
                      const OMPTaskDataTy::DependData &Data) {
assert(Data.DepKind == OMPC_DEPEND_depobj &&((void)0)
       "Expected depobj dependecy kind.")((void)0);
SmallVector<llvm::Value *, 4> Sizes;
SmallVector<LValue, 4> SizeLVals;
ASTContext &C = CGF.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
QualType KmpDependInfoPtrTy = C.getPointerType(KmpDependInfoTy);
llvm::Type *KmpDependInfoPtrT = CGF.ConvertTypeForMem(KmpDependInfoPtrTy);
{
  OMPIteratorGeneratorScope IteratorScope(
      CGF, cast_or_null<OMPIteratorExpr>(
               Data.IteratorExpr ? Data.IteratorExpr->IgnoreParenImpCasts()
                                 : nullptr));
  for (const Expr *E : Data.DepExprs) {
    LValue DepobjLVal = CGF.EmitLValue(E->IgnoreParenImpCasts());
    LValue Base = CGF.EmitLoadOfPointerLValue(
        DepobjLVal.getAddress(CGF),
        C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
    Address Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        Base.getAddress(CGF), KmpDependInfoPtrT);
    Base = CGF.MakeAddrLValue(Addr, KmpDependInfoTy, Base.getBaseInfo(),
                              Base.getTBAAInfo());
    llvm::Value *DepObjAddr = CGF.Builder.CreateGEP(
        Addr.getElementType(), Addr.getPointer(),
        llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
    LValue NumDepsBase = CGF.MakeAddrLValue(
        Address(DepObjAddr, Addr.getAlignment()), KmpDependInfoTy,
        Base.getBaseInfo(), Base.getTBAAInfo());
    // NumDeps = deps[i].base_addr;
    LValue BaseAddrLVal = CGF.EmitLValueForField(
        NumDepsBase, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
    llvm::Value *NumDeps =
        CGF.EmitLoadOfScalar(BaseAddrLVal, E->getExprLoc());
    LValue NumLVal = CGF.MakeAddrLValue(
        CGF.CreateMemTemp(C.getUIntPtrType(), "depobj.size.addr"),
        C.getUIntPtrType());
    CGF.InitTempAlloca(NumLVal.getAddress(CGF),
                       llvm::ConstantInt::get(CGF.IntPtrTy, 0));
    llvm::Value *PrevVal = CGF.EmitLoadOfScalar(NumLVal, E->getExprLoc());
    llvm::Value *Add = CGF.Builder.CreateNUWAdd(PrevVal, NumDeps);
    CGF.EmitStoreOfScalar(Add, NumLVal);
    SizeLVals.push_back(NumLVal);
  }
}
for (unsigned I = 0, E = SizeLVals.size(); I < E; ++I) {
  llvm::Value *Size =
      CGF.EmitLoadOfScalar(SizeLVals[I], Data.DepExprs[I]->getExprLoc());
  Sizes.push_back(Size);
}
return Sizes;
4776}

4778static void emitDepobjElements(CodeGenFunction &CGF, QualType &KmpDependInfoTy,
                             LValue PosLVal,
                             const OMPTaskDataTy::DependData &Data,
                             Address DependenciesArray) {
assert(Data.DepKind == OMPC_DEPEND_depobj &&((void)0)
       "Expected depobj dependecy kind.")((void)0);
ASTContext &C = CGF.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
QualType KmpDependInfoPtrTy = C.getPointerType(KmpDependInfoTy);
llvm::Type *KmpDependInfoPtrT = CGF.ConvertTypeForMem(KmpDependInfoPtrTy);
llvm::Value *ElSize = CGF.getTypeSize(KmpDependInfoTy);
{
  OMPIteratorGeneratorScope IteratorScope(
      CGF, cast_or_null<OMPIteratorExpr>(
               Data.IteratorExpr ? Data.IteratorExpr->IgnoreParenImpCasts()
                                 : nullptr));
  for (unsigned I = 0, End = Data.DepExprs.size(); I < End; ++I) {
    const Expr *E = Data.DepExprs[I];
    LValue DepobjLVal = CGF.EmitLValue(E->IgnoreParenImpCasts());
    LValue Base = CGF.EmitLoadOfPointerLValue(
        DepobjLVal.getAddress(CGF),
        C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
    Address Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
        Base.getAddress(CGF), KmpDependInfoPtrT);
    Base = CGF.MakeAddrLValue(Addr, KmpDependInfoTy, Base.getBaseInfo(),
                              Base.getTBAAInfo());

    // Get number of elements in a single depobj.
    llvm::Value *DepObjAddr = CGF.Builder.CreateGEP(
        Addr.getElementType(), Addr.getPointer(),
        llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
    LValue NumDepsBase = CGF.MakeAddrLValue(
        Address(DepObjAddr, Addr.getAlignment()), KmpDependInfoTy,
        Base.getBaseInfo(), Base.getTBAAInfo());
    // NumDeps = deps[i].base_addr;
    LValue BaseAddrLVal = CGF.EmitLValueForField(
        NumDepsBase, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
    llvm::Value *NumDeps =
        CGF.EmitLoadOfScalar(BaseAddrLVal, E->getExprLoc());

    // memcopy dependency data.
    llvm::Value *Size = CGF.Builder.CreateNUWMul(
        ElSize,
        CGF.Builder.CreateIntCast(NumDeps, CGF.SizeTy, /*isSigned=*/false));
    llvm::Value *Pos = CGF.EmitLoadOfScalar(PosLVal, E->getExprLoc());
    Address DepAddr =
        Address(CGF.Builder.CreateGEP(DependenciesArray.getElementType(),
                                      DependenciesArray.getPointer(), Pos),
                DependenciesArray.getAlignment());
    CGF.Builder.CreateMemCpy(DepAddr, Base.getAddress(CGF), Size);

    // Increase pos.
    // pos += size;
    llvm::Value *Add = CGF.Builder.CreateNUWAdd(Pos, NumDeps);
    CGF.EmitStoreOfScalar(Add, PosLVal);
  }
}
4838}

4840std::pair<llvm::Value *, Address> CGOpenMPRuntime::emitDependClause(
  CodeGenFunction &CGF, ArrayRef<OMPTaskDataTy::DependData> Dependencies,
  SourceLocation Loc) {
if (llvm::all_of(Dependencies, [](const OMPTaskDataTy::DependData &D) {
      return D.DepExprs.empty();
    }))
  return std::make_pair(nullptr, Address::invalid());
// Process list of dependencies.
ASTContext &C = CGM.getContext();
Address DependenciesArray = Address::invalid();
llvm::Value *NumOfElements = nullptr;
unsigned NumDependencies = std::accumulate(
    Dependencies.begin(), Dependencies.end(), 0,
    [](unsigned V, const OMPTaskDataTy::DependData &D) {
      return D.DepKind == OMPC_DEPEND_depobj
                 ? V
                 : (V + (D.IteratorExpr ? 0 : D.DepExprs.size()));
    });
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
bool HasDepobjDeps = false;
bool HasRegularWithIterators = false;
llvm::Value *NumOfDepobjElements = llvm::ConstantInt::get(CGF.IntPtrTy, 0);
llvm::Value *NumOfRegularWithIterators =
    llvm::ConstantInt::get(CGF.IntPtrTy, 1);
// Calculate number of depobj dependecies and regular deps with the iterators.
for (const OMPTaskDataTy::DependData &D : Dependencies) {
  if (D.DepKind == OMPC_DEPEND_depobj) {
    SmallVector<llvm::Value *, 4> Sizes =
        emitDepobjElementsSizes(CGF, KmpDependInfoTy, D);
    for (llvm::Value *Size : Sizes) {
      NumOfDepobjElements =
          CGF.Builder.CreateNUWAdd(NumOfDepobjElements, Size);
    }
    HasDepobjDeps = true;
    continue;
  }
  // Include number of iterations, if any.
  if (const auto *IE = cast_or_null<OMPIteratorExpr>(D.IteratorExpr)) {
    for (unsigned I = 0, E = IE->numOfIterators(); I < E; ++I) {
      llvm::Value *Sz = CGF.EmitScalarExpr(IE->getHelper(I).Upper);
      Sz = CGF.Builder.CreateIntCast(Sz, CGF.IntPtrTy, /*isSigned=*/false);
      NumOfRegularWithIterators =
          CGF.Builder.CreateNUWMul(NumOfRegularWithIterators, Sz);
    }
    HasRegularWithIterators = true;
    continue;
  }
}

QualType KmpDependInfoArrayTy;
if (HasDepobjDeps || HasRegularWithIterators) {
  NumOfElements = llvm::ConstantInt::get(CGM.IntPtrTy, NumDependencies,
                                         /*isSigned=*/false);
  if (HasDepobjDeps) {
    NumOfElements =
        CGF.Builder.CreateNUWAdd(NumOfDepobjElements, NumOfElements);
  }
  if (HasRegularWithIterators) {
    NumOfElements =
        CGF.Builder.CreateNUWAdd(NumOfRegularWithIterators, NumOfElements);
  }
  OpaqueValueExpr OVE(Loc,
                      C.getIntTypeForBitwidth(/*DestWidth=*/64, /*Signed=*/0),
                      VK_PRValue);
  CodeGenFunction::OpaqueValueMapping OpaqueMap(CGF, &OVE,
                                                RValue::get(NumOfElements));
  KmpDependInfoArrayTy =
      C.getVariableArrayType(KmpDependInfoTy, &OVE, ArrayType::Normal,
                             /*IndexTypeQuals=*/0, SourceRange(Loc, Loc));
  // CGF.EmitVariablyModifiedType(KmpDependInfoArrayTy);
  // Properly emit variable-sized array.
  auto *PD = ImplicitParamDecl::Create(C, KmpDependInfoArrayTy,
                                       ImplicitParamDecl::Other);
  CGF.EmitVarDecl(*PD);
  DependenciesArray = CGF.GetAddrOfLocalVar(PD);
  NumOfElements = CGF.Builder.CreateIntCast(NumOfElements, CGF.Int32Ty,
                                            /*isSigned=*/false);
} else {
  KmpDependInfoArrayTy = C.getConstantArrayType(
      KmpDependInfoTy, llvm::APInt(/*numBits=*/64, NumDependencies), nullptr,
      ArrayType::Normal, /*IndexTypeQuals=*/0);
  DependenciesArray =
      CGF.CreateMemTemp(KmpDependInfoArrayTy, ".dep.arr.addr");
  DependenciesArray = CGF.Builder.CreateConstArrayGEP(DependenciesArray, 0);
  NumOfElements = llvm::ConstantInt::get(CGM.Int32Ty, NumDependencies,
                                         /*isSigned=*/false);
}
unsigned Pos = 0;
for (unsigned I = 0, End = Dependencies.size(); I < End; ++I) {
  if (Dependencies[I].DepKind == OMPC_DEPEND_depobj ||
      Dependencies[I].IteratorExpr)
    continue;
  emitDependData(CGF, KmpDependInfoTy, &Pos, Dependencies[I],
                 DependenciesArray);
}
// Copy regular dependecies with iterators.
LValue PosLVal = CGF.MakeAddrLValue(
    CGF.CreateMemTemp(C.getSizeType(), "dep.counter.addr"), C.getSizeType());
CGF.EmitStoreOfScalar(llvm::ConstantInt::get(CGF.SizeTy, Pos), PosLVal);
for (unsigned I = 0, End = Dependencies.size(); I < End; ++I) {
  if (Dependencies[I].DepKind == OMPC_DEPEND_depobj ||
      !Dependencies[I].IteratorExpr)
    continue;
  emitDependData(CGF, KmpDependInfoTy, &PosLVal, Dependencies[I],
                 DependenciesArray);
}
// Copy final depobj arrays without iterators.
if (HasDepobjDeps) {
  for (unsigned I = 0, End = Dependencies.size(); I < End; ++I) {
    if (Dependencies[I].DepKind != OMPC_DEPEND_depobj)
      continue;
    emitDepobjElements(CGF, KmpDependInfoTy, PosLVal, Dependencies[I],
                       DependenciesArray);
  }
}
DependenciesArray = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    DependenciesArray, CGF.VoidPtrTy);
return std::make_pair(NumOfElements, DependenciesArray);
4959}

4961Address CGOpenMPRuntime::emitDepobjDependClause(
  CodeGenFunction &CGF, const OMPTaskDataTy::DependData &Dependencies,
  SourceLocation Loc) {
if (Dependencies.DepExprs.empty())
  return Address::invalid();
// Process list of dependencies.
ASTContext &C = CGM.getContext();
Address DependenciesArray = Address::invalid();
unsigned NumDependencies = Dependencies.DepExprs.size();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());

llvm::Value *Size;
// Define type kmp_depend_info[<Dependencies.size()>];
// For depobj reserve one extra element to store the number of elements.
// It is required to handle depobj(x) update(in) construct.
// kmp_depend_info[<Dependencies.size()>] deps;
llvm::Value *NumDepsVal;
CharUnits Align = C.getTypeAlignInChars(KmpDependInfoTy);
if (const auto *IE =
        cast_or_null<OMPIteratorExpr>(Dependencies.IteratorExpr)) {
  NumDepsVal = llvm::ConstantInt::get(CGF.SizeTy, 1);
  for (unsigned I = 0, E = IE->numOfIterators(); I < E; ++I) {
    llvm::Value *Sz = CGF.EmitScalarExpr(IE->getHelper(I).Upper);
    Sz = CGF.Builder.CreateIntCast(Sz, CGF.SizeTy, /*isSigned=*/false);
    NumDepsVal = CGF.Builder.CreateNUWMul(NumDepsVal, Sz);
  }
  Size = CGF.Builder.CreateNUWAdd(llvm::ConstantInt::get(CGF.SizeTy, 1),
                                  NumDepsVal);
  CharUnits SizeInBytes =
      C.getTypeSizeInChars(KmpDependInfoTy).alignTo(Align);
  llvm::Value *RecSize = CGM.getSize(SizeInBytes);
  Size = CGF.Builder.CreateNUWMul(Size, RecSize);
  NumDepsVal =
      CGF.Builder.CreateIntCast(NumDepsVal, CGF.IntPtrTy, /*isSigned=*/false);
} else {
  QualType KmpDependInfoArrayTy = C.getConstantArrayType(
      KmpDependInfoTy, llvm::APInt(/*numBits=*/64, NumDependencies + 1),
      nullptr, ArrayType::Normal, /*IndexTypeQuals=*/0);
  CharUnits Sz = C.getTypeSizeInChars(KmpDependInfoArrayTy);
  Size = CGM.getSize(Sz.alignTo(Align));
  NumDepsVal = llvm::ConstantInt::get(CGF.IntPtrTy, NumDependencies);
}
// Need to allocate on the dynamic memory.
llvm::Value *ThreadID = getThreadID(CGF, Loc);
// Use default allocator.
llvm::Value *Allocator = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
llvm::Value *Args[] = {ThreadID, Size, Allocator};

llvm::Value *Addr =
    CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                            CGM.getModule(), OMPRTL___kmpc_alloc),
                        Args, ".dep.arr.addr");
Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    Addr, CGF.ConvertTypeForMem(KmpDependInfoTy)->getPointerTo());
DependenciesArray = Address(Addr, Align);
// Write number of elements in the first element of array for depobj.
LValue Base = CGF.MakeAddrLValue(DependenciesArray, KmpDependInfoTy);
// deps[i].base_addr = NumDependencies;
LValue BaseAddrLVal = CGF.EmitLValueForField(
    Base, *std::next(KmpDependInfoRD->field_begin(), BaseAddr));
CGF.EmitStoreOfScalar(NumDepsVal, BaseAddrLVal);
llvm::PointerUnion<unsigned *, LValue *> Pos;
unsigned Idx = 1;
LValue PosLVal;
if (Dependencies.IteratorExpr) {
  PosLVal = CGF.MakeAddrLValue(
      CGF.CreateMemTemp(C.getSizeType(), "iterator.counter.addr"),
      C.getSizeType());
  CGF.EmitStoreOfScalar(llvm::ConstantInt::get(CGF.SizeTy, Idx), PosLVal,
                        /*IsInit=*/true);
  Pos = &PosLVal;
} else {
  Pos = &Idx;
}
emitDependData(CGF, KmpDependInfoTy, Pos, Dependencies, DependenciesArray);
DependenciesArray = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    CGF.Builder.CreateConstGEP(DependenciesArray, 1), CGF.VoidPtrTy);
return DependenciesArray;
5042}

5044void CGOpenMPRuntime::emitDestroyClause(CodeGenFunction &CGF, LValue DepobjLVal,
                                      SourceLocation Loc) {
ASTContext &C = CGM.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
LValue Base = CGF.EmitLoadOfPointerLValue(
    DepobjLVal.getAddress(CGF),
    C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
QualType KmpDependInfoPtrTy = C.getPointerType(KmpDependInfoTy);
Address Addr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    Base.getAddress(CGF), CGF.ConvertTypeForMem(KmpDependInfoPtrTy));
llvm::Value *DepObjAddr = CGF.Builder.CreateGEP(
    Addr.getElementType(), Addr.getPointer(),
    llvm::ConstantInt::get(CGF.IntPtrTy, -1, /*isSigned=*/true));
DepObjAddr = CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(DepObjAddr,
                                                             CGF.VoidPtrTy);
llvm::Value *ThreadID = getThreadID(CGF, Loc);
// Use default allocator.
llvm::Value *Allocator = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
llvm::Value *Args[] = {ThreadID, DepObjAddr, Allocator};

// _kmpc_free(gtid, addr, nullptr);
(void)CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                              CGM.getModule(), OMPRTL___kmpc_free),
                          Args);
5069}

5071void CGOpenMPRuntime::emitUpdateClause(CodeGenFunction &CGF, LValue DepobjLVal,
                                     OpenMPDependClauseKind NewDepKind,
                                     SourceLocation Loc) {
ASTContext &C = CGM.getContext();
QualType FlagsTy;
getDependTypes(C, KmpDependInfoTy, FlagsTy);
RecordDecl *KmpDependInfoRD =
    cast<RecordDecl>(KmpDependInfoTy->getAsTagDecl());
llvm::Type *LLVMFlagsTy = CGF.ConvertTypeForMem(FlagsTy);
llvm::Value *NumDeps;
LValue Base;
std::tie(NumDeps, Base) = getDepobjElements(CGF, DepobjLVal, Loc);

Address Begin = Base.getAddress(CGF);
// Cast from pointer to array type to pointer to single element.
llvm::Value *End = CGF.Builder.CreateGEP(
    Begin.getElementType(), Begin.getPointer(), NumDeps);
// The basic structure here is a while-do loop.
llvm::BasicBlock *BodyBB = CGF.createBasicBlock("omp.body");
llvm::BasicBlock *DoneBB = CGF.createBasicBlock("omp.done");
llvm::BasicBlock *EntryBB = CGF.Builder.GetInsertBlock();
CGF.EmitBlock(BodyBB);
llvm::PHINode *ElementPHI =
    CGF.Builder.CreatePHI(Begin.getType(), 2, "omp.elementPast");
ElementPHI->addIncoming(Begin.getPointer(), EntryBB);
Begin = Address(ElementPHI, Begin.getAlignment());
Base = CGF.MakeAddrLValue(Begin, KmpDependInfoTy, Base.getBaseInfo(),
                          Base.getTBAAInfo());
// deps[i].flags = NewDepKind;
RTLDependenceKindTy DepKind = translateDependencyKind(NewDepKind);
LValue FlagsLVal = CGF.EmitLValueForField(
    Base, *std::next(KmpDependInfoRD->field_begin(), Flags));
CGF.EmitStoreOfScalar(llvm::ConstantInt::get(LLVMFlagsTy, DepKind),
                      FlagsLVal);

// Shift the address forward by one element.
Address ElementNext =
    CGF.Builder.CreateConstGEP(Begin, /*Index=*/1, "omp.elementNext");
ElementPHI->addIncoming(ElementNext.getPointer(),
                        CGF.Builder.GetInsertBlock());
llvm::Value *IsEmpty =
    CGF.Builder.CreateICmpEQ(ElementNext.getPointer(), End, "omp.isempty");
CGF.Builder.CreateCondBr(IsEmpty, DoneBB, BodyBB);
// Done.
CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
5116}

5118void CGOpenMPRuntime::emitTaskCall(CodeGenFunction &CGF, SourceLocation Loc,
                                 const OMPExecutableDirective &D,
                                 llvm::Function *TaskFunction,
                                 QualType SharedsTy, Address Shareds,
                                 const Expr *IfCond,
                                 const OMPTaskDataTy &Data) {
if (!CGF.HaveInsertPoint())
  return;

TaskResultTy Result =
    emitTaskInit(CGF, Loc, D, TaskFunction, SharedsTy, Shareds, Data);
llvm::Value *NewTask = Result.NewTask;
llvm::Function *TaskEntry = Result.TaskEntry;
llvm::Value *NewTaskNewTaskTTy = Result.NewTaskNewTaskTTy;
LValue TDBase = Result.TDBase;
const RecordDecl *KmpTaskTQTyRD = Result.KmpTaskTQTyRD;
// Process list of dependences.
Address DependenciesArray = Address::invalid();
llvm::Value *NumOfElements;
std::tie(NumOfElements, DependenciesArray) =
    emitDependClause(CGF, Data.Dependences, Loc);

// 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
llvm::Value *ThreadID = getThreadID(CGF, Loc);
llvm::Value *UpLoc = emitUpdateLocation(CGF, Loc);
llvm::Value *TaskArgs[] = { UpLoc, ThreadID, NewTask };
llvm::Value *DepTaskArgs[7];
if (!Data.Dependences.empty()) {
  DepTaskArgs[0] = UpLoc;
  DepTaskArgs[1] = ThreadID;
  DepTaskArgs[2] = NewTask;
  DepTaskArgs[3] = NumOfElements;
  DepTaskArgs[4] = DependenciesArray.getPointer();
  DepTaskArgs[5] = CGF.Builder.getInt32(0);
  DepTaskArgs[6] = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
}
auto &&ThenCodeGen = [this, &Data, TDBase, KmpTaskTQTyRD, &TaskArgs,
                      &DepTaskArgs](CodeGenFunction &CGF, PrePostActionTy &) {
  if (!Data.Tied) {
    auto PartIdFI = std::next(KmpTaskTQTyRD->field_begin(), KmpTaskTPartId);
    LValue PartIdLVal = CGF.EmitLValueForField(TDBase, *PartIdFI);
    CGF.EmitStoreOfScalar(CGF.Builder.getInt32(0), PartIdLVal);
  }
  if (!Data.Dependences.empty()) {
    CGF.EmitRuntimeCall(
        OMPBuilder.getOrCreateRuntimeFunction(
            CGM.getModule(), OMPRTL___kmpc_omp_task_with_deps),
        DepTaskArgs);
  } else {
    CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                            CGM.getModule(), OMPRTL___kmpc_omp_task),
                        TaskArgs);
  }
  // Check if parent region is untied and build return for untied task;
  if (auto *Region =
          dyn_cast_or_null<CGOpenMPRegionInfo>(CGF.CapturedStmtInfo))
    Region->emitUntiedSwitch(CGF);
};

llvm::Value *DepWaitTaskArgs[6];
if (!Data.Dependences.empty()) {
  DepWaitTaskArgs[0] = UpLoc;
  DepWaitTaskArgs[1] = ThreadID;
  DepWaitTaskArgs[2] = NumOfElements;
  DepWaitTaskArgs[3] = DependenciesArray.getPointer();
  DepWaitTaskArgs[4] = CGF.Builder.getInt32(0);
  DepWaitTaskArgs[5] = llvm::ConstantPointerNull::get(CGF.VoidPtrTy);
}
auto &M = CGM.getModule();
auto &&ElseCodeGen = [this, &M, &TaskArgs, ThreadID, NewTaskNewTaskTTy,
                      TaskEntry, &Data, &DepWaitTaskArgs,
                      Loc](CodeGenFunction &CGF, PrePostActionTy &) {
  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
  // ndeps_noalias, kmp_depend_info_t *noalias_dep_list); if dependence info
  // is specified.
  if (!Data.Dependences.empty())
    CGF.EmitRuntimeCall(
        OMPBuilder.getOrCreateRuntimeFunction(M, OMPRTL___kmpc_omp_wait_deps),
        DepWaitTaskArgs);
  // Call proxy_task_entry(gtid, new_task);
  auto &&CodeGen = [TaskEntry, ThreadID, NewTaskNewTaskTTy,
                    Loc](CodeGenFunction &CGF, PrePostActionTy &Action) {
    Action.Enter(CGF);
    llvm::Value *OutlinedFnArgs[] = {ThreadID, NewTaskNewTaskTTy};
    CGF.CGM.getOpenMPRuntime().emitOutlinedFunctionCall(CGF, Loc, TaskEntry,
                                                        OutlinedFnArgs);
  };

  // Build void __kmpc_omp_task_begin_if0(ident_t *, kmp_int32 gtid,
  // kmp_task_t *new_task);
  // Build void __kmpc_omp_task_complete_if0(ident_t *, kmp_int32 gtid,
  // kmp_task_t *new_task);
  RegionCodeGenTy RCG(CodeGen);
  CommonActionTy Action(OMPBuilder.getOrCreateRuntimeFunction(
                            M, OMPRTL___kmpc_omp_task_begin_if0),
                        TaskArgs,
                        OMPBuilder.getOrCreateRuntimeFunction(
                            M, OMPRTL___kmpc_omp_task_complete_if0),
                        TaskArgs);
  RCG.setAction(Action);
  RCG(CGF);
};

if (IfCond) {
  emitIfClause(CGF, IfCond, ThenCodeGen, ElseCodeGen);
} else {
  RegionCodeGenTy ThenRCG(ThenCodeGen);
  ThenRCG(CGF);
}
5234}

5236void CGOpenMPRuntime::emitTaskLoopCall(CodeGenFunction &CGF, SourceLocation Loc,
                                     const OMPLoopDirective &D,
                                     llvm::Function *TaskFunction,
                                     QualType SharedsTy, Address Shareds,
                                     const Expr *IfCond,
                                     const OMPTaskDataTy &Data) {
if (!CGF.HaveInsertPoint())
  return;
TaskResultTy Result =
    emitTaskInit(CGF, Loc, D, TaskFunction, SharedsTy, Shareds, Data);
// 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
// sched, kmp_uint64 grainsize, void *task_dup);
llvm::Value *ThreadID = getThreadID(CGF, Loc);
llvm::Value *UpLoc = emitUpdateLocation(CGF, Loc);
llvm::Value *IfVal;
if (IfCond) {
  IfVal = CGF.Builder.CreateIntCast(CGF.EvaluateExprAsBool(IfCond), CGF.IntTy,
                                    /*isSigned=*/true);
} else {
  IfVal = llvm::ConstantInt::getSigned(CGF.IntTy, /*V=*/1);
}

LValue LBLVal = CGF.EmitLValueForField(
    Result.TDBase,
    *std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTLowerBound));
const auto *LBVar =
    cast<VarDecl>(cast<DeclRefExpr>(D.getLowerBoundVariable())->getDecl());
CGF.EmitAnyExprToMem(LBVar->getInit(), LBLVal.getAddress(CGF),
                     LBLVal.getQuals(),
                     /*IsInitializer=*/true);
LValue UBLVal = CGF.EmitLValueForField(
    Result.TDBase,
    *std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTUpperBound));
const auto *UBVar =
    cast<VarDecl>(cast<DeclRefExpr>(D.getUpperBoundVariable())->getDecl());
CGF.EmitAnyExprToMem(UBVar->getInit(), UBLVal.getAddress(CGF),
                     UBLVal.getQuals(),
                     /*IsInitializer=*/true);
LValue StLVal = CGF.EmitLValueForField(
    Result.TDBase,
    *std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTStride));
const auto *StVar =
    cast<VarDecl>(cast<DeclRefExpr>(D.getStrideVariable())->getDecl());
CGF.EmitAnyExprToMem(StVar->getInit(), StLVal.getAddress(CGF),
                     StLVal.getQuals(),
                     /*IsInitializer=*/true);
// Store reductions address.
LValue RedLVal = CGF.EmitLValueForField(
    Result.TDBase,
    *std::next(Result.KmpTaskTQTyRD->field_begin(), KmpTaskTReductions));
if (Data.Reductions) {
  CGF.EmitStoreOfScalar(Data.Reductions, RedLVal);
} else {
  CGF.EmitNullInitialization(RedLVal.getAddress(CGF),
                             CGF.getContext().VoidPtrTy);
}
enum { NoSchedule = 0, Grainsize = 1, NumTasks = 2 };
llvm::Value *TaskArgs[] = {
    UpLoc,
    ThreadID,
    Result.NewTask,
    IfVal,
    LBLVal.getPointer(CGF),
    UBLVal.getPointer(CGF),
    CGF.EmitLoadOfScalar(StLVal, Loc),
    llvm::ConstantInt::getSigned(
        CGF.IntTy, 1), // Always 1 because taskgroup emitted by the compiler
    llvm::ConstantInt::getSigned(
        CGF.IntTy, Data.Schedule.getPointer()
                       ? Data.Schedule.getInt() ? NumTasks : Grainsize
                       : NoSchedule),
    Data.Schedule.getPointer()
        ? CGF.Builder.CreateIntCast(Data.Schedule.getPointer(), CGF.Int64Ty,
                                    /*isSigned=*/false)
        : llvm::ConstantInt::get(CGF.Int64Ty, /*V=*/0),
    Result.TaskDupFn ? CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
                           Result.TaskDupFn, CGF.VoidPtrTy)
                     : llvm::ConstantPointerNull::get(CGF.VoidPtrTy)};
CGF.EmitRuntimeCall(OMPBuilder.getOrCreateRuntimeFunction(
                        CGM.getModule(), OMPRTL___kmpc_taskloop),
                    TaskArgs);
5320}

5322/// Emit reduction operation for each element of array (required for
5323/// array sections) LHS op = RHS.
5324/// \param Type Type of array.
5325/// \param LHSVar Variable on the left side of the reduction operation
5326/// (references element of array in original variable).
5327/// \param RHSVar Variable on the right side of the reduction operation
5328/// (references element of array in original variable).
5329/// \param RedOpGen Generator of reduction operation with use of LHSVar and
5330/// RHSVar.
5331static void EmitOMPAggregateReduction(
  CodeGenFunction &CGF, QualType Type, const VarDecl *LHSVar,
  const VarDecl *RHSVar,
  const llvm::function_ref<void(CodeGenFunction &CGF, const Expr *,
                                const Expr *, const Expr *)> &RedOpGen,
  const Expr *XExpr = nullptr, const Expr *EExpr = nullptr,
  const Expr *UpExpr = nullptr) {
// Perform element-by-element initialization.
QualType ElementTy;
Address LHSAddr = CGF.GetAddrOfLocalVar(LHSVar);
Address RHSAddr = CGF.GetAddrOfLocalVar(RHSVar);

// Drill down to the base element type on both arrays.
const ArrayType *ArrayTy = Type->getAsArrayTypeUnsafe();
llvm::Value *NumElements = CGF.emitArrayLength(ArrayTy, ElementTy, LHSAddr);

llvm::Value *RHSBegin = RHSAddr.getPointer();
llvm::Value *LHSBegin = LHSAddr.getPointer();
// Cast from pointer to array type to pointer to single element.
llvm::Value *LHSEnd =
    CGF.Builder.CreateGEP(LHSAddr.getElementType(), LHSBegin, NumElements);
// The basic structure here is a while-do loop.
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.
llvm::BasicBlock *EntryBB = CGF.Builder.GetInsertBlock();
CGF.EmitBlock(BodyBB);

CharUnits ElementSize = CGF.getContext().getTypeSizeInChars(ElementTy);

llvm::PHINode *RHSElementPHI = CGF.Builder.CreatePHI(
    RHSBegin->getType(), 2, "omp.arraycpy.srcElementPast");
RHSElementPHI->addIncoming(RHSBegin, EntryBB);
Address RHSElementCurrent =
    Address(RHSElementPHI,
            RHSAddr.getAlignment().alignmentOfArrayElement(ElementSize));

llvm::PHINode *LHSElementPHI = CGF.Builder.CreatePHI(
    LHSBegin->getType(), 2, "omp.arraycpy.destElementPast");
LHSElementPHI->addIncoming(LHSBegin, EntryBB);
Address LHSElementCurrent =
    Address(LHSElementPHI,
            LHSAddr.getAlignment().alignmentOfArrayElement(ElementSize));

// Emit copy.
CodeGenFunction::OMPPrivateScope Scope(CGF);
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.
llvm::Value *LHSElementNext = CGF.Builder.CreateConstGEP1_32(
    LHSAddr.getElementType(), LHSElementPHI, /*Idx0=*/1,
    "omp.arraycpy.dest.element");
llvm::Value *RHSElementNext = CGF.Builder.CreateConstGEP1_32(
    RHSAddr.getElementType(), RHSElementPHI, /*Idx0=*/1,
    "omp.arraycpy.src.element");
// Check whether we've reached the end.
llvm::Value *Done =
    CGF.Builder.CreateICmpEQ(LHSElementNext, LHSEnd, "omp.arraycpy.done");
CGF.Builder.CreateCondBr(Done, DoneBB, BodyBB);
LHSElementPHI->addIncoming(LHSElementNext, CGF.Builder.GetInsertBlock());
RHSElementPHI->addIncoming(RHSElementNext, CGF.Builder.GetInsertBlock());

// Done.
CGF.EmitBlock(DoneBB, /*IsFinished=*/true);
5403}

5405/// Emit reduction combiner. If the combiner is a simple expression emit it as
5406/// is, otherwise consider it as combiner of UDR decl and emit it as a call of
5407/// UDR combiner function.
5408static void emitReductionCombiner(CodeGenFunction &CGF,
                                const Expr *ReductionOp) {
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 (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);
        CodeGenFunction::OpaqueValueMapping Map(CGF, OVE, Func);
        CGF.EmitIgnoredExpr(ReductionOp);
        return;
      }
CGF.EmitIgnoredExpr(ReductionOp);
5424}

5426llvm::Function *CGOpenMPRuntime::emitReductionFunction(
  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, /*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);
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, Loc, Loc);

// Dst = (void*[n])(LHSArg);
// Src = (void*[n])(RHSArg);
Address LHS(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    CGF.Builder.CreateLoad(CGF.GetAddrOfLocalVar(&LHSArg)),
    ArgsType), CGF.getPointerAlign());
Address RHS(CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
    CGF.Builder.CreateLoad(CGF.GetAddrOfLocalVar(&RHSArg)),
    ArgsType), CGF.getPointerAlign());

//  ...
//  *(Type<i>*)lhs[i] = RedOp<i>(*(Type<i>*)lhs[i], *(Type<i>*)rhs[i]);
//  ...
CodeGenFunction::OMPPrivateScope Scope(CGF);
auto IPriv = Privates.begin();
unsigned Idx = 0;
for (unsigned I = 0, E = ReductionOps.size(); I < E; ++I, ++IPriv, ++Idx) {
  const auto *RHSVar =
      cast<VarDecl>(cast<DeclRefExpr>(RHSExprs[I])->getDecl());
  Scope.addPrivate(RHSVar, [&CGF, RHS, Idx, RHSVar]() {
    return emitAddrOfVarFromArray(CGF, RHS, Idx, RHSVar);
  });
  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();
  if (PrivTy->isVariablyModifiedType()) {
    // Get array size and emit VLA type.
    ++Idx;
    Address Elem = CGF.Builder.CreateConstArrayGEP(LHS, Idx);
    llvm::Value *Ptr = CGF.Builder.CreateLoad(Elem);
    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);
  }
}
Scope.Privatize();
IPriv = Privates.begin();
auto ILHS = LHSExprs.begin();
auto IRHS = RHSExprs.begin();
for (const Expr *E : ReductionOps) {
  if ((*IPriv)->getType()->isArrayType()) {
    // Emit reduction for array section.
    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 {
    // Emit reduction for array subscript or single variable.
    emitReductionCombiner(CGF, E);
  }
  ++IPriv;
  ++ILHS;
  ++IRHS;
}
Scope.ForceCleanup();
CGF.FinishFunction();
return Fn;
5516}

5518void CGOpenMPRuntime::emitSingleReductionCombiner(CodeGenFunction &CGF,
                                                const Expr *ReductionOp,
                                                const Expr *PrivateRef,
                                                const DeclRefExpr *LHS,
                                                const DeclRefExpr *RHS) {
if (PrivateRef->getType()->isArrayType()) {
  // Emit reduction for array section.
  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 {
  // Emit reduction for array subscript or single variable.
  emitReductionCombiner(CGF, ReductionOp);
}
5536}

5538void CGOpenMPRuntime::emitReduction(CodeGenFunction &CGF, SourceLocation Loc,
                                  ArrayRef<const Expr *> Privates,
                                  ArrayRef<const Expr *> LHSExprs,
                                  ArrayRef<const Expr *> RHSExprs,
                                  ArrayRef<const Expr *> ReductionOps,
                                  ReductionOptionsTy Options) {
if (!CGF.HaveInsertPoint())
  return;

bool WithNowait = Options.WithNowait;
bool SimpleReduction = Options.SimpleReduction;

// Next code should be emitted for reduction:
//
// static kmp_critical_name lock = { 0 };
//
// void reduce_func(void *lhs[<n>], void *rhs[<n>]) {
//  *(Type0*)lhs[0] = ReductionOperation0(*(Type0*)lhs[0], *(Type0*)rhs[0]);
//  ...
//  *(Type<n>-1*)lhs[<n>-1] = ReductionOperation<n>-1(*(Type<n>-1*)lhs[<n>-1],
//  *(Type<n>-1*)rhs[<n>-1]);
// }
//
// ...
// void *RedList[<n>] = {&<RHSExprs>[0], ..., &<RHSExprs>[<n>-1]};
// switch (__kmpc_reduce{_nowait}(<loc>, <gtid>, <n>, sizeof(RedList),
// RedList, reduce_func, &<lock>)) {
// case 1:
//  ...
//  <LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]);
//  ...
// __kmpc_end_reduce{_nowait}(<loc>, <gtid>, &<lock>);
// break;
// case 2:
//  ...
//  Atomic(<LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]));
//  ...
// [__kmpc_end_reduce(<loc>, <gtid>, &<lock>);]
// break;
// default:;
// }
//
// if SimpleReduction is true, only the next code is generated:
//  ...
//  <LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]);
//  ...

ASTContext &C = CGM.getContext();

if (SimpleReduction) {
  CodeGenFunction::RunCleanupsScope Scope(CGF);
  auto IPriv = Privates.begin();
  auto ILHS = LHSExprs.begin();
  auto IRHS = RHSExprs.begin();
  for (const Expr *E : ReductionOps) {
    emitSingleReductionCombiner(CGF, E, *IPriv, cast<DeclRefExpr>(*ILHS),
                                cast<DeclRefExpr>(*IRHS));
    ++IPriv;
    ++ILHS;
    ++IRHS;
  }
  return;
}

// 1. Build a list of reduction variables.
// void *RedList[<n>] = {<ReductionVars>[0], ..., <ReductionVars>[<n>-1]};
auto Size = RHSExprs.size();
for (const Expr *E : Privates) {
  if (E->getType()->isVariablyModifiedType())
    // Reserve place for array size.
    ++Size;
}
llvm::APInt ArraySize(/*unsigned int numBits=*/32, Size);
QualType ReductionArrayTy =
    C.getConstantArrayType(C.VoidPtrTy, ArraySize, nullptr, ArrayType::Normal,
                           /*IndexTypeQuals=*/0);
Address ReductionList =
    CGF.CreateMemTemp(ReductionArrayTy, ".omp.reduction.red_list");
auto IPriv = Privates.begin();
unsigned Idx = 0;
for (unsigned I = 0, E = RHSExprs.size(); I < E; ++I, ++IPriv, ++Idx) {
  Address Elem = CGF.Builder.CreateConstArrayGEP(ReductionList, Idx);
  CGF.Builder.CreateStore(
      CGF.Builder.CreatePointerBitCastOrAddrSpaceCast(
          CGF.EmitLValue(RHSExprs[I]).getPointer(CGF), CGF.VoidPtrTy),
      Elem);
  if ((*IPriv)->getType()->isVariablyModifiedType()) {
    // Store array size.
    ++Idx;
    Elem = CGF.Builder.CreateConstArrayGEP(ReductionList, Idx);
    llvm::Value *Size = CGF.Builder.CreateIntCast(
        CGF.getVLASize(
               CGF.getContext().getAsVariableArrayType((*IPriv)->getType()))
            .NumElts,
        CGF.SizeTy, /*isSigned=*/false);
    CGF.Builder.CreateStore(CGF.Builder.CreateIntToPtr(Size, CGF.VoidPtrTy),
                            Elem);
  }
}

// 2. Emit reduce_func().
llvm::Function *ReductionFn = emitReductionFunction(
    Loc, CGF.ConvertTypeForMem(ReductionArrayTy)->getPointerTo(), Privates,
    LHSExprs, RHSExprs, ReductionOps);

// 3. Create static kmp_critical_name lock = { 0 };
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>);
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>
    ThreadId,                              // i32 <gtid>
    CGF.Builder.getInt32(RHSExprs.size()), // i32 <n>
    ReductionArrayTySize,                  // size_type sizeof(RedList)
    RL,                                    // void *RedList
    ReductionFn, // void (*) (void *, void *) <reduce_func>
    Lock         // kmp_critical_name *&<lock>
};
llvm::Value *Res = CGF.EmitRuntimeCall(
    OMPBuilder.getOrCreateRuntimeFunction(
        CGM.getModule(),
        WithNowait ? OMPRTL___kmpc_reduce_nowait : OMPRTL___kmpc_reduce),
    Args);

// 5. Build switch(res)
llvm::BasicBlock *DefaultBB = CGF.createBasicBlock(".omp.reduction.default");
llvm::SwitchInst *SwInst =
    CGF.Builder.CreateSwitch(Res, DefaultBB, /*NumCases=*/2);

// 6. Build case 1:
//  ...
//  <LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]);
//  ...
// __kmpc_end_reduce{_nowait}(<loc>, <gtid>, &<lock>);
// break;
llvm::BasicBlock *Case1BB = CGF.createBasicBlock(".omp.reduction.case1");
SwInst->addCase(CGF.Builder.getInt32(1), Case1BB);
CGF.EmitBlock(Case1BB);

// Add emission of __kmpc_end_reduce{_nowait}(<loc>, <gtid>, &<lock>);
llvm::Value *EndArgs[] = {
    IdentTLoc, // ident_t *<loc>
    ThreadId,  // i32 <gtid>
    Lock       // kmp_critical_name *&<lock>
};
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 (const Expr *E : ReductionOps) {
    RT.emitSingleReductionCombiner(CGF, E, *IPriv, cast<DeclRefExpr>(*ILHS),
                                   cast<DeclRefExpr>(*IRHS));
    ++IPriv;
    ++ILHS;
    ++IRHS;
  }
};
RegionCodeGenTy RCG(CodeGen);
CommonActionTy Action(
    nullptr, llvm::None,
    OMPBuilder.getOrCreateRuntimeFunction(
        CGM.getModule(), WithNowait ? OMPRTL___kmpc_end_reduce_nowait
                                    : OMPRTL___kmpc_end_reduce),
    EndArgs);
RCG.setAction(Action);
RCG(CGF);

CGF.EmitBranch(DefaultBB);

// 7. Build case 2:
//  ...
//  Atomic(<LHSExprs>[i] = RedOp<i>(*<LHSExprs>[i], *<RHSExprs>[i]));
//  ...
// break;
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 ILHS = LHSExprs.begin();
  auto IRHS = RHSExprs.begin();
  auto IPriv = Privates.begin();
  for (const Expr *E : ReductionOps) {
    const Expr *XExpr = nullptr;
    const Expr *EExpr = nullptr;
    const Expr *UpExpr = nullptr;
    BinaryOperatorKind BO = BO_Comma;
    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.
    const Expr *RHSExpr = UpExpr;
    if (RHSExpr) {
      // Analyze RHS part of the whole expression.
      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 (const auto *BORHS =
              dyn_cast<BinaryOperator>(RHSExpr->IgnoreParenImpCasts())) {
        EExpr = BORHS->getRHS();
        BO = BORHS->getOpcode();
      }
    }
    if (XExpr) {
      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) {
        LValue X = CGF.EmitLValue(XExpr);
        RValue E;
        if (EExpr)
          E = CGF.EmitAnyExpr(EExpr);
        CGF.EmitOMPAtomicSimpleUpdateExpr(
            X, E, BO, /*IsXLHSInRHSPart=*/true,
            llvm::AtomicOrdering::Monotonic, Loc,
            [&CGF, UpExpr, VD, Loc](RValue XRValue) {
              CodeGenFunction::OMPPrivateScope PrivateScope(CGF);
              PrivateScope.addPrivate(
                  VD, [&CGF, VD, XRValue, Loc]() {
                    Address LHSTemp = CGF.CreateMemTemp(VD->getType());
                    CGF.emitOMPSimpleStore(
                        CGF.MakeAddrLValue(LHSTemp, VD->getType()), XRValue,
                        VD->getType().getNonReferenceType(), Loc);
                    return LHSTemp;
                  });
              (void)PrivateScope.Privatize();
              return CGF.EmitAnyExpr(UpExpr);
            });
      };
      if ((*IPriv)->getType()->isArrayType()) {
        // Emit atomic reduction for array section.
        const auto *RHSVar =
            cast<VarDecl>(cast<DeclRefExpr>(*IRHS)->getDecl());
        EmitOMPAggregateReduction(CGF, (*IPriv)->getType(), VD, RHSVar,
                                  AtomicRedGen, XExpr, EExpr, UpExpr);
      } 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 *) {
        CGOpenMPRuntime &RT = CGF.CGM.getOpenMPRuntime();
        std::string Name = RT.getName({"atomic_reduction"});
        RT.emitCriticalRegion(
            CGF, Name,
            [=](CodeGenFunction &CGF, PrePostActionTy &Action) {
              Action.Enter(CGF);
              emitReductionCombiner(CGF, E);
            },
            Loc);
      };
      if ((*IPriv)->getType()->isArrayType()) {
        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 {
        CritRedGen(CGF, nullptr, nullptr, nullptr);
      }
    }
    ++ILHS;
    ++IRHS;
    ++IPriv;
  }
};
RegionCodeGenTy AtomicRCG(AtomicCodeGen);
if (!WithNowait) {
  // Add emission of __kmpc_end_reduce(<loc>, <gtid>, &<lock>);
  llvm::Value *EndArgs[] = {
      IdentTLoc, // ident_t *<loc>
      ThreadId,  // i32 <gtid>
      Lock       // kmp_critical_name *&<lock>
  };
  CommonActionTy Action(nullptr, llvm::None,
                        OMPBuilder.getOrCreateRuntimeFunction(
                            CGM.getModule(), OMPRTL___kmpc_end_reduce),
                        EndArgs);
  AtomicRCG.setAction(Action);
  AtomicRCG(CGF);
} else {
  AtomicRCG(CGF);
}

CGF.EmitBranch(DefaultBB);
CGF.EmitBlock(DefaultBB, /*IsFinished=*/true);
5843}

5845/// Generates unique name for artificial threadprivate variables.
5846/// Format is: <Prefix> "." <Decl_mangled_name> "_" "<Decl_start_loc_raw_enc>"
5847static std::string generateUniqueName(CodeGenModule &CGM, StringRef Prefix,
                                    const Expr *Ref) {
SmallString<256> Buffer;
llvm::raw_svector_ostream Out(Buffer);
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()->getBeginLoc().getRawEncoding();
return std::string(Out.str());
5861}

5863/// Emits reduction initializer function:
5864/// \code
5865/// void @.red_init(void* %arg, void* %orig) {
5866/// %0 = bitcast void* %arg to <type>*
5867/// store <type> <init>, <type>* %0
5868/// ret void
5869/// }
5870/// \endcode
5871static llvm::Value *emitReduceInitFunction(CodeGenModule &CGM,
                                         SourceLocation Loc,
                                         ReductionCodeGen &RCG, unsigned N) {
ASTContext &C = CGM.getContext();
QualType VoidPtrTy = C.VoidPtrTy;
VoidPtrTy.addRestrict();
FunctionArgList Args;
ImplicitParamDecl Param(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, VoidPtrTy,
                        ImplicitParamDecl::Other);
ImplicitParamDecl ParamOrig(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, VoidPtrTy,
                            ImplicitParamDecl::Other);
Args.emplace_back(&Param);
Args.emplace_back(&ParamOrig);
const auto &FnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
std::string Name = CGM.getOpenMPRuntime().getName({"red_init", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                  Name, &CGM.getModule());
CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
Address PrivateAddr = CGF.EmitLoadOfPointer(
    CGF.GetAddrOfLocalVar(&Param),
    C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
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(CGM, "reduction_size", RCG.getRefExpr(N)));
  Size = CGF.EmitLoadOfScalar(SizeAddr, /*Volatile=*/false,
                              CGM.getContext().getSizeType(), Loc);
}
RCG.emitAggregateType(CGF, N, Size);
LValue OrigLVal;
// If initializer uses initializer from declare reduction construct, emit a
// pointer to the address of the original reduction item (reuired by reduction
// initializer)
if (RCG.usesReductionInitializer(N)) {
  Address SharedAddr = CGF.GetAddrOfLocalVar(&ParamOrig);
  SharedAddr = CGF.EmitLoadOfPointer(
      SharedAddr,
      CGM.getContext().VoidPtrTy.castAs<PointerType>()->getTypePtr());
  OrigLVal = CGF.MakeAddrLValue(SharedAddr, CGM.getContext().VoidPtrTy);
} else {
  OrigLVal = CGF.MakeNaturalAlignAddrLValue(
      llvm::ConstantPointerNull::get(CGM.VoidPtrTy),
      CGM.getContext().VoidPtrTy);
}
// Emit the initializer:
// %0 = bitcast void* %arg to <type>*
// store <type> <init>, <type>* %0
RCG.emitInitialization(CGF, N, PrivateAddr, OrigLVal,
                       [](CodeGenFunction &) { return false; });
CGF.FinishFunction();
return Fn;
5930}

5932/// Emits reduction combiner function:
5933/// \code
5934/// void @.red_comb(void* %arg0, void* %arg1) {
5935/// %lhs = bitcast void* %arg0 to <type>*
5936/// %rhs = bitcast void* %arg1 to <type>*
5937/// %2 = <ReductionOp>(<type>* %lhs, <type>* %rhs)
5938/// store <type> %2, <type>* %lhs
5939/// ret void
5940/// }
5941/// \endcode
5942static llvm::Value *emitReduceCombFunction(CodeGenModule &CGM,
                                         SourceLocation Loc,
                                         ReductionCodeGen &RCG, unsigned N,
                                         const Expr *ReductionOp,
                                         const Expr *LHS, const Expr *RHS,
                                         const Expr *PrivateRef) {
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, /*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);
const auto &FnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
std::string Name = CGM.getOpenMPRuntime().getName({"red_comb", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                  Name, &CGM.getModule());
CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
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(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]() {
  // Pull out the pointer to the variable.
  Address PtrAddr = CGF.EmitLoadOfPointer(
      CGF.GetAddrOfLocalVar(&ParamInOut),
      C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
  return CGF.Builder.CreateElementBitCast(
      PtrAddr, CGF.ConvertTypeForMem(LHSVD->getType()));
});
PrivateScope.addPrivate(RHSVD, [&C, &CGF, &ParamIn, RHSVD]() {
  // Pull out the pointer to the variable.
  Address PtrAddr = CGF.EmitLoadOfPointer(
      CGF.GetAddrOfLocalVar(&ParamIn),
      C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
  return CGF.Builder.CreateElementBitCast(
      PtrAddr, CGF.ConvertTypeForMem(RHSVD->getType()));
});
PrivateScope.Privatize();
// Emit the combiner body:
// %2 = <ReductionOp>(<type> *%lhs, <type> *%rhs)
// store <type> %2, <type>* %lhs
CGM.getOpenMPRuntime().emitSingleReductionCombiner(
    CGF, ReductionOp, PrivateRef, cast<DeclRefExpr>(LHS),
    cast<DeclRefExpr>(RHS));
CGF.FinishFunction();
return Fn;
6008}

6010/// Emits reduction finalizer function:
6011/// \code
6012/// void @.red_fini(void* %arg) {
6013/// %0 = bitcast void* %arg to <type>*
6014/// <destroy>(<type>* %0)
6015/// ret void
6016/// }
6017/// \endcode
6018static llvm::Value *emitReduceFiniFunction(CodeGenModule &CGM,
                                         SourceLocation Loc,
                                         ReductionCodeGen &RCG, unsigned N) {
if (!RCG.needCleanups(N))
  return nullptr;
ASTContext &C = CGM.getContext();
FunctionArgList Args;
ImplicitParamDecl Param(C, /*DC=*/nullptr, Loc, /*Id=*/nullptr, C.VoidPtrTy,
                        ImplicitParamDecl::Other);
Args.emplace_back(&Param);
const auto &FnInfo =
    CGM.getTypes().arrangeBuiltinFunctionDeclaration(C.VoidTy, Args);
llvm::FunctionType *FnTy = CGM.getTypes().GetFunctionType(FnInfo);
std::string Name = CGM.getOpenMPRuntime().getName({"red_fini", ""});
auto *Fn = llvm::Function::Create(FnTy, llvm::GlobalValue::InternalLinkage,
                                  Name, &CGM.getModule());
CGM.SetInternalFunctionAttributes(GlobalDecl(), Fn, FnInfo);
Fn->setDoesNotRecurse();
CodeGenFunction CGF(CGM);
CGF.StartFunction(GlobalDecl(), C.VoidTy, Fn, FnInfo, Args, Loc, Loc);
Address PrivateAddr = CGF.EmitLoadOfPointer(
    CGF.GetAddrOfLocalVar(&Param),
    C.getPointerType(C.VoidPtrTy).castAs<PointerType>());
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(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:
// <destroy>(<type>* %0)
RCG.emitCleanups(CGF, N, PrivateAddr);
CGF.FinishFunction(Loc);
return Fn;
6057}

6059llvm::Value *CGOpenMPRuntime::emitTaskReductionInit(
  CodeGenFunction &CGF, SourceLocation Loc, ArrayRef<const Expr *> LHSExprs,
  ArrayRef<const Expr *> RHSExprs, const OMPTaskDataTy &Data) {
if (!CGF.HaveInsertPoint() || Data.ReductionVars.empty())
  return nullptr;

// Build typedef struct:
// kmp_taskred_input {
//   void *reduce_shar; // shared reduction item
//   void *reduce_orig; // original reduction item used for initialization
//   size_t reduce_size; // size of data item
//   void *reduce_init; // data initialization routine
//   void *reduce_fini; // data finalization routine
//   void *reduce_comb; // data combiner routine
//   kmp_task_red_flags_t flags; // flags for additional info from compiler
// } kmp_taskred_input_t;
ASTContext &C = CGM.getContext();
RecordDe